Scikit Learn Docs PDF

scikit-learn user guide
Release 0.20.3
scikit-learn developers
Mar 01, 2019

CONTENTS
1 Welcome to scikit-learn 1
1.1 Installing scikit-learn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Frequently Asked Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Related Projects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 About us . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.6 Who is using scikit-learn? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.7 Release History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.8 Version 0.20.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.9 Version 0.20.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.10 Version 0.20.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.11 Version 0.20.0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.12 Version 0.19.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1.13 Version 0.19.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1.14 Version 0.19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
1.15 Previous Releases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2 scikit-learn Tutorials 129

2.1 An introduction to machine learning with scikit-learn . . . . . . . . . . . . . . . . . . . . . . . . . . 129
2.2 A tutorial on statistical-learning for scientific data processing . . . . . . . . . . . . . . . . . . . . . 135
2.3 Working With Text Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
2.4 Choosing the right estimator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
2.5 External Resources, Videos and Talks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
3 User Guide 173

3.1 Supervised learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
3.2 Unsupervised learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
3.3 Model selection and evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402
3.4 Dataset transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 535
3.5 Dataset loading utilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 583
3.6 Computing with scikit-learn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 608
4 Glossary of Common Terms and API Elements 623

4.1 General Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623
4.2 Class APIs and Estimator Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
4.3 Target Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 634
4.4 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 636
4.5 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 638
4.6 Attributes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
4.7 Data and sample properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 642
i
5 Examples 643
5.1 Miscellaneous examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643
5.2 Examples based on real world datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 678
5.3 Biclustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 739
5.4 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 751
5.5 Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 769
5.6 Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 784
5.7 Pipelines and composite estimators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 866
5.8 Covariance estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 901
5.9 Cross decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 916
5.10 Dataset examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 920
5.11 Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 929
5.12 Ensemble methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 975
5.13 Tutorial exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
5.14 Feature Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1038
5.15 Gaussian Process for Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1048
5.16 Generalized Linear Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1077
5.17 Manifold learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1164
5.18 Gaussian Mixture Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1193
5.19 Model Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1210
5.20 Multioutput methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1258
5.21 Nearest Neighbors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1261
5.22 Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1283
5.23 Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1296
5.24 Semi Supervised Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1322
5.25 Support Vector Machines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1334
5.26 Working with text documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1365
5.27 Decision Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1382
6 API Reference 1393

6.1 sklearn.base: Base classes and utility functions . . . . . . . . . . . . . . . . . . . . . . . . . . 1393
6.2 sklearn.calibration: Probability Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . 1400
6.3 sklearn.cluster: Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1404
6.4 sklearn.cluster.bicluster: Biclustering . . . . . . . . . . . . . . . . . . . . . . . . . . . 1443
6.5 sklearn.compose: Composite Estimators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1450
6.6 sklearn.covariance: Covariance Estimators . . . . . . . . . . . . . . . . . . . . . . . . . . . 1457
6.7 sklearn.cross_decomposition: Cross decomposition . . . . . . . . . . . . . . . . . . . . 1488
6.8 sklearn.datasets: Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1501
6.9 sklearn.decomposition: Matrix Decomposition . . . . . . . . . . . . . . . . . . . . . . . . 1546
6.10 sklearn.discriminant_analysis: Discriminant Analysis . . . . . . . . . . . . . . . . . . 1598
6.11 sklearn.dummy: Dummy estimators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1606
6.12 sklearn.ensemble: Ensemble Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1611
6.13 sklearn.exceptions: Exceptions and warnings . . . . . . . . . . . . . . . . . . . . . . . . . 1641
6.14 sklearn.feature_extraction: Feature Extraction . . . . . . . . . . . . . . . . . . . . . . . 1646
6.15 sklearn.feature_selection: Feature Selection . . . . . . . . . . . . . . . . . . . . . . . . 1673
6.16 sklearn.gaussian_process: Gaussian Processes . . . . . . . . . . . . . . . . . . . . . . . . 1708
6.17 sklearn.isotonic: Isotonic regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1747
6.18 sklearn.impute: Impute . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1752
6.19 sklearn.kernel_approximation Kernel Approximation . . . . . . . . . . . . . . . . . . . 1756
6.20 sklearn.kernel_ridge Kernel Ridge Regression . . . . . . . . . . . . . . . . . . . . . . . . 1766
6.21 sklearn.linear_model: Generalized Linear Models . . . . . . . . . . . . . . . . . . . . . . . 1769
6.22 sklearn.manifold: Manifold Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1865
6.23 sklearn.metrics: Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1884
6.24 sklearn.mixture: Gaussian Mixture Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 1951
ii
6.25 sklearn.model_selection: Model Selection . . . . . . . . . . . . . . . . . . . . . . . . . . 1962
6.26 sklearn.multiclass: Multiclass and multilabel classification . . . . . . . . . . . . . . . . . . 2016
6.27 sklearn.multioutput: Multioutput regression and classification . . . . . . . . . . . . . . . . 2024
6.28 sklearn.naive_bayes: Naive Bayes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2034
6.29 sklearn.neighbors: Nearest Neighbors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2048
6.30 sklearn.neural_network: Neural network models . . . . . . . . . . . . . . . . . . . . . . . 2095
6.31 sklearn.pipeline: Pipeline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2107
6.32 sklearn.preprocessing: Preprocessing and Normalization . . . . . . . . . . . . . . . . . . . 2116
6.33 sklearn.random_projection: Random projection . . . . . . . . . . . . . . . . . . . . . . . 2171
6.34 sklearn.semi_supervised Semi-Supervised Learning . . . . . . . . . . . . . . . . . . . . . 2177
6.35 sklearn.svm: Support Vector Machines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2183
6.36 sklearn.tree: Decision Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2214
6.37 sklearn.utils: Utilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2238
6.38 Recently deprecated . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2265
7 Developer’s Guide 2293

7.1 Contributing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2293
7.2 Developers’ Tips and Tricks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2314
7.3 Utilities for Developers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2318
7.4 How to optimize for speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2321
7.5 Advanced installation instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2327
7.6 Maintainer / core-developer information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2332
Bibliography 2335
Index 2343
iii
iv
CHAPTER
ONE
WELCOME TO SCIKIT-LEARN
1.1 Installing scikit-learn
Note: If you wish to contribute to the project, it’s recommended you install the latest development version.
1.1.1 Installing the latest release
Scikit-learn requires:
• Python (>= 2.7 or >= 3.4),
• NumPy (>= 1.8.2),
• SciPy (>= 0.13.3).
Warning: Scikit-learn 0.20 is the last version to support Python 2.7 and Python 3.4. Scikit-learn 0.21 will require
Python 3.5 or newer.
If you already have a working installation of numpy and scipy, the easiest way to install scikit-learn is using pip
pip install -U scikit-learn
or conda:
conda install scikit-learn
If you have not installed NumPy or SciPy yet, you can also install these using conda or pip. When using pip, please
ensure that binary wheels are used, and NumPy and SciPy are not recompiled from source, which can happen when
using particular configurations of operating system and hardware (such as Linux on a Raspberry Pi). Building numpy
and scipy from source can be complex (especially on Windows) and requires careful configuration to ensure that they
link against an optimized implementation of linear algebra routines. Instead, use a third-party distribution as described
below.
If you must install scikit-learn and its dependencies with pip, you can install it as scikit-learn[alldeps]. The
most common use case for this is in a requirements.txt file used as part of an automated build process for a
PaaS application or a Docker image. This option is not intended for manual installation from the command line.
1
scikit-learn user guide, Release 0.20.3
Note: For installing on PyPy, PyPy3-v5.10+, Numpy 1.14.0+, and scipy 1.1.0+ are required.
For installation instructions for more distributions see other distributions. For compiling the development version
from source, or building the package if no distribution is available for your architecture, see the Advanced installation
instructions.
1.1.2 Third-party Distributions
If you don’t already have a python installation with numpy and scipy, we recommend to install either via your package
manager or via a python bundle. These come with numpy, scipy, scikit-learn, matplotlib and many other helpful
scientific and data processing libraries.
Available options are:
Canopy and Anaconda for all supported platforms
Canopy and Anaconda both ship a recent version of scikit-learn, in addition to a large set of scientific python library
for Windows, Mac OSX and Linux.
Anaconda offers scikit-learn as part of its free distribution.
Warning: To upgrade or uninstall scikit-learn installed with Anaconda or conda you should not use the pip
command. Instead:
To upgrade scikit-learn:
conda update scikit-learn
To uninstall scikit-learn:
conda remove scikit-learn
Upgrading with pip install -U scikit-learn or uninstalling pip uninstall scikit-learn is

likely fail to properly remove files installed by the conda command.
pip upgrade and uninstall operations only work on packages installed via pip install.
WinPython for Windows
The WinPython project distributes scikit-learn as an additional plugin.
1.2 Frequently Asked Questions
Here we try to give some answers to questions that regularly pop up on the mailing list.
1.2.1 What is the project name (a lot of people get it wrong)?
scikit-learn, but not scikit or SciKit nor sci-kit learn. Also not scikits.learn or scikits-learn, which were previously
used.
2 Chapter 1. Welcome to scikit-learn

1.2.2 How do you pronounce the project name?
sy-kit learn. sci stands for science!
1.2.3 Why scikit?
There are multiple scikits, which are scientific toolboxes built around SciPy. You can find a list at https://scikits.
appspot.com/scikits. Apart from scikit-learn, another popular one is scikit-image.
1.2.4 How can I contribute to scikit-learn?
See Contributing. Before wanting to add a new algorithm, which is usually a major and lengthy undertaking, it is
recommended to start with known issues. Please do not contact the contributors of scikit-learn directly regarding
contributing to scikit-learn.
1.2.5 What’s the best way to get help on scikit-learn usage?
For general machine learning questions, please use Cross Validated with the [machine-learning] tag.
For scikit-learn usage questions, please use Stack Overflow with the [scikit-learn] and [python] tags. You
can alternatively use the mailing list.
Please make sure to include a minimal reproduction code snippet (ideally shorter than 10 lines) that highlights your
problem on a toy dataset (for instance from sklearn.datasets or randomly generated with functions of numpy.
random with a fixed random seed). Please remove any line of code that is not necessary to reproduce your problem.
The problem should be reproducible by simply copy-pasting your code snippet in a Python shell with scikit-learn
installed. Do not forget to include the import statements.
More guidance to write good reproduction code snippets can be found at:
https://stackoverflow.com/help/mcve
If your problem raises an exception that you do not understand (even after googling it), please make sure to include
the full traceback that you obtain when running the reproduction script.
For bug reports or feature requests, please make use of the issue tracker on GitHub.
There is also a scikit-learn Gitter channel where some users and developers might be found.
Please do not email any authors directly to ask for assistance, report bugs, or for any other issue related to
scikit-learn.
1.2.6 How should I save, export or deploy estimators for production?
See Model persistence.
1.2.7 How can I create a bunch object?
Don’t make a bunch object! They are not part of the scikit-learn API. Bunch objects are just a way to package some
numpy arrays. As a scikit-learn user you only ever need numpy arrays to feed your model with data.
For instance to train a classifier, all you need is a 2D array X for the input variables and a 1D array y for the target
variables. The array X holds the features as columns and samples as rows . The array y contains integer values to
encode the class membership of each sample in X.
1.2. Frequently Asked Questions 3

1.2.8 How can I load my own datasets into a format usable by scikit-learn?
Generally, scikit-learn works on any numeric data stored as numpy arrays or scipy sparse matrices. Other types that
are convertible to numeric arrays such as pandas DataFrame are also acceptable.
For more information on loading your data files into these usable data structures, please refer to loading external
datasets.
1.2.9 What are the inclusion criteria for new algorithms ?
We only consider well-established algorithms for inclusion. A rule of thumb is at least 3 years since publication, 200+
citations and wide use and usefulness. A technique that provides a clear-cut improvement (e.g. an enhanced data
structure or a more efficient approximation technique) on a widely-used method will also be considered for inclusion.
From the algorithms or techniques that meet the above criteria, only those which fit well within the current API of
scikit-learn, that is a fit, predict/transform interface and ordinarily having input/output that is a numpy array
or sparse matrix, are accepted.
The contributor should support the importance of the proposed addition with research papers and/or implementations
in other similar packages, demonstrate its usefulness via common use-cases/applications and corroborate performance
improvements, if any, with benchmarks and/or plots. It is expected that the proposed algorithm should outperform the
methods that are already implemented in scikit-learn at least in some areas.
Also note that your implementation need not be in scikit-learn to be used together with scikit-learn tools. You can
implement your favorite algorithm in a scikit-learn compatible way, upload it to GitHub and let us know. We will be
happy to list it under Related Projects. If you already have a package on GitHub following the scikit-learn API, you
may also be interested to look at scikit-learn-contrib.
1.2.10 Why are you so selective on what algorithms you include in scikit-learn?
Code is maintenance cost, and we need to balance the amount of code we have with the size of the team (and add to
this the fact that complexity scales non linearly with the number of features). The package relies on core developers
using their free time to fix bugs, maintain code and review contributions. Any algorithm that is added needs future
attention by the developers, at which point the original author might long have lost interest. See also What are the
inclusion criteria for new algorithms ?. For a great read about long-term maintenance issues in open-source software,
look at the Executive Summary of Roads and Bridges
1.2.11 Why did you remove HMMs from scikit-learn?
See Will you add graphical models or sequence prediction to scikit-learn?.
1.2.12 Will you add graphical models or sequence prediction to scikit-learn?
Not in the foreseeable future. scikit-learn tries to provide a unified API for the basic tasks in machine learning, with
pipelines and meta-algorithms like grid search to tie everything together. The required concepts, APIs, algorithms
and expertise required for structured learning are different from what scikit-learn has to offer. If we started doing
arbitrary structured learning, we’d need to redesign the whole package and the project would likely collapse under its
own weight.
There are two project with API similar to scikit-learn that do structured prediction:
• pystruct handles general structured learning (focuses on SSVMs on arbitrary graph structures with approximate
inference; defines the notion of sample as an instance of the graph structure)

• seqlearn handles sequences only (focuses on exact inference; has HMMs, but mostly for the sake of complete-
ness; treats a feature vector as a sample and uses an offset encoding for the dependencies between feature
vectors)
1.2.13 Will you add GPU support?
No, or at least not in the near future. The main reason is that GPU support will introduce many software dependencies
and introduce platform specific issues. scikit-learn is designed to be easy to install on a wide variety of platforms.
Outside of neural networks, GPUs don’t play a large role in machine learning today, and much larger gains in speed
can often be achieved by a careful choice of algorithms.
1.2.14 Do you support PyPy?
In case you didn’t know, PyPy is an alternative Python implementation with a built-in just-in-time compiler. Experi-
mental support for PyPy3-v5.10+ has been added, which requires Numpy 1.14.0+, and scipy 1.1.0+.
1.2.15 How do I deal with string data (or trees, graphs. . . )?
scikit-learn estimators assume you’ll feed them real-valued feature vectors. This assumption is hard-coded in pretty
much all of the library. However, you can feed non-numerical inputs to estimators in several ways.
If you have text documents, you can use a term frequency features; see Text feature extraction for the built-in text
vectorizers. For more general feature extraction from any kind of data, see Loading features from dicts and Feature
hashing.
Another common case is when you have non-numerical data and a custom distance (or similarity) metric on these data.
Examples include strings with edit distance (aka. Levenshtein distance; e.g., DNA or RNA sequences). These can be
encoded as numbers, but doing so is painful and error-prone. Working with distance metrics on arbitrary data can be
done in two ways.
Firstly, many estimators take precomputed distance/similarity matrices, so if the dataset is not too large, you can
compute distances for all pairs of inputs. If the dataset is large, you can use feature vectors with only one “feature”,
which is an index into a separate data structure, and supply a custom metric function that looks up the actual data in
this data structure. E.g., to use DBSCAN with Levenshtein distances:
>>> from leven import levenshtein

>>> import numpy as np
>>> from sklearn.cluster import dbscan
>>> data = ["ACCTCCTAGAAG", "ACCTACTAGAAGTT", "GAATATTAGGCCGA"]
>>> def lev_metric(x, y):
... i, j = int(x[0]), int(y[0]) # extract indices
... return levenshtein(data[i], data[j])
...
>>> X = np.arange(len(data)).reshape(-1, 1)
>>> X
array([[0],
[1],
[2]])
>>> # We need to specify algoritum='brute' as the default assumes
>>> # a continuous feature space.
>>> dbscan(X, metric=lev_metric, eps=5, min_samples=2, algorithm='brute')
...
([0, 1], array([ 0, 0, -1]))

(This uses the third-party edit distance package leven.)

Similar tricks can be used, with some care, for tree kernels, graph kernels, etc.
1.2.16 Why do I sometime get a crash/freeze with n_jobs > 1 under OSX or Linux?
Several scikit-learn tools such as GridSearchCV and cross_val_score rely internally on Python’s multipro-
cessing module to parallelize execution onto several Python processes by passing n_jobs > 1 as argument.
The problem is that Python multiprocessing does a fork system call without following it with an exec system
call for performance reasons. Many libraries like (some versions of) Accelerate / vecLib under OSX, (some versions
of) MKL, the OpenMP runtime of GCC, nvidia’s Cuda (and probably many others), manage their own internal thread
pool. Upon a call to fork, the thread pool state in the child process is corrupted: the thread pool believes it has many
threads while only the main thread state has been forked. It is possible to change the libraries to make them detect
when a fork happens and reinitialize the thread pool in that case: we did that for OpenBLAS (merged upstream in
master since 0.2.10) and we contributed a patch to GCC’s OpenMP runtime (not yet reviewed).
But in the end the real culprit is Python’s multiprocessing that does fork without exec to reduce the overhead
of starting and using new Python processes for parallel computing. Unfortunately this is a violation of the POSIX
standard and therefore some software editors like Apple refuse to consider the lack of fork-safety in Accelerate /
vecLib as a bug.
In Python 3.4+ it is now possible to configure multiprocessing to use the ‘forkserver’ or ‘spawn’ start methods
(instead of the default ‘fork’) to manage the process pools. To work around this issue when using scikit-learn, you
can set the JOBLIB_START_METHOD environment variable to ‘forkserver’. However the user should be aware that
using the ‘forkserver’ method prevents joblib.Parallel to call function interactively defined in a shell session.
If you have custom code that uses multiprocessing directly instead of using it via joblib you can enable the
‘forkserver’ mode globally for your program: Insert the following instructions in your main script:
import multiprocessing
# other imports, custom code, load data, define model...
if __name__ == '__main__':
multiprocessing.set_start_method('forkserver')
# call scikit-learn utils with n_jobs > 1 here
You can find more default on the new start methods in the multiprocessing documentation.
1.2.17 Why does my job use more cores than specified with n_jobs under OSX or
Linux?
This happens when vectorized numpy operations are handled by libraries such as MKL or OpenBLAS.
While scikit-learn adheres to the limit set by n_jobs, numpy operations vectorized using MKL (or OpenBLAS) will
make use of multiple threads within each scikit-learn job (thread or process).
The number of threads used by the BLAS library can be set via an environment variable. For example, to set the
maximum number of threads to some integer value N, the following environment variables should be set:
• For MKL: export MKL_NUM_THREADS=N
• For OpenBLAS: export OPENBLAS_NUM_THREADS=N

1.2.18 Why is there no support for deep or reinforcement learning / Will there be
support for deep or reinforcement learning in scikit-learn?
Deep learning and reinforcement learning both require a rich vocabulary to define an architecture, with deep learning
additionally requiring GPUs for efficient computing. However, neither of these fit within the design constraints of
scikit-learn; as a result, deep learning and reinforcement learning are currently out of scope for what scikit-learn seeks
to achieve.
You can find more information about addition of gpu support at Will you add GPU support?.
1.2.19 Why is my pull request not getting any attention?
The scikit-learn review process takes a significant amount of time, and contributors should not be discouraged by a
lack of activity or review on their pull request. We care a lot about getting things right the first time, as maintenance
and later change comes at a high cost. We rarely release any “experimental” code, so all of our contributions will be
subject to high use immediately and should be of the highest quality possible initially.
Beyond that, scikit-learn is limited in its reviewing bandwidth; many of the reviewers and core developers are working
on scikit-learn on their own time. If a review of your pull request comes slowly, it is likely because the reviewers are
busy. We ask for your understanding and request that you not close your pull request or discontinue your work solely
because of this reason.
1.2.20 How do I set a random_state for an entire execution?
For testing and replicability, it is often important to have the entire execution controlled by a single seed for the pseudo-
random number generator used in algorithms that have a randomized component. Scikit-learn does not use its own
global random state; whenever a RandomState instance or an integer random seed is not provided as an argument, it
relies on the numpy global random state, which can be set using numpy.random.seed. For example, to set an
execution’s numpy global random state to 42, one could execute the following in his or her script:
import numpy as np
np.random.seed(42)
However, a global random state is prone to modification by other code during execution. Thus, the only way to ensure
replicability is to pass RandomState instances everywhere and ensure that both estimators and cross-validation
splitters have their random_state parameter set.
1.2.21 Why do categorical variables need preprocessing in scikit-learn, compared

to other tools?
Most of scikit-learn assumes data is in NumPy arrays or SciPy sparse matrices of a single numeric dtype. These do
not explicitly represent categorical variables at present. Thus, unlike R’s data.frames or pandas.DataFrame, we require
explicit conversion of categorical features to numeric values, as discussed in Encoding categorical features. See also
Column Transformer with Mixed Types for an example of working with heterogeneous (e.g. categorical and numeric)
data.
1.2.22 Why does Scikit-learn not directly work with, for example, pan-
das.DataFrame?
The homogeneous NumPy and SciPy data objects currently expected are most efficient to process for most operations.
Extensive work would also be needed to support Pandas categorical types. Restricting input to homogeneous types
therefore reduces maintenance cost and encourages usage of efficient data structures.

1.3 Support
There are several ways to get in touch with the developers.
1.3.1 Mailing List
• The main mailing list is scikit-learn.

• There is also a commit list scikit-learn-commits, where updates to the main repository and test failures get
notified.
1.3.2 User questions
• Some scikit-learn developers support users on StackOverflow using the [scikit-learn] tag.
• For general theoretical or methodological Machine Learning questions stack exchange is probably a more suit-
able venue.
In both cases please use a descriptive question in the title field (e.g. no “Please help with scikit-learn!” as this is not a
question) and put details on what you tried to achieve, what were the expected results and what you observed instead
in the details field.
Code and data snippets are welcome. Minimalistic (up to ~20 lines long) reproduction script very helpful.
Please describe the nature of your data and the how you preprocessed it: what is the number of samples, what is the
number and type of features (i.d. categorical or numerical) and for supervised learning tasks, what target are your
trying to predict: binary, multiclass (1 out of n_classes) or multilabel (k out of n_classes) classification or
continuous variable regression.
1.3.3 Bug tracker
If you think you’ve encountered a bug, please report it to the issue tracker:
https://github.com/scikit-learn/scikit-learn/issues
Don’t forget to include:
• steps (or better script) to reproduce,
• expected outcome,
• observed outcome or python (or gdb) tracebacks
To help developers fix your bug faster, please link to a https://gist.github.com holding a standalone minimalistic python
script that reproduces your bug and optionally a minimalistic subsample of your dataset (for instance exported as CSV
files using numpy.savetxt).
Note: gists are git cloneable repositories and thus you can use git to push datafiles to them.
1.3.4 IRC
Some developers like to hang out on channel #scikit-learn on irc.freenode.net.

If you do not have an IRC client or are behind a firewall this web client works fine: https://webchat.freenode.net

1.3.5 Documentation resources
This documentation is relative to 0.20.3. Documentation for other versions can be found here.
Printable pdf documentation for old versions can be found here.
1.4 Related Projects
Projects implementing the scikit-learn estimator API are encouraged to use the scikit-learn-contrib template which
facilitates best practices for testing and documenting estimators. The scikit-learn-contrib GitHub organisation also
accepts high-quality contributions of repositories conforming to this template.
Below is a list of sister-projects, extensions and domain specific packages.
1.4.1 Interoperability and framework enhancements
These tools adapt scikit-learn for use with other technologies or otherwise enhance the functionality of scikit-learn’s
estimators.
Data formats
• sklearn_pandas bridge for scikit-learn pipelines and pandas data frame with dedicated transformers.
• sklearn_xarray provides compatibility of scikit-learn estimators with xarray data structures.
Auto-ML
• auto_ml Automated machine learning for production and analytics, built on scikit-learn and related projects.
Trains a pipeline wth all the standard machine learning steps. Tuned for prediction speed and ease of transfer to
production environments.
• auto-sklearn An automated machine learning toolkit and a drop-in replacement for a scikit-learn estimator
• TPOT An automated machine learning toolkit that optimizes a series of scikit-learn operators to design a ma-
chine learning pipeline, including data and feature preprocessors as well as the estimators. Works as a drop-in
replacement for a scikit-learn estimator.
• scikit-optimize A library to minimize (very) expensive and noisy black-box functions. It implements sev-
eral methods for sequential model-based optimization, and includes a replacement for GridSearchCV or
RandomizedSearchCV to do cross-validated parameter search using any of these strategies.
Experimentation frameworks
• REP Environment for conducting data-driven research in a consistent and reproducible way
• ML Frontend provides dataset management and SVM fitting/prediction through web-based and programmatic
interfaces.
• Scikit-Learn Laboratory A command-line wrapper around scikit-learn that makes it easy to run machine learning
experiments with multiple learners and large feature sets.
• Xcessiv is a notebook-like application for quick, scalable, and automated hyperparameter tuning and stacked
ensembling. Provides a framework for keeping track of model-hyperparameter combinations.
Model inspection and visualisation
• eli5 A library for debugging/inspecting machine learning models and explaining their predictions.
• mlxtend Includes model visualization utilities.
1.4. Related Projects 9

• scikit-plot A visualization library for quick and easy generation of common plots in data analysis and machine
learning.
• yellowbrick A suite of custom matplotlib visualizers for scikit-learn estimators to support visual feature analysis,
model selection, evaluation, and diagnostics.
Model export for production
• onnxmltools Serializes many Scikit-learn pipelines to ONNX for interchange and prediction.
• sklearn2pmml Serialization of a wide variety of scikit-learn estimators and transformers into PMML with the
help of JPMML-SkLearn library.
• sklearn-porter Transpile trained scikit-learn models to C, Java, Javascript and others.
• sklearn-compiledtrees Generate a C++ implementation of the predict function for decision trees (and ensembles)
trained by sklearn. Useful for latency-sensitive production environments.
1.4.2 Other estimators and tasks
Not everything belongs or is mature enough for the central scikit-learn project. The following are projects providing
interfaces similar to scikit-learn for additional learning algorithms, infrastructures and tasks.
Structured learning
• Seqlearn Sequence classification using HMMs or structured perceptron.
• HMMLearn Implementation of hidden markov models that was previously part of scikit-learn.
• PyStruct General conditional random fields and structured prediction.
• pomegranate Probabilistic modelling for Python, with an emphasis on hidden Markov models.
• sklearn-crfsuite Linear-chain conditional random fields (CRFsuite wrapper with sklearn-like API).
Deep neural networks etc.
• pylearn2 A deep learning and neural network library build on theano with scikit-learn like interface.
• sklearn_theano scikit-learn compatible estimators, transformers, and datasets which use Theano internally
• nolearn A number of wrappers and abstractions around existing neural network libraries
• keras Deep Learning library capable of running on top of either TensorFlow or Theano.
• lasagne A lightweight library to build and train neural networks in Theano.
• skorch A scikit-learn compatible neural network library that wraps PyTorch.
Broad scope
• mlxtend Includes a number of additional estimators as well as model visualization utilities.
• sparkit-learn Scikit-learn API and functionality for PySpark’s distributed modelling.
Other regression and classification
• xgboost Optimised gradient boosted decision tree library.
• ML-Ensemble Generalized ensemble learning (stacking, blending, subsemble, deep ensembles, etc.).
• lightning Fast state-of-the-art linear model solvers (SDCA, AdaGrad, SVRG, SAG, etc. . . ).
• py-earth Multivariate adaptive regression splines
• Kernel Regression Implementation of Nadaraya-Watson kernel regression with automatic bandwidth selection
• gplearn Genetic Programming for symbolic regression tasks.

• multiisotonic Isotonic regression on multidimensional features.

• scikit-multilearn Multi-label classification with focus on label space manipulation.
• seglearn Time series and sequence learning using sliding window segmentation.
Decomposition and clustering
• lda: Fast implementation of latent Dirichlet allocation in Cython which uses Gibbs sampling
to sample from the true posterior distribution. (scikit-learn’s sklearn.decomposition.
LatentDirichletAllocation implementation uses variational inference to sample from a tractable
approximation of a topic model’s posterior distribution.)
• Sparse Filtering Unsupervised feature learning based on sparse-filtering
• kmodes k-modes clustering algorithm for categorical data, and several of its variations.
• hdbscan HDBSCAN and Robust Single Linkage clustering algorithms for robust variable density clustering.
• spherecluster Spherical K-means and mixture of von Mises Fisher clustering routines for data on the unit hyper-
sphere.
Pre-processing
• categorical-encoding A library of sklearn compatible categorical variable encoders.
• imbalanced-learn Various methods to under- and over-sample datasets.
1.4.3 Statistical learning with Python
Other packages useful for data analysis and machine learning.

• Pandas Tools for working with heterogeneous and columnar data, relational queries, time series and basic statis-
tics.
• theano A CPU/GPU array processing framework geared towards deep learning research.
• statsmodels Estimating and analysing statistical models. More focused on statistical tests and less on prediction
than scikit-learn.
• PyMC Bayesian statistical models and fitting algorithms.
• Sacred Tool to help you configure, organize, log and reproduce experiments
• Seaborn Visualization library based on matplotlib. It provides a high-level interface for drawing attractive
statistical graphics.
• Deep Learning A curated list of deep learning software libraries.
Domain specific packages
• scikit-image Image processing and computer vision in python.

• Natural language toolkit (nltk) Natural language processing and some machine learning.
• gensim A library for topic modelling, document indexing and similarity retrieval
• NiLearn Machine learning for neuro-imaging.
• AstroML Machine learning for astronomy.
• MSMBuilder Machine learning for protein conformational dynamics time series.
• scikit-surprise A scikit for building and evaluating recommender systems.
1.4. Related Projects 11

1.4.4 Snippets and tidbits
The wiki has more!
1.5 About us
1.5.1 History
This project was started in 2007 as a Google Summer of Code project by David Cournapeau. Later that year, Matthieu
Brucher started work on this project as part of his thesis.
In 2010 Fabian Pedregosa, Gael Varoquaux, Alexandre Gramfort and Vincent Michel of INRIA took leadership of the
project and made the first public release, February the 1st 2010. Since then, several releases have appeared following
a ~3 month cycle, and a thriving international community has been leading the development.
1.5.2 Authors
The following people have been core contributors to scikit-learn’s development and maintenance:
Please do not email the authors directly to ask for assistance or report issues. Instead, please see What’s the best way
to ask questions about scikit-learn in the FAQ.
See also:
How you can contribute to the project
1.5.3 Governance
The decision making process and governance structure of scikit-learn is laid out in the governance document.
1.5.4 Citing scikit-learn
If you use scikit-learn in a scientific publication, we would appreciate citations to the following paper:
Scikit-learn: Machine Learning in Python, Pedregosa et al., JMLR 12, pp. 2825-2830, 2011.
Bibtex entry:
@article{scikit-learn,
title={Scikit-learn: Machine Learning in {P}ython},
author={Pedregosa, F. and Varoquaux, G. and Gramfort, A. and Michel, V.
and Thirion, B. and Grisel, O. and Blondel, M. and Prettenhofer, P.
and Weiss, R. and Dubourg, V. and Vanderplas, J. and Passos, A. and
Cournapeau, D. and Brucher, M. and Perrot, M. and Duchesnay, E.},
journal={Journal of Machine Learning Research},
volume={12},
pages={2825--2830},
year={2011}
}
If you want to cite scikit-learn for its API or design, you may also want to consider the following paper:

API design for machine learning software: experiences from the scikit-learn project, Buitinck et al., 2013.
Bibtex entry:
@inproceedings{sklearn_api,
author = {Lars Buitinck and Gilles Louppe and Mathieu Blondel and
Fabian Pedregosa and Andreas Mueller and Olivier Grisel and
Vlad Niculae and Peter Prettenhofer and Alexandre Gramfort
and Jaques Grobler and Robert Layton and Jake VanderPlas and
Arnaud Joly and Brian Holt and Ga{\"{e}}l Varoquaux},
title = {{API} design for machine learning software: experiences from
˓→the scikit-learn
project},
booktitle = {ECML PKDD Workshop: Languages for Data Mining and Machine
˓→Learning},
year = {2013},
pages = {108--122},
}
1.5.5 Artwork
High quality PNG and SVG logos are available in the doc/logos/ source directory.
1.5.6 Funding
INRIA actively supports this project. It has provided funding for Fabian Pedregosa (2010-2012), Jaques Grob-
ler (2012-2013) and Olivier Grisel (2013-2017) to work on this project full-time. It also hosts coding sprints and
other events. Paris-Saclay Center for Data Science funded one year
for a developer to work on the project full-time (2014-2015) and 50% of the time of Guillaume Lemaitre (2016-
2017). NYU Moore-Sloan Data Science Environment funded An-

dreas Mueller (2014-2016) to work on this project. The Moore-Sloan Data Science Environment also funds sev-
1.5. About us 13
eral students to work on the project part-time. Télécom Paris-

tech funded Manoj Kumar (2014), Tom Dupré la Tour (2015), Raghav RV (2015-2017), Thierry Guillemot (2016-
2017) and Albert Thomas (2017) to work on scikit-learn. Columbia University funds An-
dreas Müller since 2016. Andreas Müller also received a grant to improve scikit-learn
from the Alfred P. Sloan Foundation in 2017. The University of
Sydney funds Joel Nothman since July 2017. The Labex Digi-
Cosme funded Nicolas Goix (2015-2016), Tom Dupré la Tour (2015-2016 and 2017-2018), Mathurin Massias (2018-
2019) to work part time on scikit-learn during their PhDs. It also funded a scikit-learn coding sprint in 2015.
The following students were sponsored by Google to work on

scikit-learn through the Google Summer of Code program.
• 2007 - David Cournapeau
• 2011 - ‘Vlad Niculae‘_

• 2012 - ‘Vlad Niculae‘_, Immanuel Bayer.

• 2013 - Kemal Eren, Nicolas Trésegnie
• 2014 - Hamzeh Alsalhi, Issam Laradji, Maheshakya Wijewardena, Manoj Kumar.
• 2015 - Raghav RV, Wei Xue
• 2016 - Nelson Liu, YenChen Lin
It also provided funding for sprints and events around scikit-learn. If you would like to participate in the next Google
Summer of code program, please see this page.
The NeuroDebian project providing Debian packaging and contributions is supported by Dr. James V. Haxby (Dart-
mouth College).
The PSF helped find and manage funding for our 2011 Granada sprint. More information can be found here
tinyclues funded the 2011 international Granada sprint.
Donating to the project
If you are interested in donating to the project or to one of our code-sprints, you can use the Paypal button below or the
NumFOCUS Donations Page (if you use the latter, please indicate that you are donating for the scikit-learn project).
All donations will be handled by NumFOCUS, a non-profit-organization which is managed by a board of Scipy
community members. NumFOCUS’s mission is to foster scientific computing software, in particular in Python. As
a fiscal home of scikit-learn, it ensures that money is available when needed to keep the project funded and available
while in compliance with tax regulations.
The received donations for the scikit-learn project mostly will go towards covering travel-expenses for code sprints, as
well as towards the organization budget of the project1 .
Notes
The 2013 Paris international sprint
1.5. About us 15
Fig. 1.1: IAP VII/19 - DYSCO
For more information on this sprint, see here
1.5.7 Infrastructure support
• We would like to thank Rackspace for providing us with a free Rackspace Cloud account to automatically build
the documentation and the example gallery from for the development version of scikit-learn using this tool.
• We would also like to thank Shining Panda for free CPU time on their Continuous Integration server.
1.6 Who is using scikit-learn?
1.6.1 J.P.Morgan
Scikit-learn is an indispensable part of the Python machine learning toolkit at JPMorgan. It is very widely used
across all parts of the bank for classification, predictive analytics, and very many other machine learning tasks. Its
straightforward API, its breadth of algorithms, and the quality of its documentation combine to make scikit-learn
simultaneously very approachable and very powerful.
Stephen Simmons, VP, Athena Research, JPMorgan
1.6.2 Spotify
1 Regarding the organization budget in particular, we might use some of the donated funds to pay for other project expenses such as DNS,
hosting or continuous integration services.

Scikit-learn provides a toolbox with solid implementations of a bunch of state-of-the-art models and makes it easy to
plug them into existing applications. We’ve been using it quite a lot for music recommendations at Spotify and I think
it’s the most well-designed ML package I’ve seen so far.
Erik Bernhardsson, Engineering Manager Music Discovery & Machine Learning, Spotify
1.6.3 Inria
At INRIA, we use scikit-learn to support leading-edge basic research in many teams: Parietal for neuroimaging, Lear
for computer vision, Visages for medical image analysis, Privatics for security. The project is a fantastic tool to
address difficult applications of machine learning in an academic environment as it is performant and versatile, but all
easy-to-use and well documented, which makes it well suited to grad students.
Gaël Varoquaux, research at Parietal
1.6.4 betaworks
Betaworks is a NYC-based startup studio that builds new products, grows companies, and invests in others. Over
the past 8 years we’ve launched a handful of social data analytics-driven services, such as Bitly, Chartbeat, digg and
Scale Model. Consistently the betaworks data science team uses Scikit-learn for a variety of tasks. From exploratory
analysis, to product development, it is an essential part of our toolkit. Recent uses are included in digg’s new video
recommender system, and Poncho’s dynamic heuristic subspace clustering.
Gilad Lotan, Chief Data Scientist
1.6.5 Hugging Face
At Hugging Face we’re using NLP and probabilistic models to generate conversational Artificial intelligences that are
fun to chat with. Despite using deep neural nets for a few of our NLP tasks, scikit-learn is still the bread-and-butter of
our daily machine learning routine. The ease of use and predictability of the interface, as well as the straightforward
mathematical explanations that are here when you need them, is the killer feature. We use a variety of scikit-learn
models in production and they are also operationally very pleasant to work with.
Julien Chaumond, Chief Technology Officer
1.6. Who is using scikit-learn? 17

1.6.6 Evernote
Building a classifier is typically an iterative process of exploring the data, selecting the features (the attributes of the
data believed to be predictive in some way), training the models, and finally evaluating them. For many of these tasks,
we relied on the excellent scikit-learn package for Python.
Read more
Mark Ayzenshtat, VP, Augmented Intelligence
1.6.7 Télécom ParisTech
At Telecom ParisTech, scikit-learn is used for hands-on sessions and home assignments in introductory and advanced
machine learning courses. The classes are for undergrads and masters students. The great benefit of scikit-learn is its
fast learning curve that allows students to quickly start working on interesting and motivating problems.
Alexandre Gramfort, Assistant Professor
1.6.8 Booking.com
At Booking.com, we use machine learning algorithms for many different applications, such as recommending ho-
tels and destinations to our customers, detecting fraudulent reservations, or scheduling our customer service agents.
Scikit-learn is one of the tools we use when implementing standard algorithms for prediction tasks. Its API and doc-
umentations are excellent and make it easy to use. The scikit-learn developers do a great job of incorporating state of
the art implementations and new algorithms into the package. Thus, scikit-learn provides convenient access to a wide
spectrum of algorithms, and allows us to readily find the right tool for the right job.
Melanie Mueller, Data Scientist
1.6.9 AWeber

The scikit-learn toolkit is indispensable for the Data Analysis and Management team at AWeber. It allows us to do
AWesome stuff we would not otherwise have the time or resources to accomplish. The documentation is excellent,
allowing new engineers to quickly evaluate and apply many different algorithms to our data. The text feature extraction
utilities are useful when working with the large volume of email content we have at AWeber. The RandomizedPCA
implementation, along with Pipelining and FeatureUnions, allows us to develop complex machine learning algorithms
efficiently and reliably.
Anyone interested in learning more about how AWeber deploys scikit-learn in a production environment should check
out talks from PyData Boston by AWeber’s Michael Becker available at https://github.com/mdbecker/pydata_2013
Michael Becker, Software Engineer, Data Analysis and Management Ninjas
1.6.10 Yhat
The combination of consistent APIs, thorough documentation, and top notch implementation make scikit-learn our
favorite machine learning package in Python. scikit-learn makes doing advanced analysis in Python accessible to
anyone. At Yhat, we make it easy to integrate these models into your production applications. Thus eliminating the
unnecessary dev time encountered productionizing analytical work.
Greg Lamp, Co-founder Yhat
1.6.11 Rangespan
The Python scikit-learn toolkit is a core tool in the data science group at Rangespan. Its large collection of well
documented models and algorithms allow our team of data scientists to prototype fast and quickly iterate to find the
right solution to our learning problems. We find that scikit-learn is not only the right tool for prototyping, but its
careful and well tested implementation give us the confidence to run scikit-learn models in production.
Jurgen Van Gael, Data Science Director at Rangespan Ltd

1.6.12 Birchbox
At Birchbox, we face a range of machine learning problems typical to E-commerce: product recommendation, user
clustering, inventory prediction, trends detection, etc. Scikit-learn lets us experiment with many models, especially in
the exploration phase of a new project: the data can be passed around in a consistent way; models are easy to save and
reuse; updates keep us informed of new developments from the pattern discovery research community. Scikit-learn is
an important tool for our team, built the right way in the right language.
Thierry Bertin-Mahieux, Birchbox, Data Scientist
1.6.13 Bestofmedia Group
Scikit-learn is our #1 toolkit for all things machine learning at Bestofmedia. We use it for a variety of tasks (e.g. spam
fighting, ad click prediction, various ranking models) thanks to the varied, state-of-the-art algorithm implementations
packaged into it. In the lab it accelerates prototyping of complex pipelines. In production I can say it has proven to be
robust and efficient enough to be deployed for business critical components.
Eustache Diemert, Lead Scientist Bestofmedia Group
1.6.14 Change.org

At change.org we automate the use of scikit-learn’s RandomForestClassifier in our production systems to drive email
targeting that reaches millions of users across the world each week. In the lab, scikit-learn’s ease-of-use, performance,
and overall variety of algorithms implemented has proved invaluable in giving us a single reliable source to turn to for
our machine-learning needs.
Vijay Ramesh, Software Engineer in Data/science at Change.org
1.6.15 PHIMECA Engineering
At PHIMECA Engineering, we use scikit-learn estimators as surrogates for expensive-to-evaluate numerical models
(mostly but not exclusively finite-element mechanical models) for speeding up the intensive post-processing operations
involved in our simulation-based decision making framework. Scikit-learn’s fit/predict API together with its efficient
cross-validation tools considerably eases the task of selecting the best-fit estimator. We are also using scikit-learn for
illustrating concepts in our training sessions. Trainees are always impressed by the ease-of-use of scikit-learn despite
the apparent theoretical complexity of machine learning.
Vincent Dubourg, PHIMECA Engineering, PhD Engineer
1.6.16 HowAboutWe
At HowAboutWe, scikit-learn lets us implement a wide array of machine learning techniques in analysis and in pro-
duction, despite having a small team. We use scikit-learn’s classification algorithms to predict user behavior, enabling
us to (for example) estimate the value of leads from a given traffic source early in the lead’s tenure on our site. Also, our
users’ profiles consist of primarily unstructured data (answers to open-ended questions), so we use scikit-learn’s fea-
ture extraction and dimensionality reduction tools to translate these unstructured data into inputs for our matchmaking
system.
Daniel Weitzenfeld, Senior Data Scientist at HowAboutWe
1.6.17 PeerIndex
At PeerIndex we use scientific methodology to build the Influence Graph - a unique dataset that allows us to identify
who’s really influential and in which context. To do this, we have to tackle a range of machine learning and predic-
tive modeling problems. Scikit-learn has emerged as our primary tool for developing prototypes and making quick
progress. From predicting missing data and classifying tweets to clustering communities of social media users, scikit-

learn proved useful in a variety of applications. Its very intuitive interface and excellent compatibility with other
python tools makes it and indispensable tool in our daily research efforts.
Ferenc Huszar - Senior Data Scientist at Peerindex
1.6.18 DataRobot
DataRobot is building next generation predictive analytics software to make data scientists more productive, and
scikit-learn is an integral part of our system. The variety of machine learning techniques in combination with the
solid implementations that scikit-learn offers makes it a one-stop-shopping library for machine learning in Python.
Moreover, its consistent API, well-tested code and permissive licensing allow us to use it in a production environment.
Scikit-learn has literally saved us years of work we would have had to do ourselves to bring our product to market.
Jeremy Achin, CEO & Co-founder DataRobot Inc.
1.6.19 OkCupid
We’re using scikit-learn at OkCupid to evaluate and improve our matchmaking system. The range of features it has,
especially preprocessing utilities, means we can use it for a wide variety of projects, and it’s performant enough to
handle the volume of data that we need to sort through. The documentation is really thorough, as well, which makes
the library quite easy to use.
David Koh - Senior Data Scientist at OkCupid
1.6.20 Lovely
At Lovely, we strive to deliver the best apartment marketplace, with respect to our users and our listings. From
understanding user behavior, improving data quality, and detecting fraud, scikit-learn is a regular tool for gathering
insights, predictive modeling and improving our product. The easy-to-read documentation and intuitive architecture of
the API makes machine learning both explorable and accessible to a wide range of python developers. I’m constantly
recommending that more developers and scientists try scikit-learn.
Simon Frid - Data Scientist, Lead at Lovely

1.6.21 Data Publica
Data Publica builds a new predictive sales tool for commercial and marketing teams called C-Radar. We extensively
use scikit-learn to build segmentations of customers through clustering, and to predict future customers based on past
partnerships success or failure. We also categorize companies using their website communication thanks to scikit-learn
and its machine learning algorithm implementations. Eventually, machine learning makes it possible to detect weak
signals that traditional tools cannot see. All these complex tasks are performed in an easy and straightforward way
thanks to the great quality of the scikit-learn framework.
Guillaume Lebourgeois & Samuel Charron - Data Scientists at Data Publica
1.6.22 Machinalis
Scikit-learn is the cornerstone of all the machine learning projects carried at Machinalis. It has a consistent API, a
wide selection of algorithms and lots of auxiliary tools to deal with the boilerplate. We have used it in production en-
vironments on a variety of projects including click-through rate prediction, information extraction, and even counting
sheep!
In fact, we use it so much that we’ve started to freeze our common use cases into Python packages, some of them
open-sourced, like FeatureForge . Scikit-learn in one word: Awesome.
Rafael Carrascosa, Lead developer
1.6.23 solido
Scikit-learn is helping to drive Moore’s Law, via Solido. Solido creates computer-aided design tools used by the
majority of top-20 semiconductor companies and fabs, to design the bleeding-edge chips inside smartphones, auto-
mobiles, and more. Scikit-learn helps to power Solido’s algorithms for rare-event estimation, worst-case verification,
optimization, and more. At Solido, we are particularly fond of scikit-learn’s libraries for Gaussian Process models,
large-scale regularized linear regression, and classification. Scikit-learn has increased our productivity, because for
many ML problems we no longer need to “roll our own” code. This PyData 2014 talk has details.
Trent McConaghy, founder, Solido Design Automation Inc.

1.6.24 INFONEA
We employ scikit-learn for rapid prototyping and custom-made Data Science solutions within our in-memory based
Business Intelligence Software INFONEA®. As a well-documented and comprehensive collection of state-of-the-art
algorithms and pipelining methods, scikit-learn enables us to provide flexible and scalable scientific analysis solutions.
Thus, scikit-learn is immensely valuable in realizing a powerful integration of Data Science technology within self-
service business analytics.
Thorsten Kranz, Data Scientist, Coma Soft AG.
1.6.25 Dataiku
Our software, Data Science Studio (DSS), enables users to create data services that combine ETL with Machine
Learning. Our Machine Learning module integrates many scikit-learn algorithms. The scikit-learn library is a perfect
integration with DSS because it offers algorithms for virtually all business cases. Our goal is to offer a transparent and
flexible tool that makes it easier to optimize time consuming aspects of building a data service, preparing data, and
training machine learning algorithms on all types of data.
Florian Douetteau, CEO, Dataiku
1.6.26 Otto Group
Here at Otto Group, one of global Big Five B2C online retailers, we are using scikit-learn in all aspects of our daily
work from data exploration to development of machine learning application to the productive deployment of those
services. It helps us to tackle machine learning problems ranging from e-commerce to logistics. It consistent APIs
enabled us to build the Palladium REST-API framework around it and continuously deliver scikit-learn based services.
Christian Rammig, Head of Data Science, Otto Group
1.6.27 Zopa
At Zopa, the first ever Peer-to-Peer lending platform, we extensively use scikit-learn to run the business and optimize
our users’ experience. It powers our Machine Learning models involved in credit risk, fraud risk, marketing, and
pricing, and has been used for originating at least 1 billion GBP worth of Zopa loans. It is very well documented,
powerful, and simple to use. We are grateful for the capabilities it has provided, and for allowing us to deliver on our
mission of making money simple and fair.

Vlasios Vasileiou, Head of Data Science, Zopa
1.6.28 MARS
Scikit-Learn is integral to the Machine Learning Ecosystem at Mars. Whether we’re designing better recipes for
petfood or closely analysing our cocoa supply chain, Scikit-Learn is used as a tool for rapidly prototyping ideas and
taking them to production. This allows us to better understand and meet the needs of our consumers worldwide.
Scikit-Learn’s feature-rich toolset is easy to use and equips our associates with the capabilities they need to solve the
business challenges they face every day.
Michael Fitzke Next Generation Technologies Sr Leader, Mars Inc.
1.7 Release History
Release notes for current and recent releases are detailed on this page, with previous releases linked below.
Tip: Subscribe to scikit-learn releases on libraries.io to be notified when new versions are released.
1.8 Version 0.20.3
March 1, 2019
This is a bug-fix release with some minor documentation improvements and enhancements to features released in
0.20.0.
1.8.1 Changelog
sklearn.cluster
• [F IX ] Fixed a bug in cluster.KMeans where computation was single threaded when n_jobs > 1 or n_jobs =
-1. #12949 by Prabakaran Kumaresshan.
sklearn.compose
• [F IX ] Fixed a bug in compose.ColumnTransformer to handle negative indexes in the columns list of the
transformers. #12946 by Pierre Tallotte.
sklearn.covariance
• [F IX ] Fixed a regression in covariance.graphical_lasso so that the case n_features=2 is handled

correctly. #13276 by Aurélien Bellet.
1.7. Release History 25

sklearn.decomposition
• [F IX ] Fixed a bug in decomposition.sparse_encode where computation was single threaded when

n_jobs > 1 or n_jobs = -1. #13005 by Prabakaran Kumaresshan.
sklearn.datasets
• [E FFICIENCY ] sklearn.datasets.fetch_openml now loads data by streaming, avoiding high memory

usage. #13312 by Joris Van den Bossche.
sklearn.feature_extraction
• [F IX ] Fixed a bug in feature_extraction.text.CountVectorizer which would result in the sparse

feature matrix having conflicting indptr and indices precisions under very large vocabularies. #11295 by Gabriel
Vacaliuc.
sklearn.impute
• [F IX ] add support for non-numeric data in sklearn.impute.MissingIndicator which was not sup-
ported while sklearn.impute.SimpleImputer was supporting this for some imputation strategies.
#13046 by Guillaume Lemaitre.
sklearn.linear_model
• [F IX ] Fixed a bug in linear_model.MultiTaskElasticNet and linear_model.

MultiTaskLasso which were breaking when warm_start = True. #12360 by Aakanksha Joshi.
sklearn.preprocessing
• [F IX ] Fixed a bug in preprocessing.KBinsDiscretizer where strategy='kmeans' fails with an

error during transformation due to unsorted bin edges. #13134 by Sandro Casagrande.
• [F IX ] Fixed a bug in preprocessing.OneHotEncoder where the deprecation of
categorical_features was handled incorrectly in combination with handle_unknown='ignore'.
#12881 by Joris Van den Bossche.
• [F IX ] Bins whose width are too small (i.e., <= 1e-8) are removed with a warning in preprocessing.
KBinsDiscretizer. #13165 by Hanmin Qin.
sklearn.svm
• [F IX ] Fixed a bug in svm.SVC, svm.NuSVC, svm.SVR, svm.NuSVR and svm.OneClassSVM where the
scale option of parameter gamma is erroneously defined as 1 / (n_features * X.std()). It’s now
defined as 1 / (n_features * X.var()). #13221 by Hanmin Qin.

1.8.2 Code and Documentation Contributors
With thanks to:

Adrin Jalali, Agamemnon Krasoulis, Albert Thomas, Andreas Mueller, Aurélien Bellet, bertrandhaut, Bharat Raghu-
nathan, Dowon, Emmanuel Arias, Fibinse Xavier, Finn O’Shea, Gabriel Vacaliuc, Gael Varoquaux, Guillaume
Lemaitre, Hanmin Qin, joaak, Joel Nothman, Joris Van den Bossche, Jérémie Méhault, kms15, Kossori Aruku, Lak-
shya KD, maikia, Manuel López-Ibáñez, Marco Gorelli, MarcoGorelli, mferrari3, Mickaël Schoentgen, Nicolas Hug,
pavlos kallis, Pierre Glaser, pierretallotte, Prabakaran Kumaresshan, Reshama Shaikh, Rohit Kapoor, Roman Yurchak,
SandroCasagrande, Tashay Green, Thomas Fan, Vishaal Kapoor, Zhuyi Xue, Zijie (ZJ) Poh
1.9 Version 0.20.2
December 20, 2018

0.20.0.
1.9.1 Changed models
The following estimators and functions, when fit with the same data and parameters, may produce different models
from the previous version. This often occurs due to changes in the modelling logic (bug fixes or enhancements), or in
random sampling procedures.
• sklearn.neighbors when metric=='jaccard' (bug fix)
• use of 'seuclidean' or 'mahalanobis' metrics in some cases (bug fix)
1.9.2 Changelog
sklearn.compose
• [F IX ] Fixed an issue in compose.make_column_transformer which raises unexpected error when

columns is pandas Index or pandas Series. #12704 by Hanmin Qin.
sklearn.metrics
• [F IX ] Fixed a bug in metrics.pairwise_distances and metrics.

pairwise_distances_chunked where parameters V of "seuclidean" and VI of "mahalanobis"
metrics were computed after the data was split into chunks instead of being pre-computed on whole data.
#12701 by Jeremie du Boisberranger.
sklearn.neighbors
• [F IX ] Fixed sklearn.neighbors.DistanceMetric jaccard distance function to return 0 when two all-

zero vectors are compared. #12685 by Thomas Fan.
1.9. Version 0.20.2 27

sklearn.utils
• [F IX ] Calling utils.check_array on pandas.Series with categorical data, which raised an error in 0.20.0,
now returns the expected output again. #12699 by Joris Van den Bossche.
With thanks to:

adanhawth, Adrin Jalali, Albert Thomas, Andreas Mueller, Dan Stine, Feda Curic, Hanmin Qin, Jan S, jeremiedbb,
Joel Nothman, Joris Van den Bossche, josephsalmon, Katrin Leinweber, Loic Esteve, Muhammad Hassaan Rafique,
Nicolas Hug, Olivier Grisel, Paul Paczuski, Reshama Shaikh, Sam Waterbury, Shivam Kotwalia, Thomas Fan
1.10 Version 0.20.1
November 21, 2018

0.20.0. Note that we also include some API changes in this release, so you might get some extra warnings after
updating from 0.20.0 to 0.20.1.
• decomposition.IncrementalPCA (bug fix)
1.10.2 Changelog
sklearn.cluster
• [E FFICIENCY ] make cluster.MeanShift no longer try to do nested parallelism as the overhead would hurt
performance significantly when n_jobs > 1. #12159 by Olivier Grisel.
• [F IX ] Fixed a bug in cluster.DBSCAN with precomputed sparse neighbors graph, which would add explicitly
zeros on the diagonal even when already present. #12105 by Tom Dupre la Tour.
sklearn.compose
• [F IX ] Fixed an issue in compose.ColumnTransformer when stacking columns with types not convertible
to a numeric. #11912 by Adrin Jalali.
• [API C HANGE ] compose.ColumnTransformer now applies the sparse_threshold even if all trans-
formation results are sparse. #12304 by Andreas Müller.
• [API C HANGE ] compose.make_column_transformer now expects (transformer, columns) in-
stead of (columns, transformer) to keep consistent with compose.ColumnTransformer. #12339
by Adrin Jalali.

sklearn.datasets
• [F IX ] datasets.fetch_openml to correctly use the local cache. #12246 by Jan N. van Rijn.
• [F IX ] datasets.fetch_openml to correctly handle ignore attributes and row id attributes. #12330 by Jan
N. van Rijn.
• [F IX ] Fixed integer overflow in datasets.make_classification for values of n_informative pa-
rameter larger than 64. #10811 by Roman Feldbauer.
• [F IX ] Fixed olivetti faces dataset DESCR attribute to point to the right location in datasets.
fetch_olivetti_faces. #12441 by Jérémie du Boisberranger
• [F IX ] datasets.fetch_openml to retry downloading when reading from local cache fails. #12517 by
Thomas Fan.
• [F IX ] Fixed a regression in decomposition.IncrementalPCA where 0.20.0 raised an error if the number

of samples in the final batch for fitting IncrementalPCA was smaller than n_components. #12234 by Ming Li.
sklearn.ensemble
• [F IX ] Fixed a bug mostly affecting ensemble.RandomForestClassifier where

class_weight='balanced_subsample' failed with more than 32 classes. #12165 by Joel Nothman.
• [F IX ] Fixed a bug affecting ensemble.BaggingClassifier, ensemble.BaggingRegressor and
ensemble.IsolationForest, where max_features was sometimes rounded down to zero. #12388
by Connor Tann.
• [F IX ] Fixed a regression in v0.20.0 where feature_extraction.text.CountVectorizer and other

text vectorizers could error during stop words validation with custom preprocessors or tokenizers. #12393 by
Roman Yurchak.
• [F IX ] linear_model.SGDClassifier and variants with early_stopping=True would not use a

consistent validation split in the multiclass case and this would cause a crash when using those estimators as
part of parallel parameter search or cross-validation. #12122 by Olivier Grisel.
• [F IX ] Fixed a bug affecting SGDClassifier in the multiclass case. Each one-versus-all step is run in a
joblib.Parallel call and mutating a common parameter, causing a segmentation fault if called within a
backend using processes and not threads. We now use require=sharedmem at the joblib.Parallel
instance creation. #12518 by Pierre Glaser and Olivier Grisel.
sklearn.metrics
• [F IX ] Fixed a bug in metrics.pairwise.pairwise_distances_argmin_min which returned the

square root of the distance when the metric parameter was set to “euclidean”. #12481 by Jérémie du Boisber-
ranger.
1.10. Version 0.20.1 29

• [F IX ] Fixed a bug in metrics.pairwise.pairwise_distances_chunked which didn’t ensure the

diagonal is zero for euclidean distances. #12612 by Andreas Müller.
• [API C HANGE ] The metrics.calinski_harabaz_score has been renamed to metrics.
calinski_harabasz_score and will be removed in version 0.23. #12211 by Lisa Thomas, Mark Hannel
and Melissa Ferrari.
sklearn.mixture
• [F IX ] Ensure that the fit_predict method of mixture.GaussianMixture and mixture.

BayesianGaussianMixture always yield assignments consistent with fit followed by predict even
if the convergence criterion is too loose or not met. #12451 by Olivier Grisel.
sklearn.neighbors
• [F IX ] force the parallelism backend to threading for neighbors.KDTree and neighbors.BallTree

in Python 2.7 to avoid pickling errors caused by the serialization of their methods. #12171 by Thomas Moreau.
• [F IX ] Fixed bug in preprocessing.OrdinalEncoder when passing manually specified categories.

#12365 by Joris Van den Bossche.
• [F IX ] Fixed bug in preprocessing.KBinsDiscretizer where the transform method mutates the
_encoder attribute. The transform method is now thread safe. #12514 by Hanmin Qin.
• [F IX ] Fixed a bug in preprocessing.PowerTransformer where the Yeo-Johnson transform was incor-
rect for lambda parameters outside of [0, 2] #12522 by Nicolas Hug.
• [F IX ] Fixed a bug in preprocessing.OneHotEncoder where transform failed when set to ignore un-
known numpy strings of different lengths #12471 by Gabriel Marzinotto.
• [API C HANGE ] The default value of the method argument in preprocessing.power_transform will
be changed from box-cox to yeo-johnson to match preprocessing.PowerTransformer in ver-
sion 0.23. A FutureWarning is raised when the default value is used. #12317 by Eric Chang.
sklearn.utils
• [F IX ] Use float64 for mean accumulator to avoid floating point precision issues in preprocessing.
StandardScaler and decomposition.IncrementalPCA when using float32 datasets. #12338 by
bauks.
• [F IX ] Calling utils.check_array on pandas.Series, which raised an error in 0.20.0, now returns the ex-
pected output again. #12625 by Andreas Müller
Miscellaneous
• [F IX ] When using site joblib by setting the environment variable SKLEARN_SITE_JOBLIB, added compatibility
with joblib 0.11 in addition to 0.12+. #12350 by Joel Nothman and Roman Yurchak.
• [F IX ] Make sure to avoid raising FutureWarning when calling np.vstack with numpy 1.16 and later (use
list comprehensions instead of generator expressions in many locations of the scikit-learn code base). #12467
by Olivier Grisel.

• [API C HANGE ] Removed all mentions of sklearn.externals.joblib, and deprecated joblib

methods exposed in sklearn.utils, except for utils.parallel_backend and utils.
register_parallel_backend, which allow users to configure parallel computation in scikit-learn. Other
functionalities are part of joblib. package and should be used directly, by installing it. The goal of this change
is to prepare for unvendoring joblib in future version of scikit-learn. #12345 by Thomas Moreau
With thanks to:

^__^, Adrin Jalali, Andrea Navarrete, Andreas Mueller, bauks, BenjaStudio, Cheuk Ting Ho, Connossor, Corey Levin-
son, Dan Stine, daten-kieker, Denis Kataev, Dillon Gardner, Dmitry Vukolov, Dougal J. Sutherland, Edward J Brown,
Eric Chang, Federico Caselli, Gabriel Marzinotto, Gael Varoquaux, GauravAhlawat, Gustavo De Mari Pereira, Han-
min Qin, haroldfox, JackLangerman, Jacopo Notarstefano, janvanrijn, jdethurens, jeremiedbb, Joel Nothman, Joris
Van den Bossche, Koen, Kushal Chauhan, Lee Yi Jie Joel, Lily Xiong, mail-liam, Mark Hannel, melsyt, Ming Li,
Nicholas Smith, Nicolas Hug, Nikolay Shebanov, Oleksandr Pavlyk, Olivier Grisel, Peter Hausamann, Pierre Glaser,
Pulkit Maloo, Quentin Batista, Radostin Stoyanov, Ramil Nugmanov, Rebekah Kim, Reshama Shaikh, Rohan Singh,
Roman Feldbauer, Roman Yurchak, Roopam Sharma, Sam Waterbury, Scott Lowe, Sebastian Raschka, Stephen Tier-
ney, SylvainLan, TakingItCasual, Thomas Fan, Thomas Moreau, Tom Dupré la Tour, Tulio Casagrande, Utkarsh
Upadhyay, Xing Han Lu, Yaroslav Halchenko, Zach Miller
1.11 Version 0.20.0
September 25, 2018

This release packs in a mountain of bug fixes, features and enhancements for the Scikit-learn library, and improvements
to the documentation and examples. Thanks to our contributors!
This release is dedicated to the memory of Raghav Rajagopalan.
Warning: Version 0.20 is the last version of scikit-learn to support Python 2.7 and Python 3.4. Scikit-learn 0.21
will require Python 3.5 or higher.
1.11.1 Highlights
We have tried to improve our support for common data-science use-cases including missing values, categorical vari-
ables, heterogeneous data, and features/targets with unusual distributions. Missing values in features, represented by
NaNs, are now accepted in column-wise preprocessing such as scalers. Each feature is fitted disregarding NaNs, and
data containing NaNs can be transformed. The new impute module provides estimators for learning despite missing
data.
ColumnTransformer handles the case where different features or columns of a pandas.DataFrame need dif-
ferent preprocessing. String or pandas Categorical columns can now be encoded with OneHotEncoder or
OrdinalEncoder.
TransformedTargetRegressor helps when the regression target needs to be transformed to be modeled.
PowerTransformer and KBinsDiscretizer join QuantileTransformer as non-linear transformations.
Beyond this, we have added sample_weight support to several estimators (including KMeans,
BayesianRidge and KernelDensity) and improved stopping criteria in others (including MLPRegressor,
GradientBoostingRegressor and SGDRegressor).
1.11. Version 0.20.0 31

This release is also the first to be accompanied by a Glossary of Common Terms and API Elements developed by Joel
Nothman. The glossary is a reference resource to help users and contributors become familiar with the terminology
and conventions used in Scikit-learn.
Sorry if your contribution didn’t make it into the highlights. There’s a lot here. . .
• cluster.MeanShift (bug fix)
• decomposition.IncrementalPCA in Python 2 (bug fix)
• decomposition.SparsePCA (bug fix)
• ensemble.GradientBoostingClassifier (bug fix affecting feature importances)
• isotonic.IsotonicRegression (bug fix)
• linear_model.ARDRegression (bug fix)
• linear_model.LogisticRegressionCV (bug fix)
• linear_model.OrthogonalMatchingPursuit (bug fix)
• linear_model.PassiveAggressiveClassifier (bug fix)
• linear_model.PassiveAggressiveRegressor (bug fix)
• linear_model.Perceptron (bug fix)
• linear_model.SGDClassifier (bug fix)
• linear_model.SGDRegressor (bug fix)
• metrics.roc_auc_score (bug fix)
• metrics.roc_curve (bug fix)
• neural_network.BaseMultilayerPerceptron (bug fix)
• neural_network.MLPClassifier (bug fix)
• neural_network.MLPRegressor (bug fix)
• The v0.19.0 release notes failed to mention a backwards incompatibility with model_selection.
StratifiedKFold when shuffle=True due to #7823.
Details are listed in the changelog below.
(While we are trying to better inform users by providing this information, we cannot assure that this list is complete.)
1.11.3 Known Major Bugs
• #11924: linear_model.LogisticRegressionCV with solver=’lbfgs’ and multi_class=’multinomial’

may be non-deterministic or otherwise broken on macOS. This appears to be the case on Travis CI servers, but
has not been confirmed on personal MacBooks! This issue has been present in previous releases.

• #9354: metrics.pairwise.euclidean_distances (which is used several times throughout the li-

brary) gives results with poor precision, which particularly affects its use with 32-bit float inputs. This became
more problematic in versions 0.18 and 0.19 when some algorithms were changed to avoid casting 32-bit data
into 64-bit.
1.11.4 Changelog
Support for Python 3.3 has been officially dropped.
sklearn.cluster
• [M AJOR F EATURE ] cluster.AgglomerativeClustering now supports Single Linkage clustering via

linkage='single'. #9372 by Leland McInnes and Steve Astels.
• [F EATURE ] cluster.KMeans and cluster.MiniBatchKMeans now support sample weights via new
parameter sample_weight in fit function. #10933 by Johannes Hansen.
• [E FFICIENCY ] cluster.KMeans, cluster.MiniBatchKMeans and cluster.k_means passed with
algorithm='full' now enforces row-major ordering, improving runtime. #10471 by Gaurav Dhingra.
• [E FFICIENCY ] cluster.DBSCAN now is parallelized according to n_jobs regardless of algorithm. #8003
by Joël Billaud.
• [E NHANCEMENT ] cluster.KMeans now gives a warning if the number of distinct clusters found is smaller
than n_clusters. This may occur when the number of distinct points in the data set is actually smaller than
the number of cluster one is looking for. #10059 by Christian Braune.
• [F IX ] Fixed a bug where the fit method of cluster.AffinityPropagation stored cluster centers as 3d
array instead of 2d array in case of non-convergence. For the same class, fixed undefined and arbitrary behavior
in case of training data where all samples had equal similarity. #9612. By Jonatan Samoocha.
• [F IX ] Fixed a bug in cluster.spectral_clustering where the normalization of the spectrum was
using a division instead of a multiplication. #8129 by Jan Margeta, Guillaume Lemaitre, and Devansh D..
• [F IX ] Fixed a bug in cluster.k_means_elkan where the returned iteration was 1 less than the correct
value. Also added the missing n_iter_ attribute in the docstring of cluster.KMeans. #11353 by Jeremie
du Boisberranger.
• [F IX ] Fixed a bug in cluster.mean_shift where the assigned labels were not deterministic if there were
multiple clusters with the same intensities. #11901 by Adrin Jalali.
• [API C HANGE ] Deprecate pooling_func unused parameter in cluster.
AgglomerativeClustering. #9875 by Kumar Ashutosh.
sklearn.compose
• New module.
• [M AJOR F EATURE ] Added compose.ColumnTransformer, which allows to apply different transformers
to different columns of arrays or pandas DataFrames. #9012 by Andreas Müller and Joris Van den Bossche, and
#11315 by Thomas Fan.
• [M AJOR F EATURE ] Added the compose.TransformedTargetRegressor which transforms the target y
before fitting a regression model. The predictions are mapped back to the original space via an inverse transform.
#9041 by Andreas Müller and Guillaume Lemaitre.
1.11. Version 0.20.0 33

sklearn.covariance
• [E FFICIENCY ] Runtime improvements to covariance.GraphicalLasso. #9858 by Steven Brown.

• [API C HANGE ] The covariance.graph_lasso, covariance.GraphLasso and covariance.
GraphLassoCV have been renamed to covariance.graphical_lasso, covariance.
GraphicalLasso and covariance.GraphicalLassoCV respectively and will be removed in
version 0.22. #9993 by Artiem Krinitsyn
sklearn.datasets
• [M AJOR F EATURE ] Added datasets.fetch_openml to fetch datasets from OpenML. OpenML is a free,
open data sharing platform and will be used instead of mldata as it provides better service availability. #9908 by
Andreas Müller and Jan N. van Rijn.
• [F EATURE ] In datasets.make_blobs, one can now pass a list to the n_samples parameter to indicate
the number of samples to generate per cluster. #8617 by Maskani Filali Mohamed and Konstantinos Katrioplas.
• [F EATURE ] Add filename attribute to datasets that have a CSV file. #9101 by alex-33 and Maskani Filali
Mohamed.
• [F EATURE ] return_X_y parameter has been added to several dataset loaders. #10774 by Chris Catalfo.
• [F IX ] Fixed a bug in datasets.load_boston which had a wrong data point. #10795 by Takeshi
Yoshizawa.
• [F IX ] Fixed a bug in datasets.load_iris which had two wrong data points. #11082 by Sadhana Srini-
vasan and Hanmin Qin.
• [F IX ] Fixed a bug in datasets.fetch_kddcup99, where data were not properly shuffled. #9731 by Nico-
las Goix.
• [F IX ] Fixed a bug in datasets.make_circles, where no odd number of data points could be generated.
#10045 by Christian Braune.
• [API C HANGE ] Deprecated sklearn.datasets.fetch_mldata to be removed in version 0.22. ml-
data.org is no longer operational. Until removal it will remain possible to load cached datasets. #11466 by
Joel Nothman.
• [F EATURE ] decomposition.dict_learning functions and models now support positivity constraints.

This applies to the dictionary and sparse code. #6374 by John Kirkham.
• [F EATURE ] [F IX ] decomposition.SparsePCA now exposes normalize_components. When set to
True, the train and test data are centered with the train mean repsectively during the fit phase and the transform
phase. This fixes the behavior of SparsePCA. When set to False, which is the default, the previous abnormal
behaviour still holds. The False value is for backward compatibility and should not be used. #11585 by Ivan
Panico.
• [E FFICIENCY ] Efficiency improvements in decomposition.dict_learning. #11420 and others by John
Kirkham.
• [F IX ] Fix for uninformative error in decomposition.IncrementalPCA: now an error is raised if the
number of components is larger than the chosen batch size. The n_components=None case was adapted
accordingly. #6452. By Wally Gauze.
• [F IX ] Fixed a bug where the partial_fit method of decomposition.IncrementalPCA used integer
division instead of float division on Python 2. #9492 by James Bourbeau.

• [F IX ] In decomposition.PCA selecting a n_components parameter greater than the number of samples now
raises an error. Similarly, the n_components=None case now selects the minimum of n_samples and
n_features. #8484 by Wally Gauze.
• [F IX ] Fixed a bug in decomposition.PCA where users will get unexpected error with large datasets when
n_components='mle' on Python 3 versions. #9886 by Hanmin Qin.
• [F IX ] Fixed an underflow in calculating KL-divergence for decomposition.NMF #10142 by Tom Dupre la
Tour.
• [F IX ] Fixed a bug in decomposition.SparseCoder when running OMP sparse coding in parallel using
read-only memory mapped datastructures. #5956 by Vighnesh Birodkar and Olivier Grisel.
sklearn.discriminant_analysis
• [E FFICIENCY ] Memory usage improvement for _class_means and _class_cov in

discriminant_analysis. #10898 by Nanxin Chen.
sklearn.dummy
• [F EATURE ] dummy.DummyRegressor now has a return_std option in its predict method. The re-
turned standard deviations will be zeros.
• [F EATURE ] dummy.DummyClassifier and dummy.DummyRegressor now only require X to be an ob-
ject with finite length or shape. #9832 by Vrishank Bhardwaj.
• [F EATURE ] dummy.DummyClassifier and dummy.DummyRegressor can now be scored without sup-
plying test samples. #11951 by Rüdiger Busche.
sklearn.ensemble
• [F EATURE ] ensemble.BaggingRegressor and ensemble.BaggingClassifier can now be fit

with missing/non-finite values in X and/or multi-output Y to support wrapping pipelines that perform their
own imputation. #9707 by Jimmy Wan.
• [F EATURE ] ensemble.GradientBoostingClassifier and ensemble.
GradientBoostingRegressor now support early stopping via n_iter_no_change,
validation_fraction and tol. #7071 by Raghav RV
• [F EATURE ] Added named_estimators_ parameter in ensemble.VotingClassifier to access fitted
estimators. #9157 by Herilalaina Rakotoarison.
• [F IX ] Fixed a bug when fitting ensemble.GradientBoostingClassifier or ensemble.
GradientBoostingRegressor with warm_start=True which previously raised a segmentation fault
due to a non-conversion of CSC matrix into CSR format expected by decision_function. Similarly,
Fortran-ordered arrays are converted to C-ordered arrays in the dense case. #9991 by Guillaume Lemaitre.
• [F IX ] Fixed a bug in ensemble.GradientBoostingRegressor and ensemble.
GradientBoostingClassifier to have feature importances summed and then normalized, rather
than normalizing on a per-tree basis. The previous behavior over-weighted the Gini importance of features that
appear in later stages. This issue only affected feature importances. #11176 by Gil Forsyth.
• [API C HANGE ] The default value of the n_estimators parameter of ensemble.
RandomForestClassifier, ensemble.RandomForestRegressor, ensemble.
ExtraTreesClassifier, ensemble.ExtraTreesRegressor, and ensemble.
RandomTreesEmbedding will change from 10 in version 0.20 to 100 in 0.22. A FutureWarning is
raised when the default value is used. #11542 by Anna Ayzenshtat.
1.11. Version 0.20.0 35

• [API C HANGE ] Classes derived from ensemble.BaseBagging. The attribute estimators_samples_

will return a list of arrays containing the indices selected for each bootstrap instead of a list of arrays containing
the mask of the samples selected for each bootstrap. Indices allows to repeat samples while mask does not allow
this functionality. #9524 by Guillaume Lemaitre.
• [F IX ] ensemble.BaseBagging where one could not deterministically reproduce fit result using the object
attributes when random_state is set. #9723 by Guillaume Lemaitre.
• [F EATURE ] Enable the call to get_feature_names in unfitted feature_extraction.text.

CountVectorizer initialized with a vocabulary. #10908 by Mohamed Maskani.
• [E NHANCEMENT ] idf_ can now be set on a feature_extraction.text.TfidfTransformer.
#10899 by Sergey Melderis.
• [F IX ] Fixed a bug in feature_extraction.image.extract_patches_2d which would throw an
exception if max_patches was greater than or equal to the number of all possible patches rather than simply
returning the number of possible patches. #10101 by Varun Agrawal
• [F IX ] Fixed a bug in feature_extraction.text.CountVectorizer, feature_extraction.
text.TfidfVectorizer, feature_extraction.text.HashingVectorizer to support 64 bit
sparse array indexing necessary to process large datasets with more than 2·109 tokens (words or n-grams).
#9147 by Claes-Fredrik Mannby and Roman Yurchak.
• [F IX ] Fixed bug in feature_extraction.text.TfidfVectorizer which was ignoring the parame-
ter dtype. In addition, feature_extraction.text.TfidfTransformer will preserve dtype for
floating and raise a warning if dtype requested is integer. #10441 by Mayur Kulkarni and Guillaume Lemaitre.
sklearn.feature_selection
• [F EATURE ] Added select K best features functionality to feature_selection.SelectFromModel.

#6689 by Nihar Sheth and Quazi Rahman.
• [F EATURE ] Added min_features_to_select parameter to feature_selection.RFECV to bound
evaluated features counts. #11293 by Brent Yi.
• [F EATURE ] feature_selection.RFECV ’s fit method now supports groups. #9656 by Adam Greenhall.
• [F IX ] Fixed computation of n_features_to_compute for edge case with tied CV scores in
feature_selection.RFECV . #9222 by Nick Hoh.
sklearn.gaussian_process
• [E FFICIENCY ] In gaussian_process.GaussianProcessRegressor, method predict is faster

when using return_std=True in particular more when called several times in a row. #9234 by andrewww
and Minghui Liu.
sklearn.impute
• New module, adopting preprocessing.Imputer as impute.SimpleImputer with minor changes

(see under preprocessing below).
• [M AJOR F EATURE ] Added impute.MissingIndicator which generates a binary indicator for missing
values. #8075 by Maniteja Nandana and Guillaume Lemaitre.

• [F EATURE ] The impute.SimpleImputer has a new strategy, 'constant', to complete missing values
with a fixed one, given by the fill_value parameter. This strategy supports numeric and non-numeric data,
and so does the 'most_frequent' strategy now. #11211 by Jeremie du Boisberranger.
sklearn.isotonic
• [F IX ] Fixed a bug in isotonic.IsotonicRegression which incorrectly combined weights when fitting

a model to data involving points with identical X values. #9484 by Dallas Card
• [F EATURE ] linear_model.SGDClassifier, linear_model.SGDRegressor, linear_model.

PassiveAggressiveClassifier, linear_model.PassiveAggressiveRegressor and
linear_model.Perceptron now expose early_stopping, validation_fraction and
n_iter_no_change parameters, to stop optimization monitoring the score on a validation set. A new learn-
ing rate "adaptive" strategy divides the learning rate by 5 each time n_iter_no_change consecutive
epochs fail to improve the model. #9043 by Tom Dupre la Tour.
• [F EATURE ] Add sample_weight parameter to the fit method of linear_model.BayesianRidge for
weighted linear regression. #10112 by Peter St. John.
• [F IX ] Fixed a bug in logistic.logistic_regression_path to ensure that the returned coefficients
are correct when multiclass='multinomial'. Previously, some of the coefficients would override each
other, leading to incorrect results in linear_model.LogisticRegressionCV . #11724 by Nicolas Hug.
• [F IX ] Fixed a bug in linear_model.LogisticRegression where when using the parameter
multi_class='multinomial', the predict_proba method was returning incorrect probabilities in
the case of binary outcomes. #9939 by Roger Westover.
• [F IX ] Fixed a bug in linear_model.LogisticRegressionCV where the score method always com-
putes accuracy, not the metric given by the scoring parameter. #10998 by Thomas Fan.
• [F IX ] Fixed a bug in linear_model.LogisticRegressionCV where the ‘ovr’ strategy was always used
to compute cross-validation scores in the multiclass setting, even if 'multinomial' was set. #8720 by
William de Vazelhes.
• [F IX ] Fixed a bug in linear_model.OrthogonalMatchingPursuit that was broken when setting
normalize=False. #10071 by Alexandre Gramfort.
• [F IX ] Fixed a bug in linear_model.ARDRegression which caused incorrectly updated estimates for the
standard deviation and the coefficients. #10153 by Jörg Döpfert.
• [F IX ] Fixed a bug in linear_model.ARDRegression and linear_model.BayesianRidge which
caused NaN predictions when fitted with a constant target. #10095 by Jörg Döpfert.
• [F IX ] Fixed a bug in linear_model.RidgeClassifierCV where the parameter store_cv_values
was not implemented though it was documented in cv_values as a way to set up the storage of cross-
validation values for different alphas. #10297 by Mabel Villalba-Jiménez.
• [F IX ] Fixed a bug in linear_model.ElasticNet which caused the input to be overridden when using
parameter copy_X=True and check_input=False. #10581 by Yacine Mazari.
• [F IX ] Fixed a bug in sklearn.linear_model.Lasso where the coefficient had wrong shape when
fit_intercept=False. #10687 by Martin Hahn.
• [F IX ] Fixed a bug in sklearn.linear_model.LogisticRegression where the
multi_class='multinomial' with binary output with warm_start=True #10836 by Aish-
warya Srinivasan.
1.11. Version 0.20.0 37

• [F IX ] Fixed a bug in linear_model.RidgeCV where using integer alphas raised an error. #10397 by
Mabel Villalba-Jiménez.
• [F IX ] Fixed condition triggering gap computation in linear_model.Lasso and linear_model.
ElasticNet when working with sparse matrices. #10992 by Alexandre Gramfort.
• [F IX ] Fixed a bug in linear_model.SGDClassifier, linear_model.
SGDRegressor, linear_model.PassiveAggressiveClassifier, linear_model.
PassiveAggressiveRegressor and linear_model.Perceptron, where the stopping criterion
was stopping the algorithm before convergence. A parameter n_iter_no_change was added and set by
default to 5. Previous behavior is equivalent to setting the parameter to 1. #9043 by Tom Dupre la Tour.
• [F IX ] Fixed a bug where liblinear and libsvm-based estimators would segfault if passed a scipy.sparse matrix
with 64-bit indices. They now raise a ValueError. #11327 by Karan Dhingra and Joel Nothman.
• [API C HANGE ] The default values of the solver and multi_class parameters of linear_model.
LogisticRegression will change respectively from 'liblinear' and 'ovr' in version 0.20 to
'lbfgs' and 'auto' in version 0.22. A FutureWarning is raised when the default values are used. #11905
by Tom Dupre la Tour and Joel Nothman.
• [API C HANGE ] Deprecate positive=True option in linear_model.Lars as the underlying implemen-
tation is broken. Use linear_model.Lasso instead. #9837 by Alexandre Gramfort.
• [API C HANGE ] n_iter_ may vary from previous releases in linear_model.LogisticRegression
with solver='lbfgs' and linear_model.HuberRegressor. For Scipy <= 1.0.0, the optimizer could
perform more than the requested maximum number of iterations. Now both estimators will report at most
max_iter iterations even if more were performed. #10723 by Joel Nothman.
sklearn.manifold
• [E FFICIENCY ] Speed improvements for both ‘exact’ and ‘barnes_hut’ methods in manifold.TSNE. #10593
and #10610 by Tom Dupre la Tour.
• [F EATURE ] Support sparse input in manifold.Isomap.fit. #8554 by Leland McInnes.
• [F EATURE ] manifold.t_sne.trustworthiness accepts metrics other than Euclidean. #9775 by
William de Vazelhes.
• [F IX ] Fixed a bug in manifold.spectral_embedding where the normalization of the spectrum was
using a division instead of a multiplication. #8129 by Jan Margeta, Guillaume Lemaitre, and Devansh D..
• [API C HANGE ] [F EATURE ] Deprecate precomputed parameter in function manifold.t_sne.
trustworthiness. Instead, the new parameter metric should be used with any compatible metric in-
cluding ‘precomputed’, in which case the input matrix X should be a matrix of pairwise distances or squared
distances. #9775 by William de Vazelhes.
• [API C HANGE ] Deprecate precomputed parameter in function manifold.t_sne.trustworthiness.
Instead, the new parameter metric should be used with any compatible metric including ‘precomputed’, in
which case the input matrix X should be a matrix of pairwise distances or squared distances. #9775 by William
de Vazelhes.
sklearn.metrics
• [M AJOR F EATURE ] Added the metrics.davies_bouldin_score metric for evaluation of clustering mod-
els without a ground truth. #10827 by Luis Osa.
• [M AJOR F EATURE ] Added the metrics.balanced_accuracy_score metric and a corresponding
'balanced_accuracy' scorer for binary and multiclass classification. #8066 by @xyguo and Aman
Dalmia, and #10587 by Joel Nothman.

• [F EATURE ] Partial AUC is available via max_fpr parameter in metrics.roc_auc_score. #3840 by

Alexander Niederbühl.
• [F EATURE ] A scorer based on metrics.brier_score_loss is also available. #9521 by Hanmin Qin.
• [F EATURE ] Added control over the normalization in metrics.normalized_mutual_info_score and
metrics.adjusted_mutual_info_score via the average_method parameter. In version 0.22, the
default normalizer for each will become the arithmetic mean of the entropies of each clustering. #11124 by
Arya McCarthy.
• [F EATURE ] Added output_dict parameter in metrics.classification_report to return classifi-
cation statistics as dictionary. #11160 by Dan Barkhorn.
• [F EATURE ] metrics.classification_report now reports all applicable averages on the given data, in-
cluding micro, macro and weighted average as well as samples average for multilabel data. #11679 by Alexander
Pacha.
• [F EATURE ] metrics.average_precision_score now supports binary y_true other than {0, 1} or
{-1, 1} through pos_label parameter. #9980 by Hanmin Qin.
• [F EATURE ] metrics.label_ranking_average_precision_score now supports
sample_weight. #10845 by Jose Perez-Parras Toledano.
• [F EATURE ] Add dense_output parameter to metrics.pairwise.linear_kernel. When False and
both inputs are sparse, will return a sparse matrix. #10999 by Taylor G Smith.
• [E FFICIENCY ] metrics.silhouette_score and metrics.silhouette_samples are more mem-
ory efficient and run faster. This avoids some reported freezes and MemoryErrors. #11135 by Joel Nothman.
• [F IX ] Fixed a bug in metrics.precision_recall_fscore_support when truncated range(n_labels)
is passed as value for labels. #10377 by Gaurav Dhingra.
• [F IX ] Fixed a bug due to floating point error in metrics.roc_auc_score with non-integer sample weights.
#9786 by Hanmin Qin.
• [F IX ] Fixed a bug where metrics.roc_curve sometimes starts on y-axis instead of (0, 0), which is in-
consistent with the document and other implementations. Note that this will not influence the result from
metrics.roc_auc_score #10093 by alexryndin and Hanmin Qin.
• [F IX ] Fixed a bug to avoid integer overflow. Casted product to 64 bits integer in metrics.
mutual_info_score. #9772 by Kumar Ashutosh.
• [F IX ] Fixed a bug where metrics.average_precision_score will sometimes return nan when
sample_weight contains 0. #9980 by Hanmin Qin.
• [F IX ] Fixed a bug in metrics.fowlkes_mallows_score to avoid integer overflow. Casted return value
of contingency_matrix to int64 and computed product of square roots rather than square root of product. #9515
by Alan Liddell and Manh Dao.
• [API C HANGE ] Deprecate reorder parameter in metrics.auc as it’s no longer required for metrics.
roc_auc_score. Moreover using reorder=True can hide bugs due to floating point error in the input.
#9851 by Hanmin Qin.
• [API C HANGE ] In metrics.normalized_mutual_info_score and metrics.
adjusted_mutual_info_score, warn that average_method will have a new default value.
In version 0.22, the default normalizer for each will become the arithmetic mean of the entropies of
each clustering. Currently, metrics.normalized_mutual_info_score uses the default of
average_method='geometric', and metrics.adjusted_mutual_info_score uses the
default of average_method='max' to match their behaviors in version 0.19. #11124 by Arya McCarthy.
• [API C HANGE ] The batch_size parameter to metrics.pairwise_distances_argmin_min and
metrics.pairwise_distances_argmin is deprecated to be removed in v0.22. It no longer has any ef-
1.11. Version 0.20.0 39

fect, as batch size is determined by global working_memory config. See Limiting Working Memory. #10280
by Joel Nothman and Aman Dalmia.
sklearn.mixture
• [F EATURE ] Added function fit_predict to mixture.GaussianMixture and mixture.

GaussianMixture, which is essentially equivalent to calling fit and predict. #10336 by Shu Haoran
and Andrew Peng.
• [F IX ] Fixed a bug in mixture.BaseMixture where the reported n_iter_ was missing an iteration. It af-
fected mixture.GaussianMixture and mixture.BayesianGaussianMixture. #10740 by Erich
Schubert and Guillaume Lemaitre.
• [F IX ] Fixed a bug in mixture.BaseMixture and its subclasses mixture.GaussianMixture and
mixture.BayesianGaussianMixture where the lower_bound_ was not the max lower bound across
all initializations (when n_init > 1), but just the lower bound of the last initialization. #10869 by Aurélien
Géron.
sklearn.model_selection
• [F EATURE ] Add return_estimator parameter in model_selection.cross_validate to return estimators

fitted on each split. #9686 by Aurélien Bellet.
• [F EATURE ] New refit_time_ attribute will be stored in model_selection.GridSearchCV and
model_selection.RandomizedSearchCV if refit is set to True. This will allow measuring the
complete time it takes to perform hyperparameter optimization and refitting the best model on the whole dataset.
#11310 by Matthias Feurer.
• [F EATURE ] Expose error_score parameter in model_selection.cross_validate,
model_selection.cross_val_score, model_selection.learning_curve and
model_selection.validation_curve to control the behavior triggered when an error occurs in
model_selection._fit_and_score. #11576 by Samuel O. Ronsin.
• [F EATURE ] BaseSearchCV now has an experimental, private interface to support customized parameter
search strategies, through its _run_search method. See the implementations in model_selection.
GridSearchCV and model_selection.RandomizedSearchCV and please provide feedback if you
use this. Note that we do not assure the stability of this API beyond version 0.20. #9599 by Joel Nothman
• [E NHANCEMENT ] Add improved error message in model_selection.cross_val_score when multiple
metrics are passed in scoring keyword. #11006 by Ming Li.
• [API C HANGE ] The default number of cross-validation folds cv and the default number of splits n_splits in
the model_selection.KFold-like splitters will change from 3 to 5 in 0.22 as 3-fold has a lot of variance.
#11557 by Alexandre Boucaud.
• [API C HANGE ] The default of iid parameter of model_selection.GridSearchCV and
model_selection.RandomizedSearchCV will change from True to False in version 0.22 to
correspond to the standard definition of cross-validation, and the parameter will be removed in version
0.24 altogether. This parameter is of greatest practical significance where the sizes of different test sets in
cross-validation were very unequal, i.e. in group-based CV strategies. #9085 by Laurent Direr and Andreas
Müller.
• [API C HANGE ] The default value of the error_score parameter in model_selection.GridSearchCV
and model_selection.RandomizedSearchCV will change to np.NaN in version 0.22. #10677 by
Kirill Zhdanovich.

• [API C HANGE ] Changed ValueError exception raised in model_selection.ParameterSampler to a

UserWarning for case where the class is instantiated with a greater value of n_iter than the total space of
parameters in the parameter grid. n_iter now acts as an upper bound on iterations. #10982 by Juliet Lawton
• [API C HANGE ] Invalid input for model_selection.ParameterGrid now raises TypeError. #10928 by
Solutus Immensus
sklearn.multioutput
• [M AJOR F EATURE ] Added multioutput.RegressorChain for multi-target regression. #9257 by Kumar

Ashutosh.
sklearn.naive_bayes
• [M AJOR F EATURE ] Added naive_bayes.ComplementNB, which implements the Complement Naive Bayes
classifier described in Rennie et al. (2003). #8190 by Michael A. Alcorn.
• [F EATURE ] Add var_smoothing parameter in naive_bayes.GaussianNB to give a precise control over
variances calculation. #9681 by Dmitry Mottl.
• [F IX ] Fixed a bug in naive_bayes.GaussianNB which incorrectly raised error for prior list which summed
to 1. #10005 by Gaurav Dhingra.
• [F IX ] Fixed a bug in naive_bayes.MultinomialNB which did not accept vector valued pseudocounts
(alpha). #10346 by Tobias Madsen
sklearn.neighbors
• [E FFICIENCY ] neighbors.RadiusNeighborsRegressor and neighbors.

RadiusNeighborsClassifier are now parallelized according to n_jobs regardless of algorithm.
#10887 by Joël Billaud.
• [E FFICIENCY ] Nearest neighbors query methods are now more memory efficient when
algorithm='brute'. #11136 by Joel Nothman and Aman Dalmia.
• [F EATURE ] Add sample_weight parameter to the fit method of neighbors.KernelDensity to enable
weighting in kernel density estimation. #4394 by Samuel O. Ronsin.
• [F EATURE ] Novelty detection with neighbors.LocalOutlierFactor: Add a novelty param-
eter to neighbors.LocalOutlierFactor. When novelty is set to True, neighbors.
LocalOutlierFactor can then be used for novelty detection, i.e. predict on new unseen data. Available
prediction methods are predict, decision_function and score_samples. By default, novelty is
set to False, and only the fit_predict method is avaiable. By Albert Thomas.
• [F IX ] Fixed a bug in neighbors.NearestNeighbors where fitting a NearestNeighbors model fails when
a) the distance metric used is a callable and b) the input to the NearestNeighbors model is sparse. #9579 by
Thomas Kober.
• [F IX ] Fixed a bug so predict in neighbors.RadiusNeighborsRegressor can handle empty neigh-
bor set when using non uniform weights. Also raises a new warning when no neighbors are found for samples.
#9655 by Andreas Bjerre-Nielsen.
• [F IX ] [E FFICIENCY ] Fixed a bug in KDTree construction that results in faster construction and querying times.
#11556 by Jake VanderPlas
• [F IX ] Fixed a bug in neighbors.KDTree and neighbors.BallTree where pickled tree objects would
change their type to the super class BinaryTree. #11774 by Nicolas Hug.
1.11. Version 0.20.0 41

sklearn.neural_network
• [F EATURE ] Add n_iter_no_change parameter in neural_network.BaseMultilayerPerceptron,

neural_network.MLPRegressor, and neural_network.MLPClassifier to give control over
maximum number of epochs to not meet tol improvement. #9456 by Nicholas Nadeau.
• [F IX ] Fixed a bug in neural_network.BaseMultilayerPerceptron, neural_network.
MLPRegressor, and neural_network.MLPClassifier with new n_iter_no_change parameter
now at 10 from previously hardcoded 2. #9456 by Nicholas Nadeau.
• [F IX ] Fixed a bug in neural_network.MLPRegressor where fitting quit unexpectedly early due to local
minima or fluctuations. #9456 by Nicholas Nadeau
sklearn.pipeline
• [F EATURE ] The predict method of pipeline.Pipeline now passes keyword arguments on to the
pipeline’s last estimator, enabling the use of parameters such as return_std in a pipeline with caution.
#9304 by Breno Freitas.
• [API C HANGE ] pipeline.FeatureUnion now supports 'drop' as a transformer to drop features. #11144
by Thomas Fan.
• [M AJOR F EATURE ] Expanded preprocessing.OneHotEncoder to allow to encode categorical string

features as a numeric array using a one-hot (or dummy) encoding scheme, and added preprocessing.
OrdinalEncoder to convert to ordinal integers. Those two classes now handle encoding of all feature types
(also handles string-valued features) and derives the categories based on the unique values in the features instead
of the maximum value in the features. #9151 and #10521 by Vighnesh Birodkar and Joris Van den Bossche.
• [M AJOR F EATURE ] Added preprocessing.KBinsDiscretizer for turning continuous features into cat-
egorical or one-hot encoded features. #7668, #9647, #10195, #10192, #11272, #11467 and #11505. by Henry
Lin, Hanmin Qin, Tom Dupre la Tour and Giovanni Giuseppe Costa.
• [M AJOR F EATURE ] Added preprocessing.PowerTransformer, which implements the Yeo-Johnson
and Box-Cox power transformations. Power transformations try to find a set of feature-wise parametric trans-
formations to approximately map data to a Gaussian distribution centered at zero and with unit variance. This is
useful as a variance-stabilizing transformation in situations where normality and homoscedasticity are desirable.
#10210 by Eric Chang and Maniteja Nandana, and #11520 by Nicolas Hug.
• [M AJOR F EATURE ] NaN values are ignored and handled in the following preprocessing methods:
preprocessing.MaxAbsScaler, preprocessing.MinMaxScaler, preprocessing.
RobustScaler, preprocessing.StandardScaler, preprocessing.PowerTransformer,
preprocessing.QuantileTransformer classes and preprocessing.maxabs_scale,
preprocessing.minmax_scale, preprocessing.robust_scale, preprocessing.scale,
preprocessing.power_transform, preprocessing.quantile_transform functions respec-
tively addressed in issues #11011, #11005, #11308, #11206, #11306, and #10437. By Lucija Gregov and
Guillaume Lemaitre.
• [F EATURE ] preprocessing.PolynomialFeatures now supports sparse input. #10452 by Aman
Dalmia and Joel Nothman.
• [F EATURE ] preprocessing.RobustScaler and preprocessing.robust_scale can be fitted us-
ing sparse matrices. #11308 by Guillaume Lemaitre.
• [F EATURE ] preprocessing.OneHotEncoder now supports the get_feature_names method to obtain the
transformed feature names. #10181 by Nirvan Anjirbag and Joris Van den Bossche.

• [F EATURE ] A parameter check_inverse was added to preprocessing.FunctionTransformer to

ensure that func and inverse_func are the inverse of each other. #9399 by Guillaume Lemaitre.
• [F EATURE ] The transform method of sklearn.preprocessing.MultiLabelBinarizer now ig-
nores any unknown classes. A warning is raised stating the unknown classes classes found which are ignored.
#10913 by Rodrigo Agundez.
• [F IX ] Fixed bugs in preprocessing.LabelEncoder which would sometimes throw errors when
transform or inverse_transform was called with empty arrays. #10458 by Mayur Kulkarni.
• [F IX ] Fix ValueError in preprocessing.LabelEncoder when using inverse_transform on unseen
labels. #9816 by Charlie Newey.
• [F IX ] Fix bug in preprocessing.OneHotEncoder which discarded the dtype when returning a sparse
matrix output. #11042 by Daniel Morales.
• [F IX ] Fix fit and partial_fit in preprocessing.StandardScaler in the rare case when
with_mean=False and with_std=False which was crashing by calling fit more than once and giving in-
consistent results for mean_ whether the input was a sparse or a dense matrix. mean_ will be set to None
with both sparse and dense inputs. n_samples_seen_ will be also reported for both input types. #11235 by
Guillaume Lemaitre.
• [API C HANGE ] Deprecate n_values and categorical_features parameters and
active_features_, feature_indices_ and n_values_ attributes of preprocessing.
OneHotEncoder. The n_values parameter can be replaced with the new categories pa-
rameter, and the attributes with the new categories_ attribute. Selecting the categorical fea-
tures with the categorical_features parameter is now better supported using the compose.
ColumnTransformer. #10521 by Joris Van den Bossche.
• [API C HANGE ] Deprecate preprocessing.Imputer and move the corresponding module to impute.
SimpleImputer. #9726 by Kumar Ashutosh.
• [API C HANGE ] The axis parameter that was in preprocessing.Imputer is no longer present in
impute.SimpleImputer. The behavior is equivalent to axis=0 (impute along columns). Row-
wise imputation can be performed with FunctionTransformer (e.g., FunctionTransformer(lambda X:
SimpleImputer().fit_transform(X.T).T)). #10829 by Guillaume Lemaitre and Gilberto Olimpio.
• [API C HANGE ] The NaN marker for the missing values has been changed between the preprocessing.
Imputer and the impute.SimpleImputer. missing_values='NaN' should now be
missing_values=np.nan. #11211 by Jeremie du Boisberranger.
• [API C HANGE ] In preprocessing.FunctionTransformer, the default of validate will be from
True to False in 0.22. #10655 by Guillaume Lemaitre.
sklearn.svm
• [F IX ] Fixed a bug in svm.SVC where when the argument kernel is unicode in Python2, the
predict_proba method was raising an unexpected TypeError given dense inputs. #10412 by Jiongyan
Zhang.
• [API C HANGE ] Deprecate random_state parameter in svm.OneClassSVM as the underlying implemen-
tation is not random. #9497 by Albert Thomas.
• [API C HANGE ] The default value of gamma parameter of svm.SVC, NuSVC, SVR, NuSVR, OneClassSVM
will change from 'auto' to 'scale' in version 0.22 to account better for unscaled features. #8361 by
Gaurav Dhingra and Ting Neo.
1.11. Version 0.20.0 43

sklearn.tree
• [E NHANCEMENT ] Although private (and hence not assured API stability), tree._criterion.
ClassificationCriterion and tree._criterion.RegressionCriterion may now be cim-
ported and extended. #10325 by Camil Staps.
• [F IX ] Fixed a bug in tree.BaseDecisionTree with splitter=”best” where split threshold could become
infinite when values in X were near infinite. #10536 by Jonathan Ohayon.
• [F IX ] Fixed a bug in tree.MAE to ensure sample weights are being used during the calculation of tree MAE im-
purity. Previous behaviour could cause suboptimal splits to be chosen since the impurity calculation considered
all samples to be of equal weight importance. #11464 by John Stott.
sklearn.utils
• [F EATURE ] utils.check_array and utils.check_X_y now have accept_large_sparse to con-

trol whether scipy.sparse matrices with 64-bit indices should be rejected. #11327 by Karan Dhingra and Joel
Nothman.
• [E FFICIENCY ] [F IX ] Avoid copying the data in utils.check_array when the input data is a memmap (and
copy=False). #10663 by Arthur Mensch and Loïc Estève.
• [API C HANGE ] utils.check_array yield a FutureWarning indicating that arrays of bytes/strings will
be interpreted as decimal numbers beginning in version 0.22. #10229 by Ryan Lee
Multiple modules
• [F EATURE ] [API C HANGE ] More consistent outlier detection API: Add a score_samples method
in svm.OneClassSVM , ensemble.IsolationForest, neighbors.LocalOutlierFactor,
covariance.EllipticEnvelope. It allows to access raw score functions from original pa-
pers. A new offset_ parameter allows to link score_samples and decision_function
methods. The contamination parameter of ensemble.IsolationForest and neighbors.
LocalOutlierFactor decision_function methods is used to define this offset_ such that outliers
(resp. inliers) have negative (resp. positive) decision_function values. By default, contamination
is kept unchanged to 0.1 for a deprecation period. In 0.22, it will be set to “auto”, thus using method-specific
score offsets. In covariance.EllipticEnvelope decision_function method, the raw_values
parameter is deprecated as the shifted Mahalanobis distance will be always returned in 0.22. #9015 by Nicolas
Goix.
• [F EATURE ] [API C HANGE ] A behaviour parameter has been introduced in ensemble.
IsolationForest to ensure backward compatibility. In the old behaviour, the decision_function is
independent of the contamination parameter. A threshold attribute depending on the contamination
parameter is thus used. In the new behaviour the decision_function is dependent on the
contamination parameter, in such a way that 0 becomes its natural threshold to detect outliers. Set-
ting behaviour to “old” is deprecated and will not be possible in version 0.22. Beside, the behaviour parameter
will be removed in 0.24. #11553 by Nicolas Goix.
• [API C HANGE ] Added convergence warning to svm.LinearSVC and linear_model.
LogisticRegression when verbose is set to 0. #10881 by Alexandre Sevin.
• [API C HANGE ] Changed warning type from UserWarning to exceptions.ConvergenceWarning
for failing convergence in linear_model.logistic_regression_path, linear_model.
RANSACRegressor, linear_model.ridge_regression, gaussian_process.
GaussianProcessRegressor, gaussian_process.GaussianProcessClassifier,
decomposition.fastica, cross_decomposition.PLSCanonical, cluster.
AffinityPropagation, and cluster.Birch. #10306 by Jonathan Siebert.

Miscellaneous
• [M AJOR F EATURE ] A new configuration parameter, working_memory was added to control memory con-
sumption limits in chunked operations, such as the new metrics.pairwise_distances_chunked. See
Limiting Working Memory. #10280 by Joel Nothman and Aman Dalmia.
• [F EATURE ] The version of joblib bundled with Scikit-learn is now 0.12. This uses a new default multiprocess-
ing implementation, named loky. While this may incur some memory and communication overhead, it should
provide greater cross-platform stability than relying on Python standard library multiprocessing. #11741 by the
Joblib developers, especially Thomas Moreau and Olivier Grisel.
• [F EATURE ] An environment variable to use the site joblib instead of the vendored one was added (Environment
variables). The main API of joblib is now exposed in sklearn.utils. #11166 by Gael Varoquaux.
• [F EATURE ] Add almost complete PyPy 3 support. Known unsupported functionalities are datasets.
load_svmlight_file, feature_extraction.FeatureHasher and feature_extraction.
text.HashingVectorizer. For running on PyPy, PyPy3-v5.10+, Numpy 1.14.0+, and scipy 1.1.0+ are
required. #11010 by Ronan Lamy and Roman Yurchak.
• [F EATURE ] A utility method sklearn.show_versions was added to print out information relevant for
debugging. It includes the user system, the Python executable, the version of the main libraries and BLAS
binding information. #11596 by Alexandre Boucaud
• [F IX ] Fixed a bug when setting parameters on meta-estimator, involving both a wrapped estimator and its pa-
rameter. #9999 by Marcus Voss and Joel Nothman.
• [F IX ] Fixed a bug where calling sklearn.base.clone was not thread safe and could result in a “pop from
empty list” error. #9569 by Andreas Müller.
• [API C HANGE ] The default value of n_jobs is changed from 1 to None in all related functions and classes.
n_jobs=None means unset. It will generally be interpreted as n_jobs=1, unless the current joblib.
Parallel backend context specifies otherwise (See Glossary for additional information). Note that this
change happens immediately (i.e., without a deprecation cycle). #11741 by Olivier Grisel.
• [F IX ] Fixed a bug in validation helpers where passing a Dask DataFrame results in an error. #12462 by Zachariah
Miller
1.11.5 Changes to estimator checks
These changes mostly affect library developers.

• Checks for transformers now apply if the estimator implements transform, regardless of whether it inherits from
sklearn.base.TransformerMixin. #10474 by Joel Nothman.
• Classifiers are now checked for consistency between decision_function and categorical predictions. #10500 by
Narine Kokhlikyan.
• Allow tests in utils.estimator_checks.check_estimator to test functions that accept pairwise
data. #9701 by Kyle Johnson
• Allow utils.estimator_checks.check_estimator to check that there is no private settings apart
from parameters during estimator initialization. #9378 by Herilalaina Rakotoarison
• The set of checks in utils.estimator_checks.check_estimator now includes a
check_set_params test which checks that set_params is equivalent to passing parameters in
__init__ and warns if it encounters parameter validation. #7738 by Alvin Chiang
• Add invariance tests for clustering metrics. #8102 by Ankita Sinha and Guillaume Lemaitre.
• Add check_methods_subset_invariance to check_estimator, which checks that estimator
methods are invariant if applied to a data subset. #10428 by Jonathan Ohayon
1.11. Version 0.20.0 45

• Add tests in utils.estimator_checks.check_estimator to check that an estimator can handle

read-only memmap input data. #10663 by Arthur Mensch and Loïc Estève.
• check_sample_weights_pandas_series now uses 8 rather than 6 samples to accommodate for the
default number of clusters in cluster.KMeans. #10933 by Johannes Hansen.
• Estimators are now checked for whether sample_weight=None equates to sample_weight=np.
ones(...). #11558 by Sergul Aydore.
Thanks to everyone who has contributed to the maintenance and improvement of the project since version 0.19, in-
cluding:
211217613, Aarshay Jain, absolutelyNoWarranty, Adam Greenhall, Adam Kleczewski, Adam Richie-Halford, adelr,
AdityaDaflapurkar, Adrin Jalali, Aidan Fitzgerald, aishgrt1, Akash Shivram, Alan Liddell, Alan Yee, Albert Thomas,
Alexander Lenail, Alexander-N, Alexandre Boucaud, Alexandre Gramfort, Alexandre Sevin, Alex Egg, Alvaro Perez-
Diaz, Amanda, Aman Dalmia, Andreas Bjerre-Nielsen, Andreas Mueller, Andrew Peng, Angus Williams, Aniruddha
Dave, annaayzenshtat, Anthony Gitter, Antonio Quinonez, Anubhav Marwaha, Arik Pamnani, Arthur Ozga, Artiem
K, Arunava, Arya McCarthy, Attractadore, Aurélien Bellet, Aurélien Geron, Ayush Gupta, Balakumaran Manoha-
ran, Bangda Sun, Barry Hart, Bastian Venthur, Ben Lawson, Benn Roth, Breno Freitas, Brent Yi, brett koonce,
Caio Oliveira, Camil Staps, cclauss, Chady Kamar, Charlie Brummitt, Charlie Newey, chris, Chris, Chris Catalfo,
Chris Foster, Chris Holdgraf, Christian Braune, Christian Hirsch, Christian Hogan, Christopher Jenness, Clement
Joudet, cnx, cwitte, Dallas Card, Dan Barkhorn, Daniel, Daniel Ferreira, Daniel Gomez, Daniel Klevebring, Danielle
Shwed, Daniel Mohns, Danil Baibak, Darius Morawiec, David Beach, David Burns, David Kirkby, David Nichol-
son, David Pickup, Derek, Didi Bar-Zev, diegodlh, Dillon Gardner, Dillon Niederhut, dilutedsauce, dlovell, Dmitry
Mottl, Dmitry Petrov, Dor Cohen, Douglas Duhaime, Ekaterina Tuzova, Eric Chang, Eric Dean Sanchez, Erich Schu-
bert, Eunji, Fang-Chieh Chou, FarahSaeed, felix, Félix Raimundo, fenx, filipj8, FrankHui, Franz Wompner, Freija
Descamps, frsi, Gabriele Calvo, Gael Varoquaux, Gaurav Dhingra, Georgi Peev, Gil Forsyth, Giovanni Giuseppe
Costa, gkevinyen5418, goncalo-rodrigues, Gryllos Prokopis, Guillaume Lemaitre, Guillaume “Vermeille” Sanchez,
Gustavo De Mari Pereira, hakaa1, Hanmin Qin, Henry Lin, Hong, Honghe, Hossein Pourbozorg, Hristo, Hunan Ros-
tomyan, iampat, Ivan PANICO, Jaewon Chung, Jake VanderPlas, jakirkham, James Bourbeau, James Malcolm, Jamie
Cox, Jan Koch, Jan Margeta, Jan Schlüter, janvanrijn, Jason Wolosonovich, JC Liu, Jeb Bearer, jeremiedbb, Jimmy
Wan, Jinkun Wang, Jiongyan Zhang, jjabl, jkleint, Joan Massich, Joël Billaud, Joel Nothman, Johannes Hansen,
JohnStott, Jonatan Samoocha, Jonathan Ohayon, Jörg Döpfert, Joris Van den Bossche, Jose Perez-Parras Toledano,
josephsalmon, jotasi, jschendel, Julian Kuhlmann, Julien Chaumond, julietcl, Justin Shenk, Karl F, Kasper Primdal
Lauritzen, Katrin Leinweber, Kirill, ksemb, Kuai Yu, Kumar Ashutosh, Kyeongpil Kang, Kye Taylor, kyledrogo,
Leland McInnes, Léo DS, Liam Geron, Liutong Zhou, Lizao Li, lkjcalc, Loic Esteve, louib, Luciano Viola, Lucija
Gregov, Luis Osa, Luis Pedro Coelho, Luke M Craig, Luke Persola, Mabel, Mabel Villalba, Maniteja Nandana, MarkI-
wanchyshyn, Mark Roth, Markus Müller, MarsGuy, Martin Gubri, martin-hahn, martin-kokos, mathurinm, Matthias
Feurer, Max Copeland, Mayur Kulkarni, Meghann Agarwal, Melanie Goetz, Michael A. Alcorn, Minghui Liu, Ming
Li, Minh Le, Mohamed Ali Jamaoui, Mohamed Maskani, Mohammad Shahebaz, Muayyad Alsadi, Nabarun Pal, Na-
garjuna Kumar, Naoya Kanai, Narendran Santhanam, NarineK, Nathaniel Saul, Nathan Suh, Nicholas Nadeau, P.Eng.,
AVS, Nick Hoh, Nicolas Goix, Nicolas Hug, Nicolau Werneck, nielsenmarkus11, Nihar Sheth, Nikita Titov, Nilesh
Kevlani, Nirvan Anjirbag, notmatthancock, nzw, Oleksandr Pavlyk, oliblum90, Oliver Rausch, Olivier Grisel, Oren
Milman, Osaid Rehman Nasir, pasbi, Patrick Fernandes, Patrick Olden, Paul Paczuski, Pedro Morales, Peter, Peter St.
John, pierreablin, pietruh, Pinaki Nath Chowdhury, Piotr Szymański, Pradeep Reddy Raamana, Pravar D Mahajan,
pravarmahajan, QingYing Chen, Raghav RV, Rajendra arora, RAKOTOARISON Herilalaina, Rameshwar Bhaskaran,
RankyLau, Rasul Kerimov, Reiichiro Nakano, Rob, Roman Kosobrodov, Roman Yurchak, Ronan Lamy, rragundez,
Rüdiger Busche, Ryan, Sachin Kelkar, Sagnik Bhattacharya, Sailesh Choyal, Sam Radhakrishnan, Sam Steingold,
Samuel Bell, Samuel O. Ronsin, Saqib Nizam Shamsi, SATISH J, Saurabh Gupta, Scott Gigante, Sebastian Flen-
nerhag, Sebastian Raschka, Sebastien Dubois, Sébastien Lerique, Sebastin Santy, Sergey Feldman, Sergey Melderis,
Sergul Aydore, Shahebaz, Shalil Awaley, Shangwu Yao, Sharad Vijalapuram, Sharan Yalburgi, shenhanc78, Shivam
Rastogi, Shu Haoran, siftikha, Sinclert Pérez, SolutusImmensus, Somya Anand, srajan paliwal, Sriharsha Hatwar, Sri
Krishna, Stefan van der Walt, Stephen McDowell, Steven Brown, syonekura, Taehoon Lee, Takanori Hayashi, tarcusx,

Taylor G Smith, theriley106, Thomas, Thomas Fan, Thomas Heavey, Tobias Madsen, tobycheese, Tom Augspurger,
Tom Dupré la Tour, Tommy, Trevor Stephens, Trishnendu Ghorai, Tulio Casagrande, twosigmajab, Umar Farouk
Umar, Urvang Patel, Utkarsh Upadhyay, Vadim Markovtsev, Varun Agrawal, Vathsala Achar, Vilhelm von Ehren-
heim, Vinayak Mehta, Vinit, Vinod Kumar L, Viraj Mavani, Viraj Navkal, Vivek Kumar, Vlad Niculae, vqean3, Vris-
hank Bhardwaj, vufg, wallygauze, Warut Vijitbenjaronk, wdevazelhes, Wenhao Zhang, Wes Barnett, Will, William de
Vazelhes, Will Rosenfeld, Xin Xiong, Yiming (Paul) Li, ymazari, Yufeng, Zach Griffith, Zé Vinícius, Zhenqing Hu,
Zhiqing Xiao, Zijie (ZJ) Poh
1.12 Version 0.19.2
July, 2018
This release is exclusively in order to support Python 3.7.
1.12.1 Related changes
• n_iter_ may vary from previous releases in linear_model.LogisticRegression with

solver='lbfgs' and linear_model.HuberRegressor. For Scipy <= 1.0.0, the optimizer could
perform more than the requested maximum number of iterations. Now both estimators will report at most
max_iter iterations even if more were performed. #10723 by Joel Nothman.
1.13 Version 0.19.1
October 23, 2017

0.19.0.
Note there may be minor differences in TSNE output in this release (due to #9623), in the case where multiple samples
have equal distance to some sample.
1.13.1 Changelog
API changes
• Reverted the addition of metrics.ndcg_score and metrics.dcg_score which had been merged into
version 0.19.0 by error. The implementations were broken and undocumented.
• return_train_score which was added to model_selection.GridSearchCV ,
model_selection.RandomizedSearchCV and model_selection.cross_validate in
version 0.19.0 will be changing its default value from True to False in version 0.21. We found that calculating
training score could have a great effect on cross validation runtime in some cases. Users should explicitly
set return_train_score to False if prediction or scoring functions are slow, resulting in a deleterious
effect on CV runtime, or to True if they wish to use the calculated scores. #9677 by Kumar Ashutosh and Joel
Nothman.
• correlation_models and regression_models from the legacy gaussian processes implementation
have been belatedly deprecated. #9717 by Kumar Ashutosh.
1.12. Version 0.19.2 47

Bug fixes
• Avoid integer overflows in metrics.matthews_corrcoef. #9693 by Sam Steingold.

• Fixed a bug in the objective function for manifold.TSNE (both exact and with the Barnes-Hut approximation)
when n_components >= 3. #9711 by @goncalo-rodrigues.
• Fix regression in model_selection.cross_val_predict where it raised an error with
method='predict_proba' for some probabilistic classifiers. #9641 by James Bourbeau.
• Fixed a bug where datasets.make_classification modified its input weights. #9865 by Sachin
Kelkar.
• model_selection.StratifiedShuffleSplit now works with multioutput multiclass or multilabel
data with more than 1000 columns. #9922 by Charlie Brummitt.
• Fixed a bug with nested and conditional parameter setting, e.g. setting a pipeline step and its parameter at the
same time. #9945 by Andreas Müller and Joel Nothman.
Regressions in 0.19.0 fixed in 0.19.1:
• Fixed a bug where parallelised prediction in random forests was not thread-safe and could (rarely) result in
arbitrary errors. #9830 by Joel Nothman.
• Fix regression in model_selection.cross_val_predict where it no longer accepted X as a list.
#9600 by Rasul Kerimov.
• Fixed handling of cross_val_predict for binary classification with
method='decision_function'. #9593 by Reiichiro Nakano and core devs.
• Fix regression in pipeline.Pipeline where it no longer accepted steps as a tuple. #9604 by Joris Van
den Bossche.
• Fix bug where n_iter was not properly deprecated, leaving n_iter unavailable for interim use
in linear_model.SGDClassifier, linear_model.SGDRegressor, linear_model.
linear_model.Perceptron. #9558 by Andreas Müller.
• Dataset fetchers make sure temporary files are closed before removing them, which caused errors on Windows.
#9847 by Joan Massich.
• Fixed a regression in manifold.TSNE where it no longer supported metrics other than ‘euclidean’ and ‘pre-
computed’. #9623 by Oli Blum.
Enhancements
• Our test suite and utils.estimator_checks.check_estimators can now be run without Nose in-
stalled. #9697 by Joan Massich.
• To improve usability of version 0.19’s pipeline.Pipeline caching, memory now allows joblib.
Memory instances. This make use of the new utils.validation.check_memory helper. issue:9584
by Kumar Ashutosh
• Some fixes to examples: #9750, #9788, #9815
• Made a FutureWarning in SGD-based estimators less verbose. #9802 by Vrishank Bhardwaj.
With thanks to:

Joel Nothman, Loic Esteve, Andreas Mueller, Kumar Ashutosh, Vrishank Bhardwaj, Hanmin Qin, Rasul Kerimov,
James Bourbeau, Nagarjuna Kumar, Nathaniel Saul, Olivier Grisel, Roman Yurchak, Reiichiro Nakano, Sachin Kelkar,
Sam Steingold, Yaroslav Halchenko, diegodlh, felix, goncalo-rodrigues, jkleint, oliblum90, pasbi, Anthony Gitter, Ben
Lawson, Charlie Brummitt, Didi Bar-Zev, Gael Varoquaux, Joan Massich, Joris Van den Bossche, nielsenmarkus11
1.14 Version 0.19
August 12, 2017
1.14.1 Highlights
We are excited to release a number of great new features including neighbors.LocalOutlierFactor

for anomaly detection, preprocessing.QuantileTransformer for robust feature transformation, and
the multioutput.ClassifierChain meta-estimator to simply account for dependencies between classes
in multilabel problems. We have some new algorithms in existing estimators, such as multiplicative up-
date in decomposition.NMF and multinomial linear_model.LogisticRegression with L1 loss (use
solver='saga').
Cross validation is now able to return the results from multiple metric evaluations. The new model_selection.
cross_validate can return many scores on the test data as well as training set performance and timings, and we
have extended the scoring and refit parameters for grid/randomized search to handle multiple metrics.
You can also learn faster. For instance, the new option to cache transformations in pipeline.Pipeline makes
grid search over pipelines including slow transformations much more efficient. And you can predict faster: if you’re
sure you know what you’re doing, you can turn off validating that the input is finite using config_context.
We’ve made some important fixes too. We’ve fixed a longstanding implementation error in metrics.
average_precision_score, so please be cautious with prior results reported from that function. A number
of errors in the manifold.TSNE implementation have been fixed, particularly in the default Barnes-Hut approx-
imation. semi_supervised.LabelSpreading and semi_supervised.LabelPropagation have had
substantial fixes. LabelPropagation was previously broken. LabelSpreading should now correctly respect its alpha
parameter.
• cluster.KMeans with sparse X and initial centroids given (bug fix)
• cross_decomposition.PLSRegression with scale=True (bug fix)
• ensemble.GradientBoostingClassifier and ensemble.GradientBoostingRegressor
where min_impurity_split is used (bug fix)
• gradient boosting loss='quantile' (bug fix)
• ensemble.IsolationForest (bug fix)
• feature_selection.SelectFdr (bug fix)
• linear_model.RANSACRegressor (bug fix)
• linear_model.LassoLars (bug fix)
• linear_model.LassoLarsIC (bug fix)
1.14. Version 0.19 49

• manifold.TSNE (bug fix)

• neighbors.NearestCentroid (bug fix)
• semi_supervised.LabelSpreading (bug fix)
• semi_supervised.LabelPropagation (bug fix)
• tree based models where min_weight_fraction_leaf is used (enhancement)
• model_selection.StratifiedKFold with shuffle=True (this change, due to #7823 was not men-
tioned in the release notes at the time)
Details are listed in the changelog below.
(While we are trying to better inform users by providing this information, we cannot assure that this list is complete.)
1.14.3 Changelog
New features
Classifiers and regressors

• Added multioutput.ClassifierChain for multi-label classification. By Adam Kleczewski.
• Added solver 'saga' that implements the improved version of Stochastic Average Gradient, in
linear_model.LogisticRegression and linear_model.Ridge. It allows the use of L1 penalty
with multinomial logistic loss, and behaves marginally better than ‘sag’ during the first epochs of ridge and
logistic regression. #8446 by Arthur Mensch.
Other estimators
• Added the neighbors.LocalOutlierFactor class for anomaly detection based on nearest neighbors.
#5279 by Nicolas Goix and Alexandre Gramfort.
• Added preprocessing.QuantileTransformer class and preprocessing.
quantile_transform function for features normalization based on quantiles. #8363 by Denis Engemann,
Guillaume Lemaitre, Olivier Grisel, Raghav RV, Thierry Guillemot, and Gael Varoquaux.
• The new solver 'mu' implements a Multiplicate Update in decomposition.NMF, allowing the optimization
of all beta-divergences, including the Frobenius norm, the generalized Kullback-Leibler divergence and the
Itakura-Saito divergence. #5295 by Tom Dupre la Tour.
Model selection and evaluation
• model_selection.GridSearchCV and model_selection.RandomizedSearchCV now support
simultaneous evaluation of multiple metrics. Refer to the Specifying multiple metrics for evaluation section of
the user guide for more information. #7388 by Raghav RV
• Added the model_selection.cross_validate which allows evaluation of multiple metrics. This func-
tion returns a dict with more useful information from cross-validation such as the train scores, fit times and score
times. Refer to The cross_validate function and multiple metric evaluation section of the userguide for more
information. #7388 by Raghav RV
• Added metrics.mean_squared_log_error, which computes the mean square error of the logarithmic
transformation of targets, particularly useful for targets with an exponential trend. #7655 by Karan Desai.
• Added metrics.dcg_score and metrics.ndcg_score, which compute Discounted cumulative gain
(DCG) and Normalized discounted cumulative gain (NDCG). #7739 by David Gasquez.
• Added the model_selection.RepeatedKFold and model_selection.
RepeatedStratifiedKFold. #8120 by Neeraj Gangwar.

Miscellaneous
• Validation that input data contains no NaN or inf can now be suppressed using config_context, at your
own risk. This will save on runtime, and may be particularly useful for prediction time. #7548 by Joel Nothman.
• Added a test to ensure parameter listing in docstrings match the function/class signature. #9206 by Alexandre
Gramfort and Raghav RV.
Enhancements
Trees and ensembles

• The min_weight_fraction_leaf constraint in tree construction is now more efficient, taking a fast path
to declare a node a leaf if its weight is less than 2 * the minimum. Note that the constructed tree will be different
from previous versions where min_weight_fraction_leaf is used. #7441 by Nelson Liu.
• ensemble.GradientBoostingClassifier and ensemble.GradientBoostingRegressor
now support sparse input for prediction. #6101 by Ibraim Ganiev.
• ensemble.VotingClassifier now allows changing estimators by using ensemble.
VotingClassifier.set_params. An estimator can also be removed by setting it to None. #7674 by
Yichuan Liu.
• tree.export_graphviz now shows configurable number of decimal places. #8698 by Guillaume
Lemaitre.
• Added flatten_transform parameter to ensemble.VotingClassifier to change output shape of
transform method to 2 dimensional. #7794 by Ibraim Ganiev and Herilalaina Rakotoarison.
Linear, kernelized and related models
• linear_model.SGDClassifier, linear_model.SGDRegressor, linear_model.
linear_model.Perceptron now expose max_iter and tol parameters, to handle convergence more
precisely. n_iter parameter is deprecated, and the fitted estimator exposes a n_iter_ attribute, with actual
number of iterations before convergence. #5036 by Tom Dupre la Tour.
• Added average parameter to perform weight averaging in linear_model.
PassiveAggressiveClassifier. #4939 by Andrea Esuli.
• linear_model.RANSACRegressor no longer throws an error when calling fit if no inliers are found in
its first iteration. Furthermore, causes of skipped iterations are tracked in newly added attributes, n_skips_*.
#7914 by Michael Horrell.
• In gaussian_process.GaussianProcessRegressor, method predict is a lot faster with
return_std=True. #8591 by Hadrien Bertrand.
• Added return_std to predict method of linear_model.ARDRegression and linear_model.
BayesianRidge. #7838 by Sergey Feldman.
• Memory usage enhancements: Prevent cast from float32 to float64 in: linear_model.
MultiTaskElasticNet; linear_model.LogisticRegression when using newton-cg solver; and
linear_model.Ridge when using svd, sparse_cg, cholesky or lsqr solvers. #8835, #8061 by Joan Massich
and Nicolas Cordier and Thierry Guillemot.
Other predictors
• Custom metrics for the neighbors binary trees now have fewer constraints: they must take two 1d-arrays and
return a float. #6288 by Jake Vanderplas.
• algorithm='auto in neighbors estimators now chooses the most appropriate algorithm for all input
types and metrics. #9145 by Herilalaina Rakotoarison and Reddy Chinthala.
1.14. Version 0.19 51

Decomposition, manifold learning and clustering

• cluster.MiniBatchKMeans and cluster.KMeans now use significantly less memory when assigning
data points to their nearest cluster center. #7721 by Jon Crall.
• decomposition.PCA, decomposition.IncrementalPCA and decomposition.
TruncatedSVD now expose the singular values from the underlying SVD. They are stored in the
attribute singular_values_, like in decomposition.IncrementalPCA. #7685 by Tommy Löfstedt
• decomposition.NMF now faster when beta_loss=0. #9277 by @hongkahjun.
• Memory improvements for method barnes_hut in manifold.TSNE #7089 by Thomas Moreau and Olivier
Grisel.
• Optimization schedule improvements for Barnes-Hut manifold.TSNE so the results are closer to the one
from the reference implementation lvdmaaten/bhtsne by Thomas Moreau and Olivier Grisel.
• Memory usage enhancements: Prevent cast from float32 to float64 in decomposition.PCA and
decomposition.randomized_svd_low_rank. #9067 by Raghav RV.
Preprocessing and feature selection
• Added norm_order parameter to feature_selection.SelectFromModel to enable selection of the
norm order when coef_ is more than 1D. #6181 by Antoine Wendlinger.
• Added ability to use sparse matrices in feature_selection.f_regression with center=True.
#8065 by Daniel LeJeune.
• Small performance improvement to n-gram creation in feature_extraction.text by binding methods
for loops and special-casing unigrams. #7567 by Jaye Doepke
• Relax assumption on the data for the kernel_approximation.SkewedChi2Sampler. Since the
Skewed-Chi2 kernel is defined on the open interval (−𝑠𝑘𝑒𝑤𝑒𝑑𝑛𝑒𝑠𝑠; +∞)𝑑 , the transform function should not
check whether X < 0 but whether X < -self.skewedness. #7573 by Romain Brault.
• Made default kernel parameters kernel-dependent in kernel_approximation.Nystroem. #5229 by
Saurabh Bansod and Andreas Müller.
Model evaluation and meta-estimators
• pipeline.Pipeline is now able to cache transformers within a pipeline by using the memory constructor
parameter. #7990 by Guillaume Lemaitre.
• pipeline.Pipeline steps can now be accessed as attributes of its named_steps attribute. #8586 by
Herilalaina Rakotoarison.
• Added sample_weight parameter to pipeline.Pipeline.score. #7723 by Mikhail Korobov.
• Added ability to set n_jobs parameter to pipeline.make_union. A TypeError will be raised for any
other kwargs. #8028 by Alexander Booth.
• model_selection.GridSearchCV , model_selection.RandomizedSearchCV and
model_selection.cross_val_score now allow estimators with callable kernels which were
previously prohibited. #8005 by Andreas Müller .
• model_selection.cross_val_predict now returns output of the correct shape for all values of the
argument method. #7863 by Aman Dalmia.
• Added shuffle and random_state parameters to shuffle training data before taking prefixes of it based on
training sizes in model_selection.learning_curve. #7506 by Narine Kokhlikyan.
• model_selection.StratifiedShuffleSplit now works with multioutput multiclass (or multilabel)
data. #9044 by Vlad Niculae.

• Speed improvements to model_selection.StratifiedShuffleSplit. #5991 by Arthur Mensch and

Joel Nothman.
• Add shuffle parameter to model_selection.train_test_split. #8845 by themrmax
• multioutput.MultiOutputRegressor and multioutput.MultiOutputClassifier now
support online learning using partial_fit. :issue: 8053 by Peng Yu.
• Add max_train_size parameter to model_selection.TimeSeriesSplit #8282 by Aman Dalmia.
• More clustering metrics are now available through metrics.get_scorer and scoring parameters. #8117
by Raghav RV.
• A scorer based on metrics.explained_variance_score is also available. #9259 by Hanmin Qin.
Metrics
• metrics.matthews_corrcoef now support multiclass classification. #8094 by Jon Crall.
• Add sample_weight parameter to metrics.cohen_kappa_score. #8335 by Victor Poughon.
Miscellaneous
• utils.check_estimator now attempts to ensure that methods transform, predict, etc. do not set attributes
on the estimator. #7533 by Ekaterina Krivich.
• Added type checking to the accept_sparse parameter in utils.validation methods. This parameter
now accepts only boolean, string, or list/tuple of strings. accept_sparse=None is deprecated and should
be replaced by accept_sparse=False. #7880 by Josh Karnofsky.
• Make it possible to load a chunk of an svmlight formatted file by passing a range of bytes to datasets.
load_svmlight_file. #935 by Olivier Grisel.
• dummy.DummyClassifier and dummy.DummyRegressor now accept non-finite features. #8931 by
@Attractadore.
Bug fixes
Trees and ensembles

• Fixed a memory leak in trees when using trees with criterion='mae'. #8002 by Raghav RV.
• Fixed a bug where ensemble.IsolationForest uses an an incorrect formula for the average path length
#8549 by Peter Wang.
• Fixed a bug where ensemble.AdaBoostClassifier throws ZeroDivisionError while fitting data
with single class labels. #7501 by Dominik Krzeminski.
• Fixed a bug in ensemble.GradientBoostingClassifier and ensemble.
GradientBoostingRegressor where a float being compared to 0.0 using == caused a divide by
zero error. #7970 by He Chen.
• Fix a bug where ensemble.GradientBoostingClassifier and ensemble.
GradientBoostingRegressor ignored the min_impurity_split parameter. #8006 by Sebastian
Pölsterl.
• Fixed oob_score in ensemble.BaggingClassifier. #8936 by Michael Lewis
• Fixed excessive memory usage in prediction for random forests estimators. #8672 by Mike Benfield.
• Fixed a bug where sample_weight as a list broke random forests in Python 2 #8068 by @xor.
• Fixed a bug where ensemble.IsolationForest fails when max_features is less than 1. #5732 by
Ishank Gulati.
1.14. Version 0.19 53

• Fix a bug where gradient boosting with loss='quantile' computed negative errors for negative values of
ytrue - ypred leading to wrong values when calling __call__. #8087 by Alexis Mignon
• Fix a bug where ensemble.VotingClassifier raises an error when a numpy array is passed in for
weights. #7983 by Vincent Pham.
• Fixed a bug where tree.export_graphviz raised an error when the length of features_names does not
match n_features in the decision tree. #8512 by Li Li.
• Fixed a bug where linear_model.RANSACRegressor.fit may run until max_iter if it finds a large
inlier group early. #8251 by @aivision2020.
• Fixed a bug where naive_bayes.MultinomialNB and naive_bayes.BernoulliNB failed when
alpha=0. #5814 by Yichuan Liu and Herilalaina Rakotoarison.
• Fixed a bug where linear_model.LassoLars does not give the same result as the LassoLars implemen-
tation available in R (lars library). #7849 by Jair Montoya Martinez.
• Fixed a bug in linear_model.RandomizedLasso, linear_model.Lars, linear_model.
LassoLars, linear_model.LarsCV and linear_model.LassoLarsCV , where the parameter
precompute was not used consistently across classes, and some values proposed in the docstring could raise
errors. #5359 by Tom Dupre la Tour.
• Fix inconsistent results between linear_model.RidgeCV and linear_model.Ridge when using
normalize=True. #9302 by Alexandre Gramfort.
• Fix a bug where linear_model.LassoLars.fit sometimes left coef_ as a list, rather than an ndarray.
#8160 by CJ Carey.
• Fix linear_model.BayesianRidge.fit to return ridge parameter alpha_ and lambda_ consistent
with calculated coefficients coef_ and intercept_. #8224 by Peter Gedeck.
• Fixed a bug in svm.OneClassSVM where it returned floats instead of integer classes. #8676 by Vathsala
Achar.
• Fix AIC/BIC criterion computation in linear_model.LassoLarsIC. #9022 by Alexandre Gramfort and
Mehmet Basbug.
• Fixed a memory leak in our LibLinear implementation. #9024 by Sergei Lebedev
• Fix bug where stratified CV splitters did not work with linear_model.LassoCV . #8973 by Paulo Haddad.
• Fixed a bug in gaussian_process.GaussianProcessRegressor when the standard deviation and
covariance predicted without fit would fail with a unmeaningful error by default. #6573 by Quazi Marufur
Rahman and Manoj Kumar.
Other predictors
• Fix semi_supervised.BaseLabelPropagation to correctly implement LabelPropagation and
LabelSpreading as done in the referenced papers. #9239 by Andre Ambrosio Boechat, Utkarsh Upadhyay,
and Joel Nothman.
• Fixed the implementation of manifold.TSNE:
• early_exageration parameter had no effect and is now used for the first 250 optimization iterations.
• Fixed the AssertionError: Tree consistency failed exception reported in #8992.
• Improve the learning schedule to match the one from the reference implementation lvdmaaten/bhtsne. by
Thomas Moreau and Olivier Grisel.

• Fix a bug in decomposition.LatentDirichletAllocation where the perplexity method was

returning incorrect results because the transform method returns normalized document topic distributions as
of version 0.18. #7954 by Gary Foreman.
• Fix output shape and bugs with n_jobs > 1 in decomposition.SparseCoder transform and
decomposition.sparse_encode for one-dimensional data and one component. This also impacts the
output shape of decomposition.DictionaryLearning. #8086 by Andreas Müller.
• Fixed the implementation of explained_variance_ in decomposition.PCA, decomposition.
RandomizedPCA and decomposition.IncrementalPCA. #9105 by Hanmin Qin.
• Fixed the implementation of noise_variance_ in decomposition.PCA. #9108 by Hanmin Qin.
• Fixed a bug where cluster.DBSCAN gives incorrect result when input is a precomputed sparse matrix with
initial rows all zero. #8306 by Akshay Gupta
• Fix a bug regarding fitting cluster.KMeans with a sparse array X and initial centroids, where X’s means
were unnecessarily being subtracted from the centroids. #7872 by Josh Karnofsky.
• Fixes to the input validation in covariance.EllipticEnvelope. #8086 by Andreas Müller.
• Fixed a bug in covariance.MinCovDet where inputting data that produced a singular covariance matrix
would cause the helper method _c_step to throw an exception. #3367 by Jeremy Steward
• Fixed a bug in manifold.TSNE affecting convergence of the gradient descent. #8768 by David DeTomaso.
• Fixed a bug in manifold.TSNE where it stored the incorrect kl_divergence_. #6507 by Sebastian
Saeger.
• Fixed improper scaling in cross_decomposition.PLSRegression with scale=True. #7819 by
jayzed82.
• cluster.bicluster.SpectralCoclustering and cluster.bicluster.
SpectralBiclustering fit method conforms with API by accepting y and returning the object.
#6126, #7814 by Laurent Direr and Maniteja Nandana.
• Fix bug where mixture sample methods did not return as many samples as requested. #7702 by Levi John
Wolf.
• Fixed the shrinkage implementation in neighbors.NearestCentroid. #9219 by Hanmin Qin.
• For sparse matrices, preprocessing.normalize with return_norm=True will now raise a
NotImplementedError with ‘l1’ or ‘l2’ norm and with norm ‘max’ the norms returned will be the same as
for dense matrices. #7771 by Ang Lu.
• Fix a bug where feature_selection.SelectFdr did not exactly implement Benjamini-Hochberg pro-
cedure. It formerly may have selected fewer features than it should. #7490 by Peng Meng.
• Fixed a bug where linear_model.RandomizedLasso and linear_model.
RandomizedLogisticRegression breaks for sparse input. #8259 by Aman Dalmia.
• Fix a bug where feature_extraction.FeatureHasher mandatorily applied a sparse random projec-
tion to the hashed features, preventing the use of feature_extraction.text.HashingVectorizer
in a pipeline with feature_extraction.text.TfidfTransformer. #7565 by Roman Yurchak.
• Fix a bug where feature_selection.mutual_info_regression did not correctly use
n_neighbors. #8181 by Guillaume Lemaitre.
1.14. Version 0.19 55

• Fixed a bug where model_selection.BaseSearchCV.inverse_transform re-

turns self.best_estimator_.transform() instead of self.best_estimator_.
inverse_transform(). #8344 by Akshay Gupta and Rasmus Eriksson.
• Added classes_ attribute to model_selection.GridSearchCV , model_selection.
RandomizedSearchCV , grid_search.GridSearchCV, and grid_search.
RandomizedSearchCV that matches the classes_ attribute of best_estimator_. #7661 and
#8295 by Alyssa Batula, Dylan Werner-Meier, and Stephen Hoover.
• Fixed a bug where model_selection.validation_curve reused the same estimator for each parame-
ter value. #7365 by Aleksandr Sandrovskii.
• model_selection.permutation_test_score now works with Pandas types. #5697 by Stijn Tonk.
• Several fixes to input validation in multiclass.OutputCodeClassifier #8086 by Andreas Müller.
• multiclass.OneVsOneClassifier’s partial_fit now ensures all classes are provided up-front.
#6250 by Asish Panda.
• Fix multioutput.MultiOutputClassifier.predict_proba to return a list of 2d arrays, rather
than a 3d array. In the case where different target columns had different numbers of classes, a ValueError
would be raised on trying to stack matrices with different dimensions. #8093 by Peter Bull.
• Cross validation now works with Pandas datatypes that that have a read-only index. #9507 by Loic Esteve.
Metrics
• metrics.average_precision_score no longer linearly interpolates between operating points, and in-
stead weighs precisions by the change in recall since the last operating point, as per the Wikipedia entry. (#7356).
By Nick Dingwall and Gael Varoquaux.
• Fix a bug in metrics.classification._check_targets which would return 'binary' if y_true
and y_pred were both 'binary' but the union of y_true and y_pred was 'multiclass'. #8377 by
Loic Esteve.
• Fixed an integer overflow bug in metrics.confusion_matrix and hence metrics.
cohen_kappa_score. #8354, #7929 by Joel Nothman and Jon Crall.
• Fixed passing of gamma parameter to the chi2 kernel in metrics.pairwise.pairwise_kernels
#5211 by Nick Rhinehart, Saurabh Bansod and Andreas Müller.
Miscellaneous
• Fixed a bug when datasets.make_classification fails when generating more than 30 features. #8159
by Herilalaina Rakotoarison.
• Fixed a bug where datasets.make_moons gives an incorrect result when n_samples is odd. #8198 by
Josh Levy.
• Some fetch_ functions in datasets were ignoring the download_if_missing keyword. #7944 by
Ralf Gommers.
• Fix estimators to accept a sample_weight parameter of type pandas.Series in their fit function.
#7825 by Kathleen Chen.
• Fix a bug in cases where numpy.cumsum may be numerically unstable, raising an exception if instability is
identified. #7376 and #7331 by Joel Nothman and @yangarbiter.
• Fix a bug where base.BaseEstimator.__getstate__ obstructed pickling customizations of child-
classes, when used in a multiple inheritance context. #8316 by Holger Peters.
• Update Sphinx-Gallery from 0.1.4 to 0.1.7 for resolving links in documentation build with Sphinx>1.5 #8010,
#7986 by Oscar Najera

• Add data_home parameter to sklearn.datasets.fetch_kddcup99. #9289 by Loic Esteve.

• Fix dataset loaders using Python 3 version of makedirs to also work in Python 2. #9284 by Sebastin Santy.
• Several minor issues were fixed with thanks to the alerts of [lgtm.com](http://lgtm.com). #9278 by Jean Helie,
among others.
1.14.4 API changes summary
Trees and ensembles

• Gradient boosting base models are no longer estimators. By Andreas Müller.
• All tree based estimators now accept a min_impurity_decrease parameter in lieu of the
min_impurity_split, which is now deprecated. The min_impurity_decrease helps stop
splitting the nodes in which the weighted impurity decrease from splitting is no longer at least
min_impurity_decrease. #8449 by Raghav RV.
• n_iter parameter is deprecated in linear_model.SGDClassifier, linear_model.
SGDRegressor, linear_model.PassiveAggressiveClassifier, linear_model.
PassiveAggressiveRegressor and linear_model.Perceptron. By Tom Dupre la Tour.
Other predictors
• neighbors.LSHForest has been deprecated and will be removed in 0.21 due to poor performance. #9078
by Laurent Direr.
• neighbors.NearestCentroid no longer purports to support metric='precomputed' which now
raises an error. #8515 by Sergul Aydore.
• The alpha parameter of semi_supervised.LabelPropagation now has no effect and is deprecated
to be removed in 0.21. #9239 by Andre Ambrosio Boechat, Utkarsh Upadhyay, and Joel Nothman.
• Deprecate the doc_topic_distr argument of the perplexity method in decomposition.
LatentDirichletAllocation because the user no longer has access to the unnormalized document topic
distribution needed for the perplexity calculation. #7954 by Gary Foreman.
• The n_topics parameter of decomposition.LatentDirichletAllocation has been renamed to
n_components and will be removed in version 0.21. #8922 by @Attractadore.
• decomposition.SparsePCA.transform’s ridge_alpha parameter is deprecated in preference for
class parameter. #8137 by Naoya Kanai.
• cluster.DBSCAN now has a metric_params parameter. #8139 by Naoya Kanai.
• feature_selection.SelectFromModel now has a partial_fit method only if the underlying es-
timator does. By Andreas Müller.
• feature_selection.SelectFromModel now validates the threshold parameter and sets the
threshold_ attribute during the call to fit, and no longer during the call to transform`. By Andreas
Müller.
• The non_negative parameter in feature_extraction.FeatureHasher has been deprecated, and
replaced with a more principled alternative, alternate_sign. #7565 by Roman Yurchak.
• linear_model.RandomizedLogisticRegression, and linear_model.RandomizedLasso
have been deprecated and will be removed in version 0.21. #8995 by Ramana.S.
1.14. Version 0.19 57


• Deprecate the fit_params constructor input to the model_selection.GridSearchCV and
model_selection.RandomizedSearchCV in favor of passing keyword parameters to the fit methods
of those classes. Data-dependent parameters needed for model training should be passed as keyword arguments
to fit, and conforming to this convention will allow the hyperparameter selection classes to be used with tools
such as model_selection.cross_val_predict. #2879 by Stephen Hoover.
• In version 0.21, the default behavior of splitters that use the test_size and train_size parameter will
change, such that specifying train_size alone will cause test_size to be the remainder. #7459 by Nelson
Liu.
• multiclass.OneVsRestClassifier now has partial_fit, decision_function and
predict_proba methods only when the underlying estimator does. #7812 by Andreas Müller and Mikhail
Korobov.
• multiclass.OneVsRestClassifier now has a partial_fit method only if the underlying estima-
tor does. By Andreas Müller.
• The decision_function output shape for binary classification in multiclass.
OneVsRestClassifier and multiclass.OneVsOneClassifier is now (n_samples,) to
conform to scikit-learn conventions. #9100 by Andreas Müller.
• The multioutput.MultiOutputClassifier.predict_proba function used to return a 3d array
(n_samples, n_classes, n_outputs). In the case where different target columns had different numbers
of classes, a ValueError would be raised on trying to stack matrices with different dimensions. This func-
tion now returns a list of arrays where the length of the list is n_outputs, and each array is (n_samples,
n_classes) for that particular output. #8093 by Peter Bull.
• Replace attribute named_steps dict to utils.Bunch in pipeline.Pipeline to enable tab com-
pletion in interactive environment. In the case conflict value on named_steps and dict attribute, dict
behavior will be prioritized. #8481 by Herilalaina Rakotoarison.
Miscellaneous
• Deprecate the y parameter in transform and inverse_transform. The method should not accept y
parameter, as it’s used at the prediction time. #8174 by Tahar Zanouda, Alexandre Gramfort and Raghav RV.
• SciPy >= 0.13.3 and NumPy >= 1.8.2 are now the minimum supported versions for scikit-learn. The following
backported functions in utils have been removed or deprecated accordingly. #8854 and #8874 by Naoya
Kanai
• The store_covariances and covariances_ parameters of discriminant_analysis.
QuadraticDiscriminantAnalysis has been renamed to store_covariance and covariance_
to be consistent with the corresponding parameter names of the discriminant_analysis.
LinearDiscriminantAnalysis. They will be removed in version 0.21. #7998 by Jiacheng
Removed in 0.19:
– utils.fixes.argpartition
– utils.fixes.array_equal
– utils.fixes.astype
– utils.fixes.bincount
– utils.fixes.expit
– utils.fixes.frombuffer_empty
– utils.fixes.in1d
– utils.fixes.norm

– utils.fixes.rankdata
– utils.fixes.safe_copy
Deprecated in 0.19, to be removed in 0.21:
– utils.arpack.eigs
– utils.arpack.eigsh
– utils.arpack.svds
– utils.extmath.fast_dot
– utils.extmath.logsumexp
– utils.extmath.norm
– utils.extmath.pinvh
– utils.graph.graph_laplacian
– utils.random.choice
– utils.sparsetools.connected_components
– utils.stats.rankdata
• Estimators with both methods decision_function and predict_proba are now required to have a
monotonic relation between them. The method check_decision_proba_consistency has been added
in utils.estimator_checks to check their consistency. #7578 by Shubham Bhardwaj
• All checks in utils.estimator_checks, in particular utils.estimator_checks.
check_estimator now accept estimator instances. Most other checks do not accept estimator classes any
more. #9019 by Andreas Müller.
• Ensure that estimators’ attributes ending with _ are not set in the constructor but only in the fit method.
Most notably, ensemble estimators (deriving from ensemble.BaseEnsemble) now only have self.
estimators_ available after fit. #7464 by Lars Buitinck and Loic Esteve.
Thanks to everyone who has contributed to the maintenance and improvement of the project since version 0.18, in-
cluding:
Joel Nothman, Loic Esteve, Andreas Mueller, Guillaume Lemaitre, Olivier Grisel, Hanmin Qin, Raghav RV, Alexandre
Gramfort, themrmax, Aman Dalmia, Gael Varoquaux, Naoya Kanai, Tom Dupré la Tour, Rishikesh, Nelson Liu, Tae-
hoon Lee, Nelle Varoquaux, Aashil, Mikhail Korobov, Sebastin Santy, Joan Massich, Roman Yurchak, RAKOTOARI-
SON Herilalaina, Thierry Guillemot, Alexandre Abadie, Carol Willing, Balakumaran Manoharan, Josh Karnofsky,
Vlad Niculae, Utkarsh Upadhyay, Dmitry Petrov, Minghui Liu, Srivatsan, Vincent Pham, Albert Thomas, Jake Van-
derPlas, Attractadore, JC Liu, alexandercbooth, chkoar, Óscar Nájera, Aarshay Jain, Kyle Gilliam, Ramana Subra-
manyam, CJ Carey, Clement Joudet, David Robles, He Chen, Joris Van den Bossche, Karan Desai, Katie Luangkote,
Leland McInnes, Maniteja Nandana, Michele Lacchia, Sergei Lebedev, Shubham Bhardwaj, akshay0724, omtcyfz,
rickiepark, waterponey, Vathsala Achar, jbDelafosse, Ralf Gommers, Ekaterina Krivich, Vivek Kumar, Ishank Gulati,
Dave Elliott, ldirer, Reiichiro Nakano, Levi John Wolf, Mathieu Blondel, Sid Kapur, Dougal J. Sutherland, midinas,
mikebenfield, Sourav Singh, Aseem Bansal, Ibraim Ganiev, Stephen Hoover, AishwaryaRK, Steven C. Howell, Gary
Foreman, Neeraj Gangwar, Tahar, Jon Crall, dokato, Kathy Chen, ferria, Thomas Moreau, Charlie Brummitt, Nicolas
Goix, Adam Kleczewski, Sam Shleifer, Nikita Singh, Basil Beirouti, Giorgio Patrini, Manoj Kumar, Rafael Possas,
James Bourbeau, James A. Bednar, Janine Harper, Jaye, Jean Helie, Jeremy Steward, Artsiom, John Wei, Jonathan
LIgo, Jonathan Rahn, seanpwilliams, Arthur Mensch, Josh Levy, Julian Kuhlmann, Julien Aubert, Jörn Hees, Kai,
shivamgargsya, Kat Hempstalk, Kaushik Lakshmikanth, Kennedy, Kenneth Lyons, Kenneth Myers, Kevin Yap, Kir-
ill Bobyrev, Konstantin Podshumok, Arthur Imbert, Lee Murray, toastedcornflakes, Lera, Li Li, Arthur Douillard,
1.14. Version 0.19 59

Mainak Jas, tobycheese, Manraj Singh, Manvendra Singh, Marc Meketon, MarcoFalke, Matthew Brett, Matthias
Gilch, Mehul Ahuja, Melanie Goetz, Meng, Peng, Michael Dezube, Michal Baumgartner, vibrantabhi19, Artem Golu-
bin, Milen Paskov, Antonin Carette, Morikko, MrMjauh, NALEPA Emmanuel, Namiya, Antoine Wendlinger, Narine
Kokhlikyan, NarineK, Nate Guerin, Angus Williams, Ang Lu, Nicole Vavrova, Nitish Pandey, Okhlopkov Daniil
Olegovich, Andy Craze, Om Prakash, Parminder Singh, Patrick Carlson, Patrick Pei, Paul Ganssle, Paulo Haddad,
Paweł Lorek, Peng Yu, Pete Bachant, Peter Bull, Peter Csizsek, Peter Wang, Pieter Arthur de Jong, Ping-Yao, Chang,
Preston Parry, Puneet Mathur, Quentin Hibon, Andrew Smith, Andrew Jackson, 1kastner, Rameshwar Bhaskaran, Re-
becca Bilbro, Remi Rampin, Andrea Esuli, Rob Hall, Robert Bradshaw, Romain Brault, Aman Pratik, Ruifeng Zheng,
Russell Smith, Sachin Agarwal, Sailesh Choyal, Samson Tan, Samuël Weber, Sarah Brown, Sebastian Pölsterl, Se-
bastian Raschka, Sebastian Saeger, Alyssa Batula, Abhyuday Pratap Singh, Sergey Feldman, Sergul Aydore, Sharan
Yalburgi, willduan, Siddharth Gupta, Sri Krishna, Almer, Stijn Tonk, Allen Riddell, Theofilos Papapanagiotou, Alison,
Alexis Mignon, Tommy Boucher, Tommy Löfstedt, Toshihiro Kamishima, Tyler Folkman, Tyler Lanigan, Alexander
Junge, Varun Shenoy, Victor Poughon, Vilhelm von Ehrenheim, Aleksandr Sandrovskii, Alan Yee, Vlasios Vasileiou,
Warut Vijitbenjaronk, Yang Zhang, Yaroslav Halchenko, Yichuan Liu, Yuichi Fujikawa, affanv14, aivision2020, xor,
andreh7, brady salz, campustrampus, Agamemnon Krasoulis, ditenberg, elena-sharova, filipj8, fukatani, gedeck, guin-
iol, guoci, hakaa1, hongkahjun, i-am-xhy, jakirkham, jaroslaw-weber, jayzed82, jeroko, jmontoyam, jonathan.striebel,
josephsalmon, jschendel, leereeves, martin-hahn, mathurinm, mehak-sachdeva, mlewis1729, mlliou112, mthorrell,
ndingwall, nuffe, yangarbiter, plagree, pldtc325, Breno Freitas, Brett Olsen, Brian A. Alfano, Brian Burns, polmauri,
Brandon Carter, Charlton Austin, Chayant T15h, Chinmaya Pancholi, Christian Danielsen, Chung Yen, Chyi-Kwei
Yau, pravarmahajan, DOHMATOB Elvis, Daniel LeJeune, Daniel Hnyk, Darius Morawiec, David DeTomaso, David
Gasquez, David Haberthür, David Heryanto, David Kirkby, David Nicholson, rashchedrin, Deborah Gertrude Digges,
Denis Engemann, Devansh D, Dickson, Bob Baxley, Don86, E. Lynch-Klarup, Ed Rogers, Elizabeth Ferriss, Ellen-
Co2, Fabian Egli, Fang-Chieh Chou, Bing Tian Dai, Greg Stupp, Grzegorz Szpak, Bertrand Thirion, Hadrien Bertrand,
Harizo Rajaona, zxcvbnius, Henry Lin, Holger Peters, Icyblade Dai, Igor Andriushchenko, Ilya, Isaac Laughlin, Iván
Vallés, Aurélien Bellet, JPFrancoia, Jacob Schreiber, Asish Mahapatra
1.15 Previous Releases
1.15.1 Version 0.18.2
June 20, 2017
Last release with Python 2.6 support
Scikit-learn 0.18 is the last major release of scikit-learn to support Python 2.6. Later versions of scikit-learn will
require Python 2.7 or above.
Changelog
• Fixes for compatibility with NumPy 1.13.0: #7946 #8355 by Loic Esteve.
• Minor compatibility changes in the examples #9010 #8040 #9149.
Code Contributors
Aman Dalmia, Loic Esteve, Nate Guerin, Sergei Lebedev

1.15.2 Version 0.18.1
November 11, 2016
Changelog
Enhancements
• Improved sample_without_replacement speed by utilizing numpy.random.permutation for most cases.

As a result, samples may differ in this release for a fixed random state. Affected estimators:
– ensemble.BaggingClassifier
– ensemble.BaggingRegressor
– linear_model.RANSACRegressor
– model_selection.RandomizedSearchCV
– random_projection.SparseRandomProjection
This also affects the datasets.make_classification method.
Bug fixes
• Fix issue where min_grad_norm and n_iter_without_progress parameters were not being utilised
by manifold.TSNE. #6497 by Sebastian Säger
• Fix bug for svm’s decision values when decision_function_shape is ovr in svm.SVC. svm.SVC’s
decision_function was incorrect from versions 0.17.0 through 0.18.0. #7724 by Bing Tian Dai
• Attribute explained_variance_ratio of discriminant_analysis.
LinearDiscriminantAnalysis calculated with SVD and Eigen solver are now of the same length.
#7632 by JPFrancoia
• Fixes issue in Univariate feature selection where score functions were not accepting multi-label targets. #7676
by Mohammed Affan
• Fixed setting parameters when calling fit multiple times on feature_selection.SelectFromModel.
#7756 by Andreas Müller
• Fixes issue in partial_fit method of multiclass.OneVsRestClassifier when number of classes
used in partial_fit was less than the total number of classes in the data. #7786 by Srivatsan Ramesh
• Fixes issue in calibration.CalibratedClassifierCV where the sum of probabilities of each class
for a data was not 1, and CalibratedClassifierCV now handles the case where the training set has less
number of classes than the total data. #7799 by Srivatsan Ramesh
• Fix a bug where sklearn.feature_selection.SelectFdr did not exactly implement Benjamini-
Hochberg procedure. It formerly may have selected fewer features than it should. #7490 by Peng Meng.
• sklearn.manifold.LocallyLinearEmbedding now correctly handles integer inputs. #6282 by Jake
Vanderplas.
• The min_weight_fraction_leaf parameter of tree-based classifiers and regressors now assumes uniform
sample weights by default if the sample_weight argument is not passed to the fit function. Previously, the
parameter was silently ignored. #7301 by Nelson Liu.
• Numerical issue with linear_model.RidgeCV on centered data when n_features > n_samples. #6178 by
Bertrand Thirion
1.15. Previous Releases 61

• Tree splitting criterion classes’ cloning/pickling is now memory safe #7680 by Ibraim Ganiev.
• Fixed a bug where decomposition.NMF sets its n_iters_ attribute in transform(). #7553 by Ekaterina
Krivich.
• sklearn.linear_model.LogisticRegressionCV now correctly handles string labels. #5874 by
Raghav RV.
• Fixed a bug where sklearn.model_selection.train_test_split raised an error when
stratify is a list of string labels. #7593 by Raghav RV.
• Fixed a bug where sklearn.model_selection.GridSearchCV and sklearn.
model_selection.RandomizedSearchCV were not pickleable because of a pickling bug in np.
ma.MaskedArray. #7594 by Raghav RV.
• All cross-validation utilities in sklearn.model_selection now permit one time cross-validation splitters
for the cv parameter. Also non-deterministic cross-validation splitters (where multiple calls to split produce
dissimilar splits) can be used as cv parameter. The sklearn.model_selection.GridSearchCV will
cross-validate each parameter setting on the split produced by the first split call to the cross-validation splitter.
#7660 by Raghav RV.
• Fix bug where preprocessing.MultiLabelBinarizer.fit_transform returned an invalid CSR
matrix. #7750 by CJ Carey.
• Fixed a bug where metrics.pairwise.cosine_distances could return a small negative distance.
#7732 by Artsion.
API changes summary
Trees and forests

• The min_weight_fraction_leaf parameter of tree-based classifiers and regressors now assumes uniform
sample weights by default if the sample_weight argument is not passed to the fit function. Previously, the
parameter was silently ignored. #7301 by Nelson Liu.
• Tree splitting criterion classes’ cloning/pickling is now memory safe. #7680 by Ibraim Ganiev.
• Length of explained_variance_ratio of discriminant_analysis.
LinearDiscriminantAnalysis changed for both Eigen and SVD solvers. The attribute has now
a length of min(n_components, n_classes - 1). #7632 by JPFrancoia
• Numerical issue with linear_model.RidgeCV on centered data when n_features > n_samples.
#6178 by Bertrand Thirion
1.15.3 Version 0.18
September 28, 2016
Last release with Python 2.6 support
Scikit-learn 0.18 will be the last version of scikit-learn to support Python 2.6. Later versions of scikit-learn will
require Python 2.7 or above.

Model Selection Enhancements and API Changes
• The model_selection module

The new module sklearn.model_selection, which groups together the functionalities of formerly
sklearn.cross_validation, sklearn.grid_search and sklearn.learning_curve, intro-
duces new possibilities such as nested cross-validation and better manipulation of parameter searches with Pan-
das.
Many things will stay the same but there are some key differences. Read below to know more about the changes.
• Data-independent CV splitters enabling nested cross-validation
The new cross-validation splitters, defined in the sklearn.model_selection, are no longer initialized
with any data-dependent parameters such as y. Instead they expose a split method that takes in the data and
yields a generator for the different splits.
This change makes it possible to use the cross-validation splitters to perform nested cross-validation, facilitated
by model_selection.GridSearchCV and model_selection.RandomizedSearchCV utilities.
• The enhanced cv_results_ attribute
The new cv_results_ attribute (of model_selection.GridSearchCV and model_selection.
RandomizedSearchCV ) introduced in lieu of the grid_scores_ attribute is a dict of 1D arrays with
elements in each array corresponding to the parameter settings (i.e. search candidates).
The cv_results_ dict can be easily imported into pandas as a DataFrame for exploring the search results.
The cv_results_ arrays include scores for each cross-validation split (with keys such as
'split0_test_score'), as well as their mean ('mean_test_score') and standard deviation
('std_test_score').
The ranks for the search candidates (based on their mean cross-validation score) is available at
cv_results_['rank_test_score'].
The parameter values for each parameter is stored separately as numpy masked object arrays. The value, for
that search candidate, is masked if the corresponding parameter is not applicable. Additionally a list of all the
parameter dicts are stored at cv_results_['params'].
• Parameters n_folds and n_iter renamed to n_splits
Some parameter names have changed: The n_folds parameter in new model_selection.KFold,
model_selection.GroupKFold (see below for the name change), and model_selection.
StratifiedKFold is now renamed to n_splits. The n_iter parameter in model_selection.
ShuffleSplit, the new class model_selection.GroupShuffleSplit and model_selection.
StratifiedShuffleSplit is now renamed to n_splits.
• Rename of splitter classes which accepts group labels along with data
The cross-validation splitters LabelKFold, LabelShuffleSplit, LeaveOneLabelOut and
LeavePLabelOut have been renamed to model_selection.GroupKFold, model_selection.
GroupShuffleSplit, model_selection.LeaveOneGroupOut and model_selection.
LeavePGroupsOut respectively.
Note the change from singular to plural form in model_selection.LeavePGroupsOut.
• Fit parameter labels renamed to groups
The labels parameter in the split method of the newly renamed splitters model_selection.
GroupKFold, model_selection.LeaveOneGroupOut, model_selection.
LeavePGroupsOut, model_selection.GroupShuffleSplit is renamed to groups following the
new nomenclature of their class names.

• Parameter n_labels renamed to n_groups

The parameter n_labels in the newly renamed model_selection.LeavePGroupsOut is changed to
n_groups.
• Training scores and Timing information
cv_results_ also includes the training scores for each cross-validation split (with keys such
as 'split0_train_score'), as well as their mean ('mean_train_score') and stan-
dard deviation ('std_train_score'). To avoid the cost of evaluating training score, set
return_train_score=False.
Additionally the mean and standard deviation of the times taken to split, train and score the model across all the
cross-validation splits is available at the key 'mean_time' and 'std_time' respectively.
Changelog
New features
Classifiers and Regressors

• The Gaussian Process module has been reimplemented and now offers classification and regression esti-
mators through gaussian_process.GaussianProcessClassifier and gaussian_process.
GaussianProcessRegressor. Among other things, the new implementation supports kernel engineering,
gradient-based hyperparameter optimization or sampling of functions from GP prior and GP posterior. Extensive
documentation and examples are provided. By Jan Hendrik Metzen.
• Added new supervised learning algorithm: Multi-layer Perceptron #3204 by Issam H. Laradji
• Added linear_model.HuberRegressor, a linear model robust to outliers. #5291 by Manoj Kumar.
• Added the multioutput.MultiOutputRegressor meta-estimator. It converts single output regressors
to multi-output regressors by fitting one regressor per output. By Tim Head.
Other estimators
• New mixture.GaussianMixture and mixture.BayesianGaussianMixture replace former mix-
ture models, employing faster inference for sounder results. #7295 by Wei Xue and Thierry Guillemot.
• Class decomposition.RandomizedPCA is now factored into decomposition.PCA and it is avail-
able calling with parameter svd_solver='randomized'. The default number of n_iter for
'randomized' has changed to 4. The old behavior of PCA is recovered by svd_solver='full'. An
additional solver calls arpack and performs truncated (non-randomized) SVD. By default, the best solver is
selected depending on the size of the input and the number of components requested. #5299 by Giorgio Patrini.
• Added two functions for mutual information estimation: feature_selection.
mutual_info_classif and feature_selection.mutual_info_regression. These
functions can be used in feature_selection.SelectKBest and feature_selection.
SelectPercentile as score functions. By Andrea Bravi and Nikolay Mayorov.
• Added the ensemble.IsolationForest class for anomaly detection based on random forests. By Nicolas
Goix.
• Added algorithm="elkan" to cluster.KMeans implementing Elkan’s fast K-Means algorithm. By
Andreas Müller.
Model selection and evaluation
• Added metrics.cluster.fowlkes_mallows_score, the Fowlkes Mallows Index which measures the
similarity of two clusterings of a set of points By Arnaud Fouchet and Thierry Guillemot.

• Added metrics.calinski_harabaz_score, which computes the Calinski and Harabaz score to evalu-
ate the resulting clustering of a set of points. By Arnaud Fouchet and Thierry Guillemot.
• Added new cross-validation splitter model_selection.TimeSeriesSplit to handle time series data.
#6586 by YenChen Lin
• The cross-validation iterators are replaced by cross-validation splitters available from sklearn.
model_selection, allowing for nested cross-validation. See Model Selection Enhancements and API
Changes for more information. #4294 by Raghav RV.
Enhancements
Trees and ensembles

• Added a new splitting criterion for tree.DecisionTreeRegressor, the mean absolute er-
ror. This criterion can also be used in ensemble.ExtraTreesRegressor, ensemble.
RandomForestRegressor, and the gradient boosting estimators. #6667 by Nelson Liu.
• Added weighted impurity-based early stopping criterion for decision tree growth. #6954 by Nelson Liu
• The random forest, extra tree and decision tree estimators now has a method decision_path which returns
the decision path of samples in the tree. By Arnaud Joly.
• A new example has been added unveiling the decision tree structure. By Arnaud Joly.
• Random forest, extra trees, decision trees and gradient boosting estimator accept the parameter
min_samples_split and min_samples_leaf provided as a percentage of the training samples. By
yelite and Arnaud Joly.
• Gradient boosting estimators accept the parameter criterion to specify to splitting criterion used in built
decision trees. #6667 by Nelson Liu.
• The memory footprint is reduced (sometimes greatly) for ensemble.bagging.BaseBagging and classes
that inherit from it, i.e, ensemble.BaggingClassifier, ensemble.BaggingRegressor, and
ensemble.IsolationForest, by dynamically generating attribute estimators_samples_ only
when it is needed. By David Staub.
• Added n_jobs and sample_weight parameters for ensemble.VotingClassifier to fit underlying
estimators in parallel. #5805 by Ibraim Ganiev.
• In linear_model.LogisticRegression, the SAG solver is now available in the multinomial case.
#5251 by Tom Dupre la Tour.
• linear_model.RANSACRegressor, svm.LinearSVC and svm.LinearSVR now support
sample_weight. By Imaculate.
• Add parameter loss to linear_model.RANSACRegressor to measure the error on the samples for every
trial. By Manoj Kumar.
• Prediction of out-of-sample events with Isotonic Regression (isotonic.IsotonicRegression) is now
much faster (over 1000x in tests with synthetic data). By Jonathan Arfa.
• Isotonic regression (isotonic.IsotonicRegression) now uses a better algorithm to avoid O(n^2) be-
havior in pathological cases, and is also generally faster (##6691). By Antony Lee.
• naive_bayes.GaussianNB now accepts data-independent class-priors through the parameter priors.
By Guillaume Lemaitre.

• linear_model.ElasticNet and linear_model.Lasso now works with np.float32 input data

without converting it into np.float64. This allows to reduce the memory consumption. #6913 by YenChen
Lin.
• semi_supervised.LabelPropagation and semi_supervised.LabelSpreading now accept
arbitrary kernel functions in addition to strings knn and rbf. #5762 by Utkarsh Upadhyay.
• Added inverse_transform function to decomposition.NMF to compute data matrix of original shape.
By Anish Shah.
• cluster.KMeans and cluster.MiniBatchKMeans now works with np.float32 and np.
float64 input data without converting it. This allows to reduce the memory consumption by using np.
float32. #6846 by Sebastian Säger and YenChen Lin.
• preprocessing.RobustScaler now accepts quantile_range parameter. #5929 by Konstantin Pod-
shumok.
• feature_extraction.FeatureHasher now accepts string values. #6173 by Ryad Zenine and
Devashish Deshpande.
• Keyword arguments can now be supplied to func in preprocessing.FunctionTransformer by
means of the kw_args parameter. By Brian McFee.
• feature_selection.SelectKBest and feature_selection.SelectPercentile now accept
score functions that take X, y as input and return only the scores. By Nikolay Mayorov.
• multiclass.OneVsOneClassifier and multiclass.OneVsRestClassifier now support
partial_fit. By Asish Panda and Philipp Dowling.
• Added support for substituting or disabling pipeline.Pipeline and pipeline.FeatureUnion com-
ponents using the set_params interface that powers sklearn.grid_search. See Selecting dimension-
ality reduction with Pipeline and GridSearchCV By Joel Nothman and Robert McGibbon.
• The new cv_results_ attribute of model_selection.GridSearchCV (and model_selection.
RandomizedSearchCV ) can be easily imported into pandas as a DataFrame. Ref Model Selection En-
hancements and API Changes for more information. #6697 by Raghav RV.
• Generalization of model_selection.cross_val_predict. One can pass method names such as pre-
dict_proba to be used in the cross validation framework instead of the default predict. By Ori Ziv and Sears
Merritt.
• The training scores and time taken for training followed by scoring for each search candidate are now available
at the cv_results_ dict. See Model Selection Enhancements and API Changes for more information. #7325
by Eugene Chen and Raghav RV.
Metrics
• Added labels flag to metrics.log_loss to explicitly provide the labels when the number of classes in
y_true and y_pred differ. #7239 by Hong Guangguo with help from Mads Jensen and Nelson Liu.
• Support sparse contingency matrices in cluster evaluation (metrics.cluster.supervised) to scale to a
large number of clusters. #7419 by Gregory Stupp and Joel Nothman.
• Add sample_weight parameter to metrics.matthews_corrcoef. By Jatin Shah and Raghav RV.
• Speed up metrics.silhouette_score by using vectorized operations. By Manoj Kumar.
• Add sample_weight parameter to metrics.confusion_matrix. By Bernardo Stein.

Miscellaneous
• Added n_jobs parameter to feature_selection.RFECV to compute the score on the test folds in par-
allel. By Manoj Kumar
• Codebase does not contain C/C++ cython generated files: they are generated during build. Distribution packages
will still contain generated C/C++ files. By Arthur Mensch.
• Reduce the memory usage for 32-bit float input arrays of utils.sparse_func.mean_variance_axis
and utils.sparse_func.incr_mean_variance_axis by supporting cython fused types. By
YenChen Lin.
• The ignore_warnings now accept a category argument to ignore only the warnings of a specified type. By
Thierry Guillemot.
• Added parameter return_X_y and return type (data, target) : tuple option to load_iris
dataset #7049, load_breast_cancer dataset #7152, load_digits dataset, load_diabetes dataset,
load_linnerud dataset, load_boston dataset #7154 by Manvendra Singh.
• Simplification of the clone function, deprecate support for estimators that modify parameters in __init__.
#5540 by Andreas Müller.
• When unpickling a scikit-learn estimator in a different version than the one the estimator was trained with, a
UserWarning is raised, see the documentation on model persistence for more details. (#7248) By Andreas
Müller.
Bug fixes
Trees and ensembles

• Random forest, extra trees, decision trees and gradient boosting won’t accept anymore
min_samples_split=1 as at least 2 samples are required to split a decision tree node. By Arnaud
Joly
• ensemble.VotingClassifier now raises NotFittedError if predict, transform or
predict_proba are called on the non-fitted estimator. by Sebastian Raschka.
• Fix bug where ensemble.AdaBoostClassifier and ensemble.AdaBoostRegressor would per-
form poorly if the random_state was fixed (#7411). By Joel Nothman.
• Fix bug in ensembles with randomization where the ensemble would not set random_state
on base estimators in a pipeline or similar nesting. (#7411). Note, results for ensemble.
BaggingClassifier ensemble.BaggingRegressor, ensemble.AdaBoostClassifier and
ensemble.AdaBoostRegressor will now differ from previous versions. By Joel Nothman.
• Fixed incorrect gradient computation for loss='squared_epsilon_insensitive' in
linear_model.SGDClassifier and linear_model.SGDRegressor (#6764). By Wenhua
Yang.
• Fix bug in linear_model.LogisticRegressionCV where solver='liblinear' did not accept
class_weights='balanced. (#6817). By Tom Dupre la Tour.
• Fix bug in neighbors.RadiusNeighborsClassifier where an error occurred when there were out-
liers being labelled and a weight function specified (#6902). By LeonieBorne.
• Fix linear_model.ElasticNet sparse decision function to match output with dense in the multioutput
case.

• decomposition.RandomizedPCA default number of iterated_power is 4 instead of 3. #5141 by Giorgio

Patrini.
• utils.extmath.randomized_svd performs 4 power iterations by default, instead or 0. In practice this
is enough for obtaining a good approximation of the true eigenvalues/vectors in the presence of noise. When
n_components is small (< .1 * min(X.shape)) n_iter is set to 7, unless the user specifies a higher number.
This improves precision with few components. #5299 by Giorgio Patrini.
• Whiten/non-whiten inconsistency between components of decomposition.PCA and decomposition.
RandomizedPCA (now factored into PCA, see the New features) is fixed. components_ are stored with no
whitening. #5299 by Giorgio Patrini.
• Fixed bug in manifold.spectral_embedding where diagonal of unnormalized Laplacian matrix was
incorrectly set to 1. #4995 by Peter Fischer.
• Fixed incorrect initialization of utils.arpack.eigsh on all occurrences. Affects cluster.
bicluster.SpectralBiclustering, decomposition.KernelPCA, manifold.
LocallyLinearEmbedding, and manifold.SpectralEmbedding (#5012). By Peter Fischer.
• Attribute explained_variance_ratio_ calculated with the SVD solver of
discriminant_analysis.LinearDiscriminantAnalysis now returns correct results. By
JPFrancoia
• preprocessing.data._transform_selected now always passes a copy of X to transform function
when copy=True (#7194). By Caio Oliveira.
• model_selection.StratifiedKFold now raises error if all n_labels for individual classes is less than
n_folds. #6182 by Devashish Deshpande.
• Fixed bug in model_selection.StratifiedShuffleSplit where train and test sample could overlap
in some edge cases, see #6121 for more details. By Loic Esteve.
• Fix in sklearn.model_selection.StratifiedShuffleSplit to return splits of size
train_size and test_size in all cases (#6472). By Andreas Müller.
• Cross-validation of OneVsOneClassifier and OneVsRestClassifier now works with precomputed
kernels. #7350 by Russell Smith.
• Fix incomplete predict_proba method delegation from model_selection.GridSearchCV to
linear_model.SGDClassifier (#7159) by Yichuan Liu.
Metrics
• Fix bug in metrics.silhouette_score in which clusters of size 1 were incorrectly scored. They should
get a score of 0. By Joel Nothman.
• Fix bug in metrics.silhouette_samples so that it now works with arbitrary labels, not just those
ranging from 0 to n_clusters - 1.
• Fix bug where expected and adjusted mutual information were incorrect if cluster contingency cells exceeded
2**16. By Joel Nothman.
• metrics.pairwise.pairwise_distances now converts arrays to boolean arrays when required in
scipy.spatial.distance. #5460 by Tom Dupre la Tour.
• Fix sparse input support in metrics.silhouette_score as well as example exam-
ples/text/document_clustering.py. By YenChen Lin.
• metrics.roc_curve and metrics.precision_recall_curve no longer round y_score values
when creating ROC curves; this was causing problems for users with very small differences in scores (#7353).

Miscellaneous
• model_selection.tests._search._check_param_grid now works correctly with all types that
extends/implements Sequence (except string), including range (Python 3.x) and xrange (Python 2.x). #7323 by
Viacheslav Kovalevskyi.
• utils.extmath.randomized_range_finder is more numerically stable when many power iterations
are requested, since it applies LU normalization by default. If n_iter<2 numerical issues are unlikely, thus
no normalization is applied. Other normalization options are available: 'none', 'LU' and 'QR'. #5141 by
Giorgio Patrini.
• Fix a bug where some formats of scipy.sparse matrix, and estimators with them as parameters, could not
be passed to base.clone. By Loic Esteve.
• datasets.load_svmlight_file now is able to read long int QID values. #7101 by Ibraim Ganiev.
API changes summary

• residual_metric has been deprecated in linear_model.RANSACRegressor. Use loss instead.
By Manoj Kumar.
• Access to public attributes .X_ and .y_ has been deprecated in isotonic.IsotonicRegression. By
Jonathan Arfa.
• The old mixture.DPGMM is deprecated in favor of the new mixture.BayesianGaussianMixture
(with the parameter weight_concentration_prior_type='dirichlet_process'). The new
class solves the computational problems of the old class and computes the Gaussian mixture with a Dirich-
let process prior faster than before. #7295 by Wei Xue and Thierry Guillemot.
• The old mixture.VBGMM is deprecated in favor of the new mixture.BayesianGaussianMixture
(with the parameter weight_concentration_prior_type='dirichlet_distribution'). The
new class solves the computational problems of the old class and computes the Variational Bayesian Gaussian
mixture faster than before. #6651 by Wei Xue and Thierry Guillemot.
• The old mixture.GMM is deprecated in favor of the new mixture.GaussianMixture. The new class
computes the Gaussian mixture faster than before and some of computational problems have been solved. #6666
by Wei Xue and Thierry Guillemot.
• The sklearn.cross_validation, sklearn.grid_search and sklearn.learning_curve
have been deprecated and the classes and functions have been reorganized into the sklearn.
model_selection module. Ref Model Selection Enhancements and API Changes for more information.
#4294 by Raghav RV.
• The grid_scores_ attribute of model_selection.GridSearchCV and model_selection.
RandomizedSearchCV is deprecated in favor of the attribute cv_results_. Ref Model Selection En-
hancements and API Changes for more information. #6697 by Raghav RV.
• The parameters n_iter or n_folds in old CV splitters are replaced by the new parameter n_splits since
it can provide a consistent and unambiguous interface to represent the number of train-test splits. #7187 by
YenChen Lin.
• classes parameter was renamed to labels in metrics.hamming_loss. #7260 by Sebastián Vanrell.

• The splitter classes LabelKFold, LabelShuffleSplit, LeaveOneLabelOut and

LeavePLabelsOut are renamed to model_selection.GroupKFold, model_selection.
GroupShuffleSplit, model_selection.LeaveOneGroupOut and model_selection.
LeavePGroupsOut respectively. Also the parameter labels in the split method of the newly renamed
splitters model_selection.LeaveOneGroupOut and model_selection.LeavePGroupsOut
is renamed to groups. Additionally in model_selection.LeavePGroupsOut, the parameter
n_labels is renamed to n_groups. #6660 by Raghav RV.
• Error and loss names for scoring parameters are now prefixed by 'neg_', such as
neg_mean_squared_error. The unprefixed versions are deprecated and will be removed in version 0.20.
#7261 by Tim Head.
Code Contributors
Aditya Joshi, Alejandro, Alexander Fabisch, Alexander Loginov, Alexander Minyushkin, Alexander Rudy, Alexan-
dre Abadie, Alexandre Abraham, Alexandre Gramfort, Alexandre Saint, alexfields, Alvaro Ulloa, alyssaq, Amlan
Kar, Andreas Mueller, andrew giessel, Andrew Jackson, Andrew McCulloh, Andrew Murray, Anish Shah, Arafat,
Archit Sharma, Ariel Rokem, Arnaud Joly, Arnaud Rachez, Arthur Mensch, Ash Hoover, asnt, b0noI, Behzad Tabib-
ian, Bernardo, Bernhard Kratzwald, Bhargav Mangipudi, blakeflei, Boyuan Deng, Brandon Carter, Brett Naul, Brian
McFee, Caio Oliveira, Camilo Lamus, Carol Willing, Cass, CeShine Lee, Charles Truong, Chyi-Kwei Yau, CJ Carey,
codevig, Colin Ni, Dan Shiebler, Daniel, Daniel Hnyk, David Ellis, David Nicholson, David Staub, David Thaler,
David Warshaw, Davide Lasagna, Deborah, definitelyuncertain, Didi Bar-Zev, djipey, dsquareindia, edwinENSAE,
Elias Kuthe, Elvis DOHMATOB, Ethan White, Fabian Pedregosa, Fabio Ticconi, fisache, Florian Wilhelm, Francis,
Francis O’Donovan, Gael Varoquaux, Ganiev Ibraim, ghg, Gilles Louppe, Giorgio Patrini, Giovanni Cherubin, Gio-
vanni Lanzani, Glenn Qian, Gordon Mohr, govin-vatsan, Graham Clenaghan, Greg Reda, Greg Stupp, Guillaume
Lemaitre, Gustav Mörtberg, halwai, Harizo Rajaona, Harry Mavroforakis, hashcode55, hdmetor, Henry Lin, Hob-
son Lane, Hugo Bowne-Anderson, Igor Andriushchenko, Imaculate, Inki Hwang, Isaac Sijaranamual, Ishank Gulati,
Issam Laradji, Iver Jordal, jackmartin, Jacob Schreiber, Jake Vanderplas, James Fiedler, James Routley, Jan Zikes,
Janna Brettingen, jarfa, Jason Laska, jblackburne, jeff levesque, Jeffrey Blackburne, Jeffrey04, Jeremy Hintz, jere-
mynixon, Jeroen, Jessica Yung, Jill-Jênn Vie, Jimmy Jia, Jiyuan Qian, Joel Nothman, johannah, John, John Boersma,
John Kirkham, John Moeller, jonathan.striebel, joncrall, Jordi, Joseph Munoz, Joshua Cook, JPFrancoia, jrfiedler,
JulianKahnert, juliathebrave, kaichogami, KamalakerDadi, Kenneth Lyons, Kevin Wang, kingjr, kjell, Konstantin
Podshumok, Kornel Kielczewski, Krishna Kalyan, krishnakalyan3, Kvle Putnam, Kyle Jackson, Lars Buitinck, ldavid,
LeiG, LeightonZhang, Leland McInnes, Liang-Chi Hsieh, Lilian Besson, lizsz, Loic Esteve, Louis Tiao, Léonie Borne,
Mads Jensen, Maniteja Nandana, Manoj Kumar, Manvendra Singh, Marco, Mario Krell, Mark Bao, Mark Szepieniec,
Martin Madsen, MartinBpr, MaryanMorel, Massil, Matheus, Mathieu Blondel, Mathieu Dubois, Matteo, Matthias Ek-
man, Max Moroz, Michael Scherer, michiaki ariga, Mikhail Korobov, Moussa Taifi, mrandrewandrade, Mridul Seth,
nadya-p, Naoya Kanai, Nate George, Nelle Varoquaux, Nelson Liu, Nick James, NickleDave, Nico, Nicolas Goix,
Nikolay Mayorov, ningchi, nlathia, okbalefthanded, Okhlopkov, Olivier Grisel, Panos Louridas, Paul Strickland, Per-
rine Letellier, pestrickland, Peter Fischer, Pieter, Ping-Yao, Chang, practicalswift, Preston Parry, Qimu Zheng, Rachit
Kansal, Raghav RV, Ralf Gommers, Ramana.S, Rammig, Randy Olson, Rob Alexander, Robert Lutz, Robin Schucker,
Rohan Jain, Ruifeng Zheng, Ryan Yu, Rémy Léone, saihttam, Saiwing Yeung, Sam Shleifer, Samuel St-Jean, Sar-
taj Singh, Sasank Chilamkurthy, saurabh.bansod, Scott Andrews, Scott Lowe, seales, Sebastian Raschka, Sebastian
Saeger, Sebastián Vanrell, Sergei Lebedev, shagun Sodhani, shanmuga cv, Shashank Shekhar, shawpan, shengxid-
uan, Shota, shuckle16, Skipper Seabold, sklearn-ci, SmedbergM, srvanrell, Sébastien Lerique, Taranjeet, themrmax,
Thierry, Thierry Guillemot, Thomas, Thomas Hallock, Thomas Moreau, Tim Head, tKammy, toastedcornflakes, Tom,
TomDLT, Toshihiro Kamishima, tracer0tong, Trent Hauck, trevorstephens, Tue Vo, Varun, Varun Jewalikar, Viach-
eslav, Vighnesh Birodkar, Vikram, Villu Ruusmann, Vinayak Mehta, walter, waterponey, Wenhua Yang, Wenjian
Huang, Will Welch, wyseguy7, xyguo, yanlend, Yaroslav Halchenko, yelite, Yen, YenChenLin, Yichuan Liu, Yoav
Ram, Yoshiki, Zheng RuiFeng, zivori, Óscar Nájera

1.15.4 Version 0.17.1
February 18, 2016
Changelog
Bug fixes
• Upgrade vendored joblib to version 0.9.4 that fixes an important bug in joblib.Parallel that can silently
yield to wrong results when working on datasets larger than 1MB: https://github.com/joblib/joblib/blob/0.9.4/
CHANGES.rst
• Fixed reading of Bunch pickles generated with scikit-learn version <= 0.16. This can affect users who have
already downloaded a dataset with scikit-learn 0.16 and are loading it with scikit-learn 0.17. See #6196 for how
this affected datasets.fetch_20newsgroups. By Loic Esteve.
• Fixed a bug that prevented using ROC AUC score to perform grid search on several CPU / cores on large arrays.
See #6147 By Olivier Grisel.
• Fixed a bug that prevented to properly set the presort parameter in ensemble.
GradientBoostingRegressor. See #5857 By Andrew McCulloh.
• Fixed a joblib error when evaluating the perplexity of a decomposition.
LatentDirichletAllocation model. See #6258 By Chyi-Kwei Yau.
1.15.5 Version 0.17
November 5, 2015
Changelog
New features
• All the Scaler classes but preprocessing.RobustScaler can be fitted online by calling partial_fit. By
Giorgio Patrini.
• The new class ensemble.VotingClassifier implements a “majority rule” / “soft voting” ensemble
classifier to combine estimators for classification. By Sebastian Raschka.
• The new class preprocessing.RobustScaler provides an alternative to preprocessing.
StandardScaler for feature-wise centering and range normalization that is robust to outliers. By Thomas
Unterthiner.
• The new class preprocessing.MaxAbsScaler provides an alternative to preprocessing.
MinMaxScaler for feature-wise range normalization when the data is already centered or sparse. By Thomas
Unterthiner.
• The new class preprocessing.FunctionTransformer turns a Python function into a Pipeline-
compatible transformer object. By Joe Jevnik.
• The new classes cross_validation.LabelKFold and cross_validation.
LabelShuffleSplit generate train-test folds, respectively similar to cross_validation.KFold and
cross_validation.ShuffleSplit, except that the folds are conditioned on a label array. By Brian
McFee, Jean Kossaifi and Gilles Louppe.

• decomposition.LatentDirichletAllocation implements the Latent Dirichlet Allocation topic

model with online variational inference. By Chyi-Kwei Yau, with code based on an implementation by Matt
Hoffman. (#3659)
• The new solver sag implements a Stochastic Average Gradient descent and is available in both
linear_model.LogisticRegression and linear_model.Ridge. This solver is very efficient for
large datasets. By Danny Sullivan and Tom Dupre la Tour. (#4738)
• The new solver cd implements a Coordinate Descent in decomposition.NMF. Previous solver based on
Projected Gradient is still available setting new parameter solver to pg, but is deprecated and will be removed
in 0.19, along with decomposition.ProjectedGradientNMF and parameters sparseness, eta,
beta and nls_max_iter. New parameters alpha and l1_ratio control L1 and L2 regularization, and
shuffle adds a shuffling step in the cd solver. By Tom Dupre la Tour and Mathieu Blondel.
Enhancements
• manifold.TSNE now supports approximate optimization via the Barnes-Hut method, leading to much faster
fitting. By Christopher Erick Moody. (#4025)
• cluster.mean_shift_.MeanShift now supports parallel execution, as implemented in the
mean_shift function. By Martino Sorbaro.
• naive_bayes.GaussianNB now supports fitting with sample_weight. By Jan Hendrik Metzen.
• dummy.DummyClassifier now supports a prior fitting strategy. By Arnaud Joly.
• Added a fit_predict method for mixture.GMM and subclasses. By Cory Lorenz.
• Added the metrics.label_ranking_loss metric. By Arnaud Joly.
• Added the metrics.cohen_kappa_score metric.
• Added a warm_start constructor parameter to the bagging ensemble models to increase the size of the en-
semble. By Tim Head.
• Added option to use multi-output regression metrics without averaging. By Konstantin Shmelkov and Michael
Eickenberg.
• Added stratify option to cross_validation.train_test_split for stratified splitting. By
Miroslav Batchkarov.
• The tree.export_graphviz function now supports aesthetic improvements for tree.
DecisionTreeClassifier and tree.DecisionTreeRegressor, including options for coloring
nodes by their majority class or impurity, showing variable names, and using node proportions instead of raw
sample counts. By Trevor Stephens.
• Improved speed of newton-cg solver in linear_model.LogisticRegression, by avoiding loss com-
putation. By Mathieu Blondel and Tom Dupre la Tour.
• The class_weight="auto" heuristic in classifiers supporting class_weight was deprecated and re-
placed by the class_weight="balanced" option, which has a simpler formula and interpretation. By
Hanna Wallach and Andreas Müller.
• Add class_weight parameter to automatically weight samples by class frequency for linear_model.
PassiveAggressiveClassifier. By Trevor Stephens.
• Added backlinks from the API reference pages to the user guide. By Andreas Müller.
• The labels parameter to sklearn.metrics.f1_score, sklearn.metrics.fbeta_score,
sklearn.metrics.recall_score and sklearn.metrics.precision_score has been ex-

tended. It is now possible to ignore one or more labels, such as where a multiclass problem has a majority
class to ignore. By Joel Nothman.
• Add sample_weight support to linear_model.RidgeClassifier. By Trevor Stephens.
• Provide an option for sparse output from sklearn.metrics.pairwise.cosine_similarity. By
Jaidev Deshpande.
• Add minmax_scale to provide a function interface for MinMaxScaler. By Thomas Unterthiner.
• dump_svmlight_file now handles multi-label datasets. By Chih-Wei Chang.
• RCV1 dataset loader (sklearn.datasets.fetch_rcv1). By Tom Dupre la Tour.
• The “Wisconsin Breast Cancer” classical two-class classification dataset is now included in scikit-learn, avail-
able with sklearn.dataset.load_breast_cancer.
• Upgraded to joblib 0.9.3 to benefit from the new automatic batching of short tasks. This makes it possible for
scikit-learn to benefit from parallelism when many very short tasks are executed in parallel, for instance by the
grid_search.GridSearchCV meta-estimator with n_jobs > 1 used with a large grid of parameters on
a small dataset. By Vlad Niculae, Olivier Grisel and Loic Esteve.
• For more details about changes in joblib 0.9.3 see the release notes: https://github.com/joblib/joblib/blob/master/
CHANGES.rst#release-093
• Improved speed (3 times per iteration) of decomposition.DictLearning with coordinate descent
method from linear_model.Lasso. By Arthur Mensch.
• Parallel processing (threaded) for queries of nearest neighbors (using the ball-tree) by Nikolay Mayorov.
• Allow datasets.make_multilabel_classification to output a sparse y. By Kashif Rasul.
• cluster.DBSCAN now accepts a sparse matrix of precomputed distances, allowing memory-efficient distance
precomputation. By Joel Nothman.
• tree.DecisionTreeClassifier now exposes an apply method for retrieving the leaf indices samples
are predicted as. By Daniel Galvez and Gilles Louppe.
• Speed up decision tree regressors, random forest regressors, extra trees regressors and gradient boosting estima-
tors by computing a proxy of the impurity improvement during the tree growth. The proxy quantity is such that
the split that maximizes this value also maximizes the impurity improvement. By Arnaud Joly, Jacob Schreiber
and Gilles Louppe.
• Speed up tree based methods by reducing the number of computations needed when computing the impurity
measure taking into account linear relationship of the computed statistics. The effect is particularly visible with
extra trees and on datasets with categorical or sparse features. By Arnaud Joly.
• ensemble.GradientBoostingRegressor and ensemble.GradientBoostingClassifier
now expose an apply method for retrieving the leaf indices each sample ends up in under each try. By Ja-
cob Schreiber.
• Add sample_weight support to linear_model.LinearRegression. By Sonny Hu. (##4881)
• Add n_iter_without_progress to manifold.TSNE to control the stopping criterion. By Santi Vil-
lalba. (#5186)
• Added optional parameter random_state in linear_model.Ridge , to set the seed of the pseudo random
generator used in sag solver. By Tom Dupre la Tour.
• Added optional parameter warm_start in linear_model.LogisticRegression. If set to True, the
solvers lbfgs, newton-cg and sag will be initialized with the coefficients computed in the previous fit. By
Tom Dupre la Tour.

• Added sample_weight support to linear_model.LogisticRegression for the lbfgs,

newton-cg, and sag solvers. By Valentin Stolbunov. Support added to the liblinear solver. By Manoj
Kumar.
• Added optional parameter presort to ensemble.GradientBoostingRegressor and ensemble.
GradientBoostingClassifier, keeping default behavior the same. This allows gradient boosters to
turn off presorting when building deep trees or using sparse data. By Jacob Schreiber.
• Altered metrics.roc_curve to drop unnecessary thresholds by default. By Graham Clenaghan.
• Added feature_selection.SelectFromModel meta-transformer which can be used along with es-
timators that have coef_ or feature_importances_ attribute to select important features of the input data. By
Maheshakya Wijewardena, Joel Nothman and Manoj Kumar.
• Added metrics.pairwise.laplacian_kernel. By Clyde Fare.
• covariance.GraphLasso allows separate control of the convergence criterion for the Elastic-Net subprob-
lem via the enet_tol parameter.
• Improved verbosity in decomposition.DictionaryLearning.
• ensemble.RandomForestClassifier and ensemble.RandomForestRegressor no longer ex-
plicitly store the samples used in bagging, resulting in a much reduced memory footprint for storing random
forest models.
• Added positive option to linear_model.Lars and linear_model.lars_path to force coeffi-
cients to be positive. (#5131)
• Added the X_norm_squared parameter to metrics.pairwise.euclidean_distances to provide
precomputed squared norms for X.
• Added the fit_predict method to pipeline.Pipeline.
• Added the preprocessing.min_max_scale function.
Bug fixes
• Fixed non-determinism in dummy.DummyClassifier with sparse multi-label output. By Andreas Müller.

• Fixed the output shape of linear_model.RANSACRegressor to (n_samples, ). By Andreas Müller.
• Fixed bug in decomposition.DictLearning when n_jobs < 0. By Andreas Müller.
• Fixed bug where grid_search.RandomizedSearchCV could consume a lot of memory for large discrete
grids. By Joel Nothman.
• Fixed bug in linear_model.LogisticRegressionCV where penalty was ignored in the final fit. By
Manoj Kumar.
• Fixed bug in ensemble.forest.ForestClassifier while computing oob_score and X is a
sparse.csc_matrix. By Ankur Ankan.
• All regressors now consistently handle and warn when given y that is of shape (n_samples, 1). By Andreas
Müller and Henry Lin. (#5431)
• Fix in cluster.KMeans cluster reassignment for sparse input by Lars Buitinck.
• Fixed a bug in lda.LDA that could cause asymmetric covariance matrices when using shrinkage. By Martin
Billinger.
• Fixed cross_validation.cross_val_predict for estimators with sparse predictions. By Buddha
Prakash.

• Fixed the predict_proba method of linear_model.LogisticRegression to use soft-max instead

of one-vs-rest normalization. By Manoj Kumar. (#5182)
• Fixed the partial_fit method of linear_model.SGDClassifier when called with
average=True. By Andrew Lamb. (#5282)
• Dataset fetchers use different filenames under Python 2 and Python 3 to avoid pickling compatibility issues. By
Olivier Grisel. (#5355)
• Fixed a bug in naive_bayes.GaussianNB which caused classification results to depend on scale. By Jake
Vanderplas.
• Fixed temporarily linear_model.Ridge, which was incorrect when fitting the intercept in the case of
sparse data. The fix automatically changes the solver to ‘sag’ in this case. #5360 by Tom Dupre la Tour.
• Fixed a performance bug in decomposition.RandomizedPCA on data with a large number of features
and fewer samples. (#4478) By Andreas Müller, Loic Esteve and Giorgio Patrini.
• Fixed bug in cross_decomposition.PLS that yielded unstable and platform dependent output, and failed
on fit_transform. By Arthur Mensch.
• Fixes to the Bunch class used to store datasets.
• Fixed ensemble.plot_partial_dependence ignoring the percentiles parameter.
• Providing a set as vocabulary in CountVectorizer no longer leads to inconsistent results when pickling.
• Fixed the conditions on when a precomputed Gram matrix needs to be recomputed in linear_model.
LinearRegression, linear_model.OrthogonalMatchingPursuit, linear_model.Lasso
and linear_model.ElasticNet.
• Fixed inconsistent memory layout in the coordinate descent solver that affected linear_model.
DictionaryLearning and covariance.GraphLasso. (#5337) By Olivier Grisel.
• manifold.LocallyLinearEmbedding no longer ignores the reg parameter.
• Nearest Neighbor estimators with custom distance metrics can now be pickled. (#4362)
• Fixed a bug in pipeline.FeatureUnion where transformer_weights were not properly handled
when performing grid-searches.
• Fixed a bug in linear_model.LogisticRegression and linear_model.
LogisticRegressionCV when using class_weight='balanced'``òr
``class_weight='auto'. By Tom Dupre la Tour.
• Fixed bug #5495 when doing OVR(SVC(decision_function_shape=”ovr”)). Fixed by Elvis Dohmatob.
API changes summary
• Attribute data_min, data_max and data_range in preprocessing.MinMaxScaler are deprecated and

won’t be available from 0.19. Instead, the class now exposes data_min_, data_max_ and data_range_. By
Giorgio Patrini.
• All Scaler classes now have an scale_ attribute, the feature-wise rescaling applied by their transform methods.
The old attribute std_ in preprocessing.StandardScaler is deprecated and superseded by scale_; it
won’t be available in 0.19. By Giorgio Patrini.
• svm.SVC` and svm.NuSVC now have an decision_function_shape parameter to make their decision
function of shape (n_samples, n_classes) by setting decision_function_shape='ovr'. This
will be the default behavior starting in 0.19. By Andreas Müller.

• Passing 1D data arrays as input to estimators is now deprecated as it caused confusion in how the array ele-
ments should be interpreted as features or as samples. All data arrays are now expected to be explicitly shaped
(n_samples, n_features). By Vighnesh Birodkar.
• lda.LDA and qda.QDA have been moved to discriminant_analysis.
LinearDiscriminantAnalysis and discriminant_analysis.
QuadraticDiscriminantAnalysis.
• The store_covariance and tol parameters have been moved from the fit method to the constructor in
discriminant_analysis.LinearDiscriminantAnalysis and the store_covariances and
tol parameters have been moved from the fit method to the constructor in discriminant_analysis.
QuadraticDiscriminantAnalysis.
• Models inheriting from _LearntSelectorMixin will no longer support the transform methods. (i.e, Ran-
domForests, GradientBoosting, LogisticRegression, DecisionTrees, SVMs and SGD related models). Wrap
these models around the metatransfomer feature_selection.SelectFromModel to remove features
(according to coefs_ or feature_importances_) which are below a certain threshold value instead.
• cluster.KMeans re-runs cluster-assignments in case of non-convergence, to ensure consistency of
predict(X) and labels_. By Vighnesh Birodkar.
• Classifier and Regressor models are now tagged as such using the _estimator_type attribute.
• Cross-validation iterators always provide indices into training and test set, not boolean masks.
• The decision_function on all regressors was deprecated and will be removed in 0.19. Use predict
instead.
• datasets.load_lfw_pairs is deprecated and will be removed in 0.19. Use datasets.
fetch_lfw_pairs instead.
• The deprecated hmm module was removed.
• The deprecated Bootstrap cross-validation iterator was removed.
• The deprecated Ward and WardAgglomerative classes have been removed. Use clustering.
AgglomerativeClustering instead.
• cross_validation.check_cv is now a public function.
• The property residues_ of linear_model.LinearRegression is deprecated and will be removed in
0.19.
• The deprecated n_jobs parameter of linear_model.LinearRegression has been moved to the con-
structor.
• Removed deprecated class_weight parameter from linear_model.SGDClassifier’s fit method.
Use the construction parameter instead.
• The deprecated support for the sequence of sequences (or list of lists) multilabel format was removed. To convert
to and from the supported binary indicator matrix format, use MultiLabelBinarizer.
• The behavior of calling the inverse_transform method of Pipeline.pipeline will change in 0.19.
It will no longer reshape one-dimensional input to two-dimensional input.
• The deprecated attributes indicator_matrix_, multilabel_ and classes_ of preprocessing.
LabelBinarizer were removed.
• Using gamma=0 in svm.SVC and svm.SVR to automatically set the gamma to 1. / n_features is dep-
recated and will be removed in 0.19. Use gamma="auto" instead.

Code Contributors
Aaron Schumacher, Adithya Ganesh, akitty, Alexandre Gramfort, Alexey Grigorev, Ali Baharev, Allen Riddell, Ando
Saabas, Andreas Mueller, Andrew Lamb, Anish Shah, Ankur Ankan, Anthony Erlinger, Ari Rouvinen, Arnaud Joly,
Arnaud Rachez, Arthur Mensch, banilo, Barmaley.exe, benjaminirving, Boyuan Deng, Brett Naul, Brian McFee,
Buddha Prakash, Chi Zhang, Chih-Wei Chang, Christof Angermueller, Christoph Gohlke, Christophe Bourguignat,
Christopher Erick Moody, Chyi-Kwei Yau, Cindy Sridharan, CJ Carey, Clyde-fare, Cory Lorenz, Dan Blanchard,
Daniel Galvez, Daniel Kronovet, Danny Sullivan, Data1010, David, David D Lowe, David Dotson, djipey, Dmitry
Spikhalskiy, Donne Martin, Dougal J. Sutherland, Dougal Sutherland, edson duarte, Eduardo Caro, Eric Larson, Eric
Martin, Erich Schubert, Fernando Carrillo, Frank C. Eckert, Frank Zalkow, Gael Varoquaux, Ganiev Ibraim, Gilles
Louppe, Giorgio Patrini, giorgiop, Graham Clenaghan, Gryllos Prokopis, gwulfs, Henry Lin, Hsuan-Tien Lin, Im-
manuel Bayer, Ishank Gulati, Jack Martin, Jacob Schreiber, Jaidev Deshpande, Jake Vanderplas, Jan Hendrik Metzen,
Jean Kossaifi, Jeffrey04, Jeremy, jfraj, Jiali Mei, Joe Jevnik, Joel Nothman, John Kirkham, John Wittenauer, Joseph,
Joshua Loyal, Jungkook Park, KamalakerDadi, Kashif Rasul, Keith Goodman, Kian Ho, Konstantin Shmelkov, Kyler
Brown, Lars Buitinck, Lilian Besson, Loic Esteve, Louis Tiao, maheshakya, Maheshakya Wijewardena, Manoj Ku-
mar, MarkTab marktab.net, Martin Ku, Martin Spacek, MartinBpr, martinosorb, MaryanMorel, Masafumi Oyamada,
Mathieu Blondel, Matt Krump, Matti Lyra, Maxim Kolganov, mbillinger, mhg, Michael Heilman, Michael Patterson,
Miroslav Batchkarov, Nelle Varoquaux, Nicolas, Nikolay Mayorov, Olivier Grisel, Omer Katz, Óscar Nájera, Pauli
Virtanen, Peter Fischer, Peter Prettenhofer, Phil Roth, pianomania, Preston Parry, Raghav RV, Rob Zinkov, Robert
Layton, Rohan Ramanath, Saket Choudhary, Sam Zhang, santi, saurabh.bansod, scls19fr, Sebastian Raschka, Sebas-
tian Saeger, Shivan Sornarajah, SimonPL, sinhrks, Skipper Seabold, Sonny Hu, sseg, Stephen Hoover, Steven De
Gryze, Steven Seguin, Theodore Vasiloudis, Thomas Unterthiner, Tiago Freitas Pereira, Tian Wang, Tim Head, Timo-
thy Hopper, tokoroten, Tom Dupré la Tour, Trevor Stephens, Valentin Stolbunov, Vighnesh Birodkar, Vinayak Mehta,
Vincent, Vincent Michel, vstolbunov, wangz10, Wei Xue, Yucheng Low, Yury Zhauniarovich, Zac Stewart, zhai_pro,
Zichen Wang
1.15.6 Version 0.16.1
April 14, 2015
Changelog
Bug fixes
• Allow input data larger than block_size in covariance.LedoitWolf by Andreas Müller.

• Fix a bug in isotonic.IsotonicRegression deduplication that caused unstable result in
calibration.CalibratedClassifierCV by Jan Hendrik Metzen.
• Fix sorting of labels in func:preprocessing.label_binarize by Michael Heilman.
• Fix several stability and convergence issues in cross_decomposition.CCA and
cross_decomposition.PLSCanonical by Andreas Müller
• Fix a bug in cluster.KMeans when precompute_distances=False on fortran-ordered data.
• Fix a speed regression in ensemble.RandomForestClassifier’s predict and predict_proba
by Andreas Müller.
• Fix a regression where utils.shuffle converted lists and dataframes to arrays, by Olivier Grisel
1.15.7 Version 0.16
March 26, 2015

Highlights
• Speed improvements (notably in cluster.DBSCAN ), reduced memory requirements, bug-fixes and better
default settings.
• Multinomial Logistic regression and a path algorithm in linear_model.LogisticRegressionCV .
• Out-of core learning of PCA via decomposition.IncrementalPCA.
• Probability callibration of classifiers using calibration.CalibratedClassifierCV .
• cluster.Birch clustering method for large-scale datasets.
• Scalable approximate nearest neighbors search with Locality-sensitive hashing forests in neighbors.
LSHForest.
• Improved error messages and better validation when using malformed input data.
• More robust integration with pandas dataframes.
Changelog
New features
• The new neighbors.LSHForest implements locality-sensitive hashing for approximate nearest neighbors
search. By Maheshakya Wijewardena.
• Added svm.LinearSVR. This class uses the liblinear implementation of Support Vector Regression which is
much faster for large sample sizes than svm.SVR with linear kernel. By Fabian Pedregosa and Qiang Luo.
• Incremental fit for GaussianNB.
• Added sample_weight support to dummy.DummyClassifier and dummy.DummyRegressor. By
Arnaud Joly.
• Added the metrics.label_ranking_average_precision_score metrics. By Arnaud Joly.
• Add the metrics.coverage_error metrics. By Arnaud Joly.
• Added linear_model.LogisticRegressionCV . By Manoj Kumar, Fabian Pedregosa, Gael Varoquaux
and Alexandre Gramfort.
• Added warm_start constructor parameter to make it possible for any trained forest model to grow additional
trees incrementally. By Laurent Direr.
• Added sample_weight support to ensemble.GradientBoostingClassifier and ensemble.
GradientBoostingRegressor. By Peter Prettenhofer.
• Added decomposition.IncrementalPCA, an implementation of the PCA algorithm that supports out-
of-core learning with a partial_fit method. By Kyle Kastner.
• Averaged SGD for SGDClassifier and SGDRegressor By Danny Sullivan.
• Added cross_val_predict function which computes cross-validated estimates. By Luis Pedro Coelho
• Added linear_model.TheilSenRegressor, a robust generalized-median-based estimator. By Florian
Wilhelm.
• Added metrics.median_absolute_error, a robust metric. By Gael Varoquaux and Florian Wilhelm.
• Add cluster.Birch, an online clustering algorithm. By Manoj Kumar, Alexandre Gramfort and Joel Noth-
man.

• Added shrinkage support to discriminant_analysis.LinearDiscriminantAnalysis using two

new solvers. By Clemens Brunner and Martin Billinger.
• Added kernel_ridge.KernelRidge, an implementation of kernelized ridge regression. By Mathieu
Blondel and Jan Hendrik Metzen.
• All solvers in linear_model.Ridge now support sample_weight. By Mathieu Blondel.
• Added cross_validation.PredefinedSplit cross-validation for fixed user-provided cross-validation
folds. By Thomas Unterthiner.
• Added calibration.CalibratedClassifierCV , an approach for calibrating the predicted probabili-
ties of a classifier. By Alexandre Gramfort, Jan Hendrik Metzen, Mathieu Blondel and Balazs Kegl.
Enhancements
• Add option return_distance in hierarchical.ward_tree to return distances between nodes for

both structured and unstructured versions of the algorithm. By Matteo Visconti di Oleggio Castello. The same
option was added in hierarchical.linkage_tree. By Manoj Kumar
• Add support for sample weights in scorer objects. Metrics with sample weight support will automatically benefit
from it. By Noel Dawe and Vlad Niculae.
• Added newton-cg and lbfgs solver support in linear_model.LogisticRegression. By Manoj Ku-
mar.
• Add selection="random" parameter to implement stochastic coordinate descent for linear_model.
Lasso, linear_model.ElasticNet and related. By Manoj Kumar.
• Add sample_weight parameter to metrics.jaccard_similarity_score and metrics.
log_loss. By Jatin Shah.
• Support sparse multilabel indicator representation in preprocessing.LabelBinarizer and
multiclass.OneVsRestClassifier (by Hamzeh Alsalhi with thanks to Rohit Sivaprasad), as
well as evaluation metrics (by Joel Nothman).
• Add sample_weight parameter to metrics.jaccard_similarity_score. By Jatin Shah.
• Add support for multiclass in metrics.hinge_loss. Added labels=None as optional parameter. By Saurabh
Jha.
• Add sample_weight parameter to metrics.hinge_loss. By Saurabh Jha.
• Add multi_class="multinomial" option in linear_model.LogisticRegression to imple-
ment a Logistic Regression solver that minimizes the cross-entropy or multinomial loss instead of the default
One-vs-Rest setting. Supports lbfgs and newton-cg solvers. By Lars Buitinck and Manoj Kumar. Solver option
newton-cg by Simon Wu.
• DictVectorizer can now perform fit_transform on an iterable in a single pass, when giving the option
sort=False. By Dan Blanchard.
• GridSearchCV and RandomizedSearchCV can now be configured to work with estimators that may fail
and raise errors on individual folds. This option is controlled by the error_score parameter. This does not affect
errors raised on re-fit. By Michal Romaniuk.
• Add digits parameter to metrics.classification_report to allow report to show different precision of floating
point numbers. By Ian Gilmore.
• Add a quantile prediction strategy to the dummy.DummyRegressor. By Aaron Staple.
• Add handle_unknown option to preprocessing.OneHotEncoder to handle unknown categorical fea-
tures more gracefully during transform. By Manoj Kumar.

• Added support for sparse input data to decision trees and their ensembles. By Fares Hedyati and Arnaud Joly.
• Optimized cluster.AffinityPropagation by reducing the number of memory allocations of large
temporary data-structures. By Antony Lee.
• Parellization of the computation of feature importances in random forest. By Olivier Grisel and Arnaud Joly.
• Add n_iter_ attribute to estimators that accept a max_iter attribute in their constructor. By Manoj Kumar.
• Added decision function for multiclass.OneVsOneClassifier By Raghav RV and Kyle Beauchamp.
• neighbors.kneighbors_graph and radius_neighbors_graph support non-Euclidean metrics.
By Manoj Kumar
• Parameter connectivity in cluster.AgglomerativeClustering and family now accept callables
that return a connectivity matrix. By Manoj Kumar.
• Sparse support for paired_distances. By Joel Nothman.
• cluster.DBSCAN now supports sparse input and sample weights and has been optimized: the inner loop has
been rewritten in Cython and radius neighbors queries are now computed in batch. By Joel Nothman and Lars
Buitinck.
• Add class_weight parameter to automatically weight samples by class frequency for
ensemble.RandomForestClassifier, tree.DecisionTreeClassifier, ensemble.
ExtraTreesClassifier and tree.ExtraTreeClassifier. By Trevor Stephens.
• grid_search.RandomizedSearchCV now does sampling without replacement if all parameters are given
as lists. By Andreas Müller.
• Parallelized calculation of pairwise_distances is now supported for scipy metrics and custom callables.
By Joel Nothman.
• Allow the fitting and scoring of all clustering algorithms in pipeline.Pipeline. By Andreas Müller.
• More robust seeding and improved error messages in cluster.MeanShift by Andreas Müller.
• Make the stopping criterion for mixture.GMM, mixture.DPGMM and mixture.VBGMM less dependent on
the number of samples by thresholding the average log-likelihood change instead of its sum over all samples.
By Hervé Bredin.
• The outcome of manifold.spectral_embedding was made deterministic by flipping the sign of eigen-
vectors. By Hasil Sharma.
• Significant performance and memory usage improvements in preprocessing.PolynomialFeatures.
By Eric Martin.
• Numerical stability improvements for preprocessing.StandardScaler and preprocessing.
scale. By Nicolas Goix
• svm.SVC fitted on sparse input now implements decision_function. By Rob Zinkov and Andreas
Müller.
• cross_validation.train_test_split now preserves the input type, instead of converting to numpy
arrays.
Documentation improvements
• Added example of using FeatureUnion for heterogeneous input. By Matt Terry

• Documentation on scorers was improved, to highlight the handling of loss functions. By Matt Pico.
• A discrepancy between liblinear output and scikit-learn’s wrappers is now noted. By Manoj Kumar.

• Improved documentation generation: examples referring to a class or function are now shown in a gallery on
the class/function’s API reference page. By Joel Nothman.
• More explicit documentation of sample generators and of data transformation. By Joel Nothman.
• sklearn.neighbors.BallTree and sklearn.neighbors.KDTree used to point to empty pages
stating that they are aliases of BinaryTree. This has been fixed to show the correct class docs. By Manoj Kumar.
• Added silhouette plots for analysis of KMeans clustering using metrics.silhouette_samples and
metrics.silhouette_score. See Selecting the number of clusters with silhouette analysis on KMeans
clustering
Bug fixes
• Metaestimators now support ducktyping for the presence of decision_function,

predict_proba and other methods. This fixes behavior of grid_search.GridSearchCV,
grid_search.RandomizedSearchCV, pipeline.Pipeline, feature_selection.RFE,
feature_selection.RFECV when nested. By Joel Nothman
• The scoring attribute of grid-search and cross-validation methods is no longer ignored when a
grid_search.GridSearchCV is given as a base estimator or the base estimator doesn’t have predict.
• The function hierarchical.ward_tree now returns the children in the same order for both the structured
and unstructured versions. By Matteo Visconti di Oleggio Castello.
• feature_selection.RFECV now correctly handles cases when step is not equal to 1. By Nikolay
Mayorov
• The decomposition.PCA now undoes whitening in its inverse_transform. Also, its components_
now always have unit length. By Michael Eickenberg.
• Fix incomplete download of the dataset when datasets.download_20newsgroups is called. By Manoj
Kumar.
• Various fixes to the Gaussian processes subpackage by Vincent Dubourg and Jan Hendrik Metzen.
• Calling partial_fit with class_weight=='auto' throws an appropriate error message and suggests
a work around. By Danny Sullivan.
• RBFSampler with gamma=g formerly approximated rbf_kernel with gamma=g/2.; the definition of
gamma is now consistent, which may substantially change your results if you use a fixed value. (If you cross-
validated over gamma, it probably doesn’t matter too much.) By Dougal Sutherland.
• Pipeline object delegate the classes_ attribute to the underlying estimator. It allows, for instance, to make
bagging of a pipeline object. By Arnaud Joly
• neighbors.NearestCentroid now uses the median as the centroid when metric is set to manhattan.
It was using the mean before. By Manoj Kumar
• Fix numerical stability issues in linear_model.SGDClassifier and linear_model.
SGDRegressor by clipping large gradients and ensuring that weight decay rescaling is always positive (for
large l2 regularization and large learning rate values). By Olivier Grisel
• When compute_full_tree is set to “auto”, the full tree is built when n_clusters is high and is early stopped when
n_clusters is low, while the behavior should be vice-versa in cluster.AgglomerativeClustering (and
friends). This has been fixed By Manoj Kumar
• Fix lazy centering of data in linear_model.enet_path and linear_model.lasso_path. It was
centered around one. It has been changed to be centered around the origin. By Manoj Kumar

• Fix handling of precomputed affinity matrices in cluster.AgglomerativeClustering when using

connectivity constraints. By Cathy Deng
• Correct partial_fit handling of class_prior for sklearn.naive_bayes.MultinomialNB and
sklearn.naive_bayes.BernoulliNB. By Trevor Stephens.
• Fixed a crash in metrics.precision_recall_fscore_support when using unsorted labels in the
multi-label setting. By Andreas Müller.
• Avoid skipping the first nearest neighbor in the methods radius_neighbors, kneighbors,
kneighbors_graph and radius_neighbors_graph in sklearn.neighbors.
NearestNeighbors and family, when the query data is not the same as fit data. By Manoj Kumar.
• Fix log-density calculation in the mixture.GMM with tied covariance. By Will Dawson
• Fixed a scaling error in feature_selection.SelectFdr where a factor n_features was missing. By
Andrew Tulloch
• Fix zero division in neighbors.KNeighborsRegressor and related classes when using distance weight-
ing and having identical data points. By Garret-R.
• Fixed round off errors with non positive-definite covariance matrices in GMM. By Alexis Mignon.
• Fixed a error in the computation of conditional probabilities in naive_bayes.BernoulliNB. By Hanna
Wallach.
• Make the method radius_neighbors of neighbors.NearestNeighbors return the samples lying
on the boundary for algorithm='brute'. By Yan Yi.
• Flip sign of dual_coef_ of svm.SVC to make it consistent with the documentation and
decision_function. By Artem Sobolev.
• Fixed handling of ties in isotonic.IsotonicRegression. We now use the weighted average of targets
(secondary method). By Andreas Müller and Michael Bommarito.
API changes summary
• GridSearchCV and cross_val_score and other meta-estimators don’t convert pandas DataFrames into
arrays any more, allowing DataFrame specific operations in custom estimators.
• multiclass.fit_ovr, multiclass.predict_ovr, predict_proba_ovr, multiclass.
fit_ovo, multiclass.predict_ovo, multiclass.fit_ecoc and multiclass.
predict_ecoc are deprecated. Use the underlying estimators instead.
• Nearest neighbors estimators used to take arbitrary keyword arguments and pass these to their distance metric.
This will no longer be supported in scikit-learn 0.18; use the metric_params argument instead.
• n_jobs parameter of the fit method shifted to the constructor of the LinearRegression class.
• The predict_proba method of multiclass.OneVsRestClassifier now returns two probabilities
per sample in the multiclass case; this is consistent with other estimators and with the method’s documenta-
tion, but previous versions accidentally returned only the positive probability. Fixed by Will Lamond and Lars
Buitinck.
• Change default value of precompute in ElasticNet and Lasso to False. Setting precompute to “auto” was
found to be slower when n_samples > n_features since the computation of the Gram matrix is computationally
expensive and outweighs the benefit of fitting the Gram for just one alpha. precompute="auto" is now
deprecated and will be removed in 0.18 By Manoj Kumar.
• Expose positive option in linear_model.enet_path and linear_model.enet_path which
constrains coefficients to be positive. By Manoj Kumar.

• Users should now supply an explicit average parameter to sklearn.metrics.f1_score, sklearn.

metrics.fbeta_score, sklearn.metrics.recall_score and sklearn.metrics.
precision_score when performing multiclass or multilabel (i.e. not binary) classification. By Joel
Nothman.
• scoring parameter for cross validation now accepts ‘f1_micro’, ‘f1_macro’ or ‘f1_weighted’. ‘f1’ is now for
binary classification only. Similar changes apply to ‘precision’ and ‘recall’. By Joel Nothman.
• The fit_intercept, normalize and return_models parameters in linear_model.enet_path
and linear_model.lasso_path have been removed. They were deprecated since 0.14
• From now onwards, all estimators will uniformly raise NotFittedError (utils.validation.
NotFittedError), when any of the predict like methods are called before the model is fit. By Raghav
RV.
• Input data validation was refactored for more consistent input validation. The check_arrays function was
replaced by check_array and check_X_y. By Andreas Müller.
• Allow X=None in the methods radius_neighbors, kneighbors, kneighbors_graph and
radius_neighbors_graph in sklearn.neighbors.NearestNeighbors and family. If set to
None, then for every sample this avoids setting the sample itself as the first nearest neighbor. By Manoj Kumar.
• Add parameter include_self in neighbors.kneighbors_graph and neighbors.
radius_neighbors_graph which has to be explicitly set by the user. If set to True, then the
sample itself is considered as the first nearest neighbor.
• thresh parameter is deprecated in favor of new tol parameter in GMM, DPGMM and VBGMM. See Enhancements
section for details. By Hervé Bredin.
• Estimators will treat input with dtype object as numeric when possible. By Andreas Müller
• Estimators now raise ValueError consistently when fitted on empty data (less than 1 sample or less than 1 feature
for 2D input). By Olivier Grisel.
• The shuffle option of linear_model.SGDClassifier, linear_model.SGDRegressor,
linear_model.Perceptron, linear_model.PassiveAggressiveClassifier and
linear_model.PassiveAggressiveRegressor now defaults to True.
• cluster.DBSCAN now uses a deterministic initialization. The random_state parameter is deprecated. By
Erich Schubert.
Code Contributors
A. Flaxman, Aaron Schumacher, Aaron Staple, abhishek thakur, Akshay, akshayah3, Aldrian Obaja, Alexander
Fabisch, Alexandre Gramfort, Alexis Mignon, Anders Aagaard, Andreas Mueller, Andreas van Cranenburgh, An-
drew Tulloch, Andrew Walker, Antony Lee, Arnaud Joly, banilo, Barmaley.exe, Ben Davies, Benedikt Koehler, bhsu,
Boris Feld, Borja Ayerdi, Boyuan Deng, Brent Pedersen, Brian Wignall, Brooke Osborn, Calvin Giles, Cathy Deng,
Celeo, cgohlke, chebee7i, Christian Stade-Schuldt, Christof Angermueller, Chyi-Kwei Yau, CJ Carey, Clemens Brun-
ner, Daiki Aminaka, Dan Blanchard, danfrankj, Danny Sullivan, David Fletcher, Dmitrijs Milajevs, Dougal J. Suther-
land, Erich Schubert, Fabian Pedregosa, Florian Wilhelm, floydsoft, Félix-Antoine Fortin, Gael Varoquaux, Garrett-R,
Gilles Louppe, gpassino, gwulfs, Hampus Bengtsson, Hamzeh Alsalhi, Hanna Wallach, Harry Mavroforakis, Hasil
Sharma, Helder, Herve Bredin, Hsiang-Fu Yu, Hugues SALAMIN, Ian Gilmore, Ilambharathi Kanniah, Imran Haque,
isms, Jake VanderPlas, Jan Dlabal, Jan Hendrik Metzen, Jatin Shah, Javier López Peña, jdcaballero, Jean Kossaifi, Jeff
Hammerbacher, Joel Nothman, Jonathan Helmus, Joseph, Kaicheng Zhang, Kevin Markham, Kyle Beauchamp, Kyle
Kastner, Lagacherie Matthieu, Lars Buitinck, Laurent Direr, leepei, Loic Esteve, Luis Pedro Coelho, Lukas Michel-
bacher, maheshakya, Manoj Kumar, Manuel, Mario Michael Krell, Martin, Martin Billinger, Martin Ku, Mateusz
Susik, Mathieu Blondel, Matt Pico, Matt Terry, Matteo Visconti dOC, Matti Lyra, Max Linke, Mehdi Cherti, Michael
Bommarito, Michael Eickenberg, Michal Romaniuk, MLG, mr.Shu, Nelle Varoquaux, Nicola Montecchio, Nicolas,
Nikolay Mayorov, Noel Dawe, Okal Billy, Olivier Grisel, Óscar Nájera, Paolo Puggioni, Peter Prettenhofer, Pratap

Vardhan, pvnguyen, queqichao, Rafael Carrascosa, Raghav R V, Rahiel Kasim, Randall Mason, Rob Zinkov, Robert
Bradshaw, Saket Choudhary, Sam Nicholls, Samuel Charron, Saurabh Jha, sethdandridge, sinhrks, snuderl, Stefan
Otte, Stefan van der Walt, Steve Tjoa, swu, Sylvain Zimmer, tejesh95, terrycojones, Thomas Delteil, Thomas Un-
terthiner, Tomas Kazmar, trevorstephens, tttthomasssss, Tzu-Ming Kuo, ugurcaliskan, ugurthemaster, Vinayak Mehta,
Vincent Dubourg, Vjacheslav Murashkin, Vlad Niculae, wadawson, Wei Xue, Will Lamond, Wu Jiang, x0l, Xinfan
Meng, Yan Yi, Yu-Chin
1.15.8 Version 0.15.2
September 4, 2014
Bug fixes
• Fixed handling of the p parameter of the Minkowski distance that was previously ignored in nearest neighbors
models. By Nikolay Mayorov.
• Fixed duplicated alphas in linear_model.LassoLars with early stopping on 32 bit Python. By Olivier
Grisel and Fabian Pedregosa.
• Fixed the build under Windows when scikit-learn is built with MSVC while NumPy is built with MinGW. By
Olivier Grisel and Federico Vaggi.
• Fixed an array index overflow bug in the coordinate descent solver. By Gael Varoquaux.
• Better handling of numpy 1.9 deprecation warnings. By Gael Varoquaux.
• Removed unnecessary data copy in cluster.KMeans. By Gael Varoquaux.
• Explicitly close open files to avoid ResourceWarnings under Python 3. By Calvin Giles.
• The transform of discriminant_analysis.LinearDiscriminantAnalysis now projects the
input on the most discriminant directions. By Martin Billinger.
• Fixed potential overflow in _tree.safe_realloc by Lars Buitinck.
• Performance optimization in isotonic.IsotonicRegression. By Robert Bradshaw.
• nose is non-longer a runtime dependency to import sklearn, only for running the tests. By Joel Nothman.
• Many documentation and website fixes by Joel Nothman, Lars Buitinck Matt Pico, and others.
1.15.9 Version 0.15.1
August 1, 2014
Bug fixes
• Made cross_validation.cross_val_score use cross_validation.KFold instead of

cross_validation.StratifiedKFold on multi-output classification problems. By Nikolay Mayorov.
• Support unseen labels preprocessing.LabelBinarizer to restore the default behavior of 0.14.1 for
backward compatibility. By Hamzeh Alsalhi.
• Fixed the cluster.KMeans stopping criterion that prevented early convergence detection. By Edward Raff
and Gael Varoquaux.
• Fixed the behavior of multiclass.OneVsOneClassifier. in case of ties at the per-class vote level by
computing the correct per-class sum of prediction scores. By Andreas Müller.

• Made cross_validation.cross_val_score and grid_search.GridSearchCV accept Python

lists as input data. This is especially useful for cross-validation and model selection of text processing pipelines.
By Andreas Müller.
• Fixed data input checks of most estimators to accept input data that implements the NumPy __array__
protocol. This is the case for for pandas.Series and pandas.DataFrame in recent versions of pandas.
By Gael Varoquaux.
• Fixed a regression for linear_model.SGDClassifier with class_weight="auto" on data with
non-contiguous labels. By Olivier Grisel.
1.15.10 Version 0.15
July 15, 2014
Highlights
• Many speed and memory improvements all across the code

• Huge speed and memory improvements to random forests (and extra trees) that also benefit better from parallel
computing.
• Incremental fit to BernoulliRBM
• Added cluster.AgglomerativeClustering for hierarchical agglomerative clustering with average
linkage, complete linkage and ward strategies.
• Added linear_model.RANSACRegressor for robust regression models.
• Added dimensionality reduction with manifold.TSNE which can be used to visualize high-dimensional data.
Changelog
New features
• Added ensemble.BaggingClassifier and ensemble.BaggingRegressor meta-estimators for

ensembling any kind of base estimator. See the Bagging section of the user guide for details and examples.
By Gilles Louppe.
• New unsupervised feature selection algorithm feature_selection.VarianceThreshold, by Lars
Buitinck.
• Added linear_model.RANSACRegressor meta-estimator for the robust fitting of regression models. By
Johannes Schönberger.
• Added cluster.AgglomerativeClustering for hierarchical agglomerative clustering with average
linkage, complete linkage and ward strategies, by Nelle Varoquaux and Gael Varoquaux.
• Shorthand constructors pipeline.make_pipeline and pipeline.make_union were added by Lars
Buitinck.
• Shuffle option for cross_validation.StratifiedKFold. By Jeffrey Blackburne.
• Incremental learning (partial_fit) for Gaussian Naive Bayes by Imran Haque.
• Added partial_fit to BernoulliRBM By Danny Sullivan.
• Added learning_curve utility to chart performance with respect to training size. See Plotting Learning
Curves. By Alexander Fabisch.

• Add positive option in LassoCV and ElasticNetCV . By Brian Wignall and Alexandre Gramfort.
• Added linear_model.MultiTaskElasticNetCV and linear_model.MultiTaskLassoCV . By
Manoj Kumar.
• Added manifold.TSNE. By Alexander Fabisch.
Enhancements
• Add sparse input support to ensemble.AdaBoostClassifier and ensemble.

AdaBoostRegressor meta-estimators. By Hamzeh Alsalhi.
• Memory improvements of decision trees, by Arnaud Joly.
• Decision trees can now be built in best-first manner by using max_leaf_nodes as the stopping criteria.
Refactored the tree code to use either a stack or a priority queue for tree building. By Peter Prettenhofer and
Gilles Louppe.
• Decision trees can now be fitted on fortran- and c-style arrays, and non-continuous arrays without the need to
make a copy. If the input array has a different dtype than np.float32, a fortran- style copy will be made
since fortran-style memory layout has speed advantages. By Peter Prettenhofer and Gilles Louppe.
• Speed improvement of regression trees by optimizing the the computation of the mean square error criterion.
This lead to speed improvement of the tree, forest and gradient boosting tree modules. By Arnaud Joly
• The img_to_graph and grid_tograph functions in sklearn.feature_extraction.image now
return np.ndarray instead of np.matrix when return_as=np.ndarray. See the Notes section for
more information on compatibility.
• Changed the internal storage of decision trees to use a struct array. This fixed some small bugs, while improving
code and providing a small speed gain. By Joel Nothman.
• Reduce memory usage and overhead when fitting and predicting with forests of randomized trees in parallel
with n_jobs != 1 by leveraging new threading backend of joblib 0.8 and releasing the GIL in the tree fitting
Cython code. By Olivier Grisel and Gilles Louppe.
• Speed improvement of the sklearn.ensemble.gradient_boosting module. By Gilles Louppe and
Peter Prettenhofer.
• Various enhancements to the sklearn.ensemble.gradient_boosting module: a warm_start ar-
gument to fit additional trees, a max_leaf_nodes argument to fit GBM style trees, a monitor fit argument
to inspect the estimator during training, and refactoring of the verbose code. By Peter Prettenhofer.
• Faster sklearn.ensemble.ExtraTrees by caching feature values. By Arnaud Joly.
• Faster depth-based tree building algorithm such as decision tree, random forest, extra trees or gradient tree
boosting (with depth based growing strategy) by avoiding trying to split on found constant features in the sample
subset. By Arnaud Joly.
• Add min_weight_fraction_leaf pre-pruning parameter to tree-based methods: the minimum weighted
fraction of the input samples required to be at a leaf node. By Noel Dawe.
• Added metrics.pairwise_distances_argmin_min, by Philippe Gervais.
• Added predict method to cluster.AffinityPropagation and cluster.MeanShift, by Mathieu
Blondel.
• Vector and matrix multiplications have been optimised throughout the library by Denis Engemann, and Alexan-
dre Gramfort. In particular, they should take less memory with older NumPy versions (prior to 1.7.2).
• Precision-recall and ROC examples now use train_test_split, and have more explanation of why these metrics
are useful. By Kyle Kastner

• The training algorithm for decomposition.NMF is faster for sparse matrices and has much lower memory
complexity, meaning it will scale up gracefully to large datasets. By Lars Buitinck.
• Added svd_method option with default value to “randomized” to decomposition.FactorAnalysis to
save memory and significantly speedup computation by Denis Engemann, and Alexandre Gramfort.
• Changed cross_validation.StratifiedKFold to try and preserve as much of the original ordering of
samples as possible so as not to hide overfitting on datasets with a non-negligible level of samples dependency.
By Daniel Nouri and Olivier Grisel.
• Add multi-output support to gaussian_process.GaussianProcess by John Novak.
• Support for precomputed distance matrices in nearest neighbor estimators by Robert Layton and Joel Nothman.
• Norm computations optimized for NumPy 1.6 and later versions by Lars Buitinck. In particular, the k-means
algorithm no longer needs a temporary data structure the size of its input.
• dummy.DummyClassifier can now be used to predict a constant output value. By Manoj Kumar.
• dummy.DummyRegressor has now a strategy parameter which allows to predict the mean, the median of the
training set or a constant output value. By Maheshakya Wijewardena.
• Multi-label classification output in multilabel indicator format is now supported by metrics.
roc_auc_score and metrics.average_precision_score by Arnaud Joly.
• Significant performance improvements (more than 100x speedup for large problems) in isotonic.
IsotonicRegression by Andrew Tulloch.
• Speed and memory usage improvements to the SGD algorithm for linear models: it now uses threads, not
separate processes, when n_jobs>1. By Lars Buitinck.
• Grid search and cross validation allow NaNs in the input arrays so that preprocessors such as
preprocessing.Imputer can be trained within the cross validation loop, avoiding potentially skewed
results.
• Ridge regression can now deal with sample weights in feature space (only sample space until then). By Michael
Eickenberg. Both solutions are provided by the Cholesky solver.
• Several classification and regression metrics now support weighted samples with the new
sample_weight argument: metrics.accuracy_score, metrics.zero_one_loss,
metrics.precision_score, metrics.average_precision_score, metrics.
f1_score, metrics.fbeta_score, metrics.recall_score, metrics.roc_auc_score,
metrics.explained_variance_score, metrics.mean_squared_error, metrics.
mean_absolute_error, metrics.r2_score. By Noel Dawe.
• Speed up of the sample generator datasets.make_multilabel_classification. By Joel Nothman.
Documentation improvements
• The Working With Text Data tutorial has now been worked in to the main documentation’s tutorial section.
Includes exercises and skeletons for tutorial presentation. Original tutorial created by several authors including
Olivier Grisel, Lars Buitinck and many others. Tutorial integration into the scikit-learn documentation by Jaques
Grobler
• Added Computational Performance documentation. Discussion and examples of prediction latency / throughput
and different factors that have influence over speed. Additional tips for building faster models and choosing a
relevant compromise between speed and predictive power. By Eustache Diemert.

Bug fixes
• Fixed bug in decomposition.MiniBatchDictionaryLearning : partial_fit was not working

properly.
• Fixed bug in linear_model.stochastic_gradient : l1_ratio was used as (1.0 - l1_ratio)
.
• Fixed bug in multiclass.OneVsOneClassifier with string labels
• Fixed a bug in LassoCV and ElasticNetCV : they would not pre-compute the Gram matrix with
precompute=True or precompute="auto" and n_samples > n_features. By Manoj Kumar.
• Fixed incorrect estimation of the degrees of freedom in feature_selection.f_regression when vari-
ates are not centered. By Virgile Fritsch.
• Fixed a race condition in parallel processing with pre_dispatch != "all" (for instance, in
cross_val_score). By Olivier Grisel.
• Raise error in cluster.FeatureAgglomeration and cluster.WardAgglomeration when no
samples are given, rather than returning meaningless clustering.
• Fixed bug in gradient_boosting.GradientBoostingRegressor with loss='huber': gamma
might have not been initialized.
• Fixed feature importances as computed with a forest of randomized trees when fit with sample_weight !=
None and/or with bootstrap=True. By Gilles Louppe.
API changes summary
• sklearn.hmm is deprecated. Its removal is planned for the 0.17 release.

• Use of covariance.EllipticEnvelop has now been removed after deprecation. Please use
covariance.EllipticEnvelope instead.
• cluster.Ward is deprecated. Use cluster.AgglomerativeClustering instead.
• cluster.WardClustering is deprecated. Use
• cluster.AgglomerativeClustering instead.
• cross_validation.Bootstrap is deprecated. cross_validation.KFold or
cross_validation.ShuffleSplit are recommended instead.
• Direct support for the sequence of sequences (or list of lists) multilabel format is deprecated. To convert to and
from the supported binary indicator matrix format, use MultiLabelBinarizer. By Joel Nothman.
• Add score method to PCA following the model of probabilistic PCA and deprecate ProbabilisticPCA
model whose score implementation is not correct. The computation now also exploits the matrix inversion
lemma for faster computation. By Alexandre Gramfort.
• The score method of FactorAnalysis now returns the average log-likelihood of the samples. Use
score_samples to get log-likelihood of each sample. By Alexandre Gramfort.
• Generating boolean masks (the setting indices=False) from cross-validation generators is deprecated. Sup-
port for masks will be removed in 0.17. The generators have produced arrays of indices by default since 0.10.
By Joel Nothman.
• 1-d arrays containing strings with dtype=object (as used in Pandas) are now considered valid classification
targets. This fixes a regression from version 0.13 in some classifiers. By Joel Nothman.
• Fix wrong explained_variance_ratio_ attribute in RandomizedPCA. By Alexandre Gramfort.

• Fit alphas for each l1_ratio instead of mean_l1_ratio in linear_model.ElasticNetCV and

linear_model.LassoCV . This changes the shape of alphas_ from (n_alphas,) to (n_l1_ratio,
n_alphas) if the l1_ratio provided is a 1-D array like object of length greater than one. By Manoj Kumar.
• Fix linear_model.ElasticNetCV and linear_model.LassoCV when fitting intercept and input
data is sparse. The automatic grid of alphas was not computed correctly and the scaling with normalize was
wrong. By Manoj Kumar.
• Fix wrong maximal number of features drawn (max_features) at each split for decision trees, random forests
and gradient tree boosting. Previously, the count for the number of drawn features started only after one non
constant features in the split. This bug fix will affect computational and generalization performance of those
algorithms in the presence of constant features. To get back previous generalization performance, you should
modify the value of max_features. By Arnaud Joly.
• Fix wrong maximal number of features drawn (max_features) at each split for ensemble.
ExtraTreesClassifier and ensemble.ExtraTreesRegressor. Previously, only non constant
features in the split was counted as drawn. Now constant features are counted as drawn. Furthermore at least
one feature must be non constant in order to make a valid split. This bug fix will affect computational and gen-
eralization performance of extra trees in the presence of constant features. To get back previous generalization
performance, you should modify the value of max_features. By Arnaud Joly.
• Fix utils.compute_class_weight when class_weight=="auto". Previously it was broken for
input of non-integer dtype and the weighted array that was returned was wrong. By Manoj Kumar.
• Fix cross_validation.Bootstrap to return ValueError when n_train + n_test > n. By
Ronald Phlypo.
People
List of contributors for release 0.15 by number of commits.

• 312 Olivier Grisel
• 275 Lars Buitinck
• 221 Gael Varoquaux
• 148 Arnaud Joly
• 134 Johannes Schönberger
• 119 Gilles Louppe
• 113 Joel Nothman
• 111 Alexandre Gramfort
• 95 Jaques Grobler
• 89 Denis Engemann
• 83 Peter Prettenhofer
• 83 Alexander Fabisch
• 62 Mathieu Blondel
• 60 Eustache Diemert
• 60 Nelle Varoquaux
• 49 Michael Bommarito
• 45 Manoj-Kumar-S

• 28 Kyle Kastner
• 26 Andreas Mueller
• 22 Noel Dawe
• 21 Maheshakya Wijewardena
• 21 Brooke Osborn
• 21 Hamzeh Alsalhi
• 21 Jake VanderPlas
• 21 Philippe Gervais
• 19 Bala Subrahmanyam Varanasi
• 12 Ronald Phlypo
• 10 Mikhail Korobov
• 8 Thomas Unterthiner
• 8 Jeffrey Blackburne
• 8 eltermann
• 8 bwignall
• 7 Ankit Agrawal
• 7 CJ Carey
• 6 Daniel Nouri
• 6 Chen Liu
• 6 Michael Eickenberg
• 6 ugurthemaster
• 5 Aaron Schumacher
• 5 Baptiste Lagarde
• 5 Rajat Khanduja
• 5 Robert McGibbon
• 5 Sergio Pascual
• 4 Alexis Metaireau
• 4 Ignacio Rossi
• 4 Virgile Fritsch
• 4 Sebastian Säger
• 4 Ilambharathi Kanniah
• 4 sdenton4
• 4 Robert Layton
• 4 Alyssa
• 4 Amos Waterland
• 3 Andrew Tulloch

• 3 murad
• 3 Steven Maude
• 3 Karol Pysniak
• 3 Jacques Kvam
• 3 cgohlke
• 3 cjlin
• 3 Michael Becker
• 3 hamzeh
• 3 Eric Jacobsen
• 3 john collins
• 3 kaushik94
• 3 Erwin Marsi
• 2 csytracy
• 2 LK
• 2 Vlad Niculae
• 2 Laurent Direr
• 2 Erik Shilts
• 2 Raul Garreta
• 2 Yoshiki Vázquez Baeza
• 2 Yung Siang Liau
• 2 abhishek thakur
• 2 James Yu
• 2 Rohit Sivaprasad
• 2 Roland Szabo
• 2 amormachine
• 2 Alexis Mignon
• 2 Oscar Carlsson
• 2 Nantas Nardelli
• 2 jess010
• 2 kowalski87
• 2 Andrew Clegg
• 2 Federico Vaggi
• 2 Simon Frid
• 2 Félix-Antoine Fortin
• 1 Ralf Gommers
• 1 t-aft

• 1 Ronan Amicel
• 1 Rupesh Kumar Srivastava
• 1 Ryan Wang
• 1 Samuel Charron
• 1 Samuel St-Jean
• 1 Fabian Pedregosa
• 1 Skipper Seabold
• 1 Stefan Walk
• 1 Stefan van der Walt
• 1 Stephan Hoyer
• 1 Allen Riddell
• 1 Valentin Haenel
• 1 Vijay Ramesh
• 1 Will Myers
• 1 Yaroslav Halchenko
• 1 Yoni Ben-Meshulam
• 1 Yury V. Zaytsev
• 1 adrinjalali
• 1 ai8rahim
• 1 alemagnani
• 1 alex
• 1 benjamin wilson
• 1 chalmerlowe
• 1 dzikie drożdże
• 1 jamestwebber
• 1 matrixorz
• 1 popo
• 1 samuela
• 1 François Boulogne
• 1 Alexander Measure
• 1 Ethan White
• 1 Guilherme Trein
• 1 Hendrik Heuer
• 1 IvicaJovic
• 1 Jan Hendrik Metzen
• 1 Jean Michel Rouly

• 1 Eduardo Ariño de la Rubia

• 1 Jelle Zijlstra
• 1 Eddy L O Jansson
• 1 Denis
• 1 John
• 1 John Schmidt
• 1 Jorge Cañardo Alastuey
• 1 Joseph Perla
• 1 Joshua Vredevoogd
• 1 José Ricardo
• 1 Julien Miotte
• 1 Kemal Eren
• 1 Kenta Sato
• 1 David Cournapeau
• 1 Kyle Kelley
• 1 Daniele Medri
• 1 Laurent Luce
• 1 Laurent Pierron
• 1 Luis Pedro Coelho
• 1 DanielWeitzenfeld
• 1 Craig Thompson
• 1 Chyi-Kwei Yau
• 1 Matthew Brett
• 1 Matthias Feurer
• 1 Max Linke
• 1 Chris Filo Gorgolewski
• 1 Charles Earl
• 1 Michael Hanke
• 1 Michele Orrù
• 1 Bryan Lunt
• 1 Brian Kearns
• 1 Paul Butler
• 1 Paweł Mandera
• 1 Peter
• 1 Andrew Ash
• 1 Pietro Zambelli

• 1 staubda
1.15.11 Version 0.14
August 7, 2013
Changelog
• Missing values with sparse and dense matrices can be imputed with the transformer preprocessing.
Imputer by Nicolas Trésegnie.
• The core implementation of decisions trees has been rewritten from scratch, allowing for faster tree induction
and lower memory consumption in all tree-based estimators. By Gilles Louppe.
• Added ensemble.AdaBoostClassifier and ensemble.AdaBoostRegressor, by Noel Dawe and
Gilles Louppe. See the AdaBoost section of the user guide for details and examples.
• Added grid_search.RandomizedSearchCV and grid_search.ParameterSampler for random-
ized hyperparameter optimization. By Andreas Müller.
• Added biclustering algorithms (sklearn.cluster.bicluster.SpectralCoclustering and
sklearn.cluster.bicluster.SpectralBiclustering), data generation methods (sklearn.
datasets.make_biclusters and sklearn.datasets.make_checkerboard), and scoring met-
rics (sklearn.metrics.consensus_score). By Kemal Eren.
• Added Restricted Boltzmann Machines (neural_network.BernoulliRBM ). By Yann Dauphin.
• Python 3 support by Justin Vincent, Lars Buitinck, Subhodeep Moitra and Olivier Grisel. All tests now pass
under Python 3.3.
• Ability to pass one penalty (alpha value) per target in linear_model.Ridge, by @eickenberg and Mathieu
Blondel.
• Fixed sklearn.linear_model.stochastic_gradient.py L2 regularization issue (minor practical
significance). By Norbert Crombach and Mathieu Blondel .
• Added an interactive version of Andreas Müller’s Machine Learning Cheat Sheet (for scikit-learn) to the docu-
mentation. See Choosing the right estimator. By Jaques Grobler.
• grid_search.GridSearchCV and cross_validation.cross_val_score now support the use
of advanced scoring function such as area under the ROC curve and f-beta scores. See The scoring parameter:
defining model evaluation rules for details. By Andreas Müller and Lars Buitinck. Passing a function from
sklearn.metrics as score_func is deprecated.
• Multi-label classification output is now supported by metrics.accuracy_score,
metrics.zero_one_loss, metrics.f1_score, metrics.fbeta_score, metrics.
classification_report, metrics.precision_score and metrics.recall_score by
Arnaud Joly.
• Two new metrics metrics.hamming_loss and metrics.jaccard_similarity_score are added
with multi-label support by Arnaud Joly.
• Speed and memory usage improvements in feature_extraction.text.CountVectorizer and
feature_extraction.text.TfidfVectorizer, by Jochen Wersdörfer and Roman Sinayev.
• The min_df parameter in feature_extraction.text.CountVectorizer and
feature_extraction.text.TfidfVectorizer, which used to be 2, has been reset to 1 to
avoid unpleasant surprises (empty vocabularies) for novice users who try it out on tiny document collections. A
value of at least 2 is still recommended for practical use.

• svm.LinearSVC, linear_model.SGDClassifier and linear_model.SGDRegressor now

have a sparsify method that converts their coef_ into a sparse matrix, meaning stored models trained
using these estimators can be made much more compact.
• linear_model.SGDClassifier now produces multiclass probability estimates when trained under log
loss or modified Huber loss.
• Hyperlinks to documentation in example code on the website by Martin Luessi.
• Fixed bug in preprocessing.MinMaxScaler causing incorrect scaling of the features for non-default
feature_range settings. By Andreas Müller.
• max_features in tree.DecisionTreeClassifier, tree.DecisionTreeRegressor and all
derived ensemble estimators now supports percentage values. By Gilles Louppe.
• Performance improvements in isotonic.IsotonicRegression by Nelle Varoquaux.
• metrics.accuracy_score has an option normalize to return the fraction or the number of correctly clas-
sified sample by Arnaud Joly.
• Added metrics.log_loss that computes log loss, aka cross-entropy loss. By Jochen Wersdörfer and Lars
Buitinck.
• A bug that caused ensemble.AdaBoostClassifier’s to output incorrect probabilities has been fixed.
• Feature selectors now share a mixin providing consistent transform, inverse_transform and
get_support methods. By Joel Nothman.
• A fitted grid_search.GridSearchCV or grid_search.RandomizedSearchCV can now generally
be pickled. By Joel Nothman.
• Refactored and vectorized implementation of metrics.roc_curve and metrics.
precision_recall_curve. By Joel Nothman.
• The new estimator sklearn.decomposition.TruncatedSVD performs dimensionality reduction using
SVD on sparse matrices, and can be used for latent semantic analysis (LSA). By Lars Buitinck.
• Added self-contained example of out-of-core learning on text data Out-of-core classification of text documents.
By Eustache Diemert.
• The default number of components for sklearn.decomposition.RandomizedPCA is now correctly
documented to be n_features. This was the default behavior, so programs using it will continue to work as
they did.
• sklearn.cluster.KMeans now fits several orders of magnitude faster on sparse data (the speedup depends
on the sparsity). By Lars Buitinck.
• Reduce memory footprint of FastICA by Denis Engemann and Alexandre Gramfort.
• Verbose output in sklearn.ensemble.gradient_boosting now uses a column format and prints
progress in decreasing frequency. It also shows the remaining time. By Peter Prettenhofer.
• sklearn.ensemble.gradient_boosting provides out-of-bag improvement oob_improvement_
rather than the OOB score for model selection. An example that shows how to use OOB estimates to select the
number of trees was added. By Peter Prettenhofer.
• Most metrics now support string labels for multiclass classification by Arnaud Joly and Lars Buitinck.
• New OrthogonalMatchingPursuitCV class by Alexandre Gramfort and Vlad Niculae.
• Fixed a bug in sklearn.covariance.GraphLassoCV : the ‘alphas’ parameter now works as expected
when given a list of values. By Philippe Gervais.

• Fixed an important bug in sklearn.covariance.GraphLassoCV that prevented all folds provided by

a CV object to be used (only the first 3 were used). When providing a CV object, execution time may thus
increase significantly compared to the previous version (bug results are correct now). By Philippe Gervais.
• cross_validation.cross_val_score and the grid_search module is now tested with multi-
output data by Arnaud Joly.
• datasets.make_multilabel_classification can now return the output in label indicator multil-
abel format by Arnaud Joly.
• K-nearest neighbors, neighbors.KNeighborsRegressor and neighbors.
RadiusNeighborsRegressor, and radius neighbors, neighbors.RadiusNeighborsRegressor
and neighbors.RadiusNeighborsClassifier support multioutput data by Arnaud Joly.
• Random state in LibSVM-based estimators (svm.SVC, NuSVC, OneClassSVM, svm.SVR, svm.NuSVR)
can now be controlled. This is useful to ensure consistency in the probability estimates for the classifiers trained
with probability=True. By Vlad Niculae.
• Out-of-core learning support for discrete naive Bayes classifiers sklearn.naive_bayes.
MultinomialNB and sklearn.naive_bayes.BernoulliNB by adding the partial_fit
method by Olivier Grisel.
• New website design and navigation by Gilles Louppe, Nelle Varoquaux, Vincent Michel and Andreas Müller.
• Improved documentation on multi-class, multi-label and multi-output classification by Yannick Schwartz and
Arnaud Joly.
• Better input and error handling in the metrics module by Arnaud Joly and Joel Nothman.
• Speed optimization of the hmm module by Mikhail Korobov
• Significant speed improvements for sklearn.cluster.DBSCAN by cleverless
API changes summary
• The auc_score was renamed roc_auc_score.

• Testing scikit-learn with sklearn.test() is deprecated. Use nosetests sklearn from the command
line.
• Feature importances in tree.DecisionTreeClassifier, tree.DecisionTreeRegressor and all
derived ensemble estimators are now computed on the fly when accessing the feature_importances_
attribute. Setting compute_importances=True is no longer required. By Gilles Louppe.
• linear_model.lasso_path and linear_model.enet_path can return its results in the same format
as that of linear_model.lars_path. This is done by setting the return_models parameter to False.
By Jaques Grobler and Alexandre Gramfort
• grid_search.IterGrid was renamed to grid_search.ParameterGrid.
• Fixed bug in KFold causing imperfect class balance in some cases. By Alexandre Gramfort and Tadej Janež.
• sklearn.neighbors.BallTree has been refactored, and a sklearn.neighbors.KDTree has been
added which shares the same interface. The Ball Tree now works with a wide variety of distance metrics.
Both classes have many new methods, including single-tree and dual-tree queries, breadth-first and depth-first
searching, and more advanced queries such as kernel density estimation and 2-point correlation functions. By
Jake Vanderplas
• Support for scipy.spatial.cKDTree within neighbors queries has been removed, and the functionality replaced
with the new KDTree class.

• sklearn.neighbors.KernelDensity has been added, which performs efficient kernel density estima-
tion with a variety of kernels.
• sklearn.decomposition.KernelPCA now always returns output with n_components components,
unless the new parameter remove_zero_eig is set to True. This new behavior is consistent with the way
kernel PCA was always documented; previously, the removal of components with zero eigenvalues was tacitly
performed on all data.
• gcv_mode="auto" no longer tries to perform SVD on a densified sparse matrix in sklearn.
linear_model.RidgeCV .
• Sparse matrix support in sklearn.decomposition.RandomizedPCA is now deprecated in favor of the
new TruncatedSVD.
• cross_validation.KFold and cross_validation.StratifiedKFold now enforce n_folds >=
2 otherwise a ValueError is raised. By Olivier Grisel.
• datasets.load_files’s charset and charset_errors parameters were renamed encoding and
decode_errors.
• Attribute oob_score_ in sklearn.ensemble.GradientBoostingRegressor and
sklearn.ensemble.GradientBoostingClassifier is deprecated and has been replaced by
oob_improvement_ .
• Attributes in OrthogonalMatchingPursuit have been deprecated (copy_X, Gram, . . . ) and precompute_gram
renamed precompute for consistency. See #2224.
• sklearn.preprocessing.StandardScaler now converts integer input to float, and raises a warning.
Previously it rounded for dense integer input.
• sklearn.multiclass.OneVsRestClassifier now has a decision_function method. This
will return the distance of each sample from the decision boundary for each class, as long as the underlying
estimators implement the decision_function method. By Kyle Kastner.
• Better input validation, warning on unexpected shapes for y.
People

• 187 Andreas Mueller
• 124 Arnaud Joly
• 102 Noel Dawe
• 99 Kemal Eren
• 79 Joel Nothman
• 75 Jake VanderPlas
• 71 Vlad Niculae

• 38 Nicolas Trésegnie
• 35 eustache
• 27 Denis Engemann
• 25 Yann N. Dauphin
• 19 Justin Vincent
• 17 Robert Layton
• 15 Doug Coleman
• 13 Robert Marchman
• 11 Philippe Gervais
• 10 Jim Holmström
• 10 Tadej Janež
• 10 syhw
• 9 Steven De Gryze
• 8 sergeyf
• 7 Ben Root
• 7 Hrishikesh Huilgolkar
• 6 Kyle Kastner
• 6 Martin Luessi
• 6 Rob Speer
• 5 Federico Vaggi
• 5 Raul Garreta
• 5 Rob Zinkov
• 4 Ken Geis
• 3 A. Flaxman
• 3 Denton Cockburn
• 3 Dougal Sutherland
• 3 Ian Ozsvald
• 3 Johannes Schönberger
• 3 Robert McGibbon
• 3 Roman Sinayev

• 3 Szabo Roland
• 2 Diego Molla
• 2 Imran Haque
• 2 Jochen Wersdörfer
• 2 Sergey Karayev
• 2 Yannick Schwartz
• 2 jamestwebber
• 1 Abhijeet Kolhe
• 1 Alexander Fabisch
• 1 Bastiaan van den Berg
• 1 Benjamin Peterson
• 1 Daniel Velkov
• 1 Fazlul Shahriar
• 1 Felix Brockherde
• 1 Harikrishnan S
• 1 Jack Hale
• 1 JakeMick
• 1 James McDermott
• 1 John Benediktsson
• 1 John Zwinck
• 1 Joshua Vredevoogd
• 1 Justin Pati
• 1 Kevin Hughes
• 1 Kyle Kelley
• 1 Matthias Ekman
• 1 Miroslav Shubernetskiy
• 1 Naoki Orii
• 1 Norbert Crombach
• 1 Rafael Cunha de Almeida
• 1 Rolando Espinoza La fuente
• 1 Seamus Abshere
• 1 Sergey Feldman
• 1 Sergio Medina
• 1 Stefano Lattarini
• 1 Steve Koch

• 1 Sturla Molden
• 1 Thomas Jarosch
1.15.12 Version 0.13.1
February 23, 2013

The 0.13.1 release only fixes some bugs and does not add any new functionality.
Changelog
• Fixed a testing error caused by the function cross_validation.train_test_split being interpreted

as a test by Yaroslav Halchenko.
• Fixed a bug in the reassignment of small clusters in the cluster.MiniBatchKMeans by Gael Varoquaux.
• Fixed default value of gamma in decomposition.KernelPCA by Lars Buitinck.
• Updated joblib to 0.7.0d by Gael Varoquaux.
• Fixed scaling of the deviance in ensemble.GradientBoostingClassifier by Peter Prettenhofer.
• Better tie-breaking in multiclass.OneVsOneClassifier by Andreas Müller.
• Other small improvements to tests and documentation.
People
List of contributors for release 0.13.1 by number of commits.

• 12 Andreas Müller
• 5 Robert Marchman
• 2 Hrishikesh Huilgolkar
• 1 Bastiaan van den Berg
• 1 Diego Molla
• 1 Gilles Louppe
• 1 Rafael Cunha de Almeida
• 1 Rolando Espinoza La fuente
• 1 Vlad Niculae

1.15.13 Version 0.13
January 21, 2013
New Estimator Classes
• dummy.DummyClassifier and dummy.DummyRegressor, two data-independent predictors by Mathieu

Blondel. Useful to sanity-check your estimators. See Dummy estimators in the user guide. Multioutput support
added by Arnaud Joly.
• decomposition.FactorAnalysis, a transformer implementing the classical factor analysis, by Chris-
tian Osendorfer and Alexandre Gramfort. See Factor Analysis in the user guide.
• feature_extraction.FeatureHasher, a transformer implementing the “hashing trick” for fast,
low-memory feature extraction from string fields by Lars Buitinck and feature_extraction.text.
HashingVectorizer for text documents by Olivier Grisel See Feature hashing and Vectorizing a large
text corpus with the hashing trick for the documentation and sample usage.
• pipeline.FeatureUnion, a transformer that concatenates results of several other transformers by Andreas
Müller. See FeatureUnion: composite feature spaces in the user guide.
• random_projection.GaussianRandomProjection, random_projection.
SparseRandomProjection and the function random_projection.
johnson_lindenstrauss_min_dim. The first two are transformers implementing Gaussian and
sparse random projection matrix by Olivier Grisel and Arnaud Joly. See Random Projection in the user guide.
• kernel_approximation.Nystroem, a transformer for approximating arbitrary kernels by Andreas
Müller. See Nystroem Method for Kernel Approximation in the user guide.
• preprocessing.OneHotEncoder, a transformer that computes binary encodings of categorical features
by Andreas Müller. See Encoding categorical features in the user guide.
• linear_model.PassiveAggressiveClassifier and linear_model.
PassiveAggressiveRegressor, predictors implementing an efficient stochastic optimization for
linear models by Rob Zinkov and Mathieu Blondel. See Passive Aggressive Algorithms in the user guide.
• ensemble.RandomTreesEmbedding, a transformer for creating high-dimensional sparse representations
using ensembles of totally random trees by Andreas Müller. See Totally Random Trees Embedding in the user
guide.
• manifold.SpectralEmbedding and function manifold.spectral_embedding, implementing
the “laplacian eigenmaps” transformation for non-linear dimensionality reduction by Wei Li. See Spectral
Embedding in the user guide.
• isotonic.IsotonicRegression by Fabian Pedregosa, Alexandre Gramfort and Nelle Varoquaux,
Changelog
• metrics.zero_one_loss (formerly metrics.zero_one) now has option for normalized output that
reports the fraction of misclassifications, rather than the raw number of misclassifications. By Kyle Beauchamp.
• tree.DecisionTreeClassifier and all derived ensemble models now support sample weighting, by
Noel Dawe and Gilles Louppe.
• Speedup improvement when using bootstrap samples in forests of randomized trees, by Peter Prettenhofer and
Gilles Louppe.
• Partial dependence plots for Gradient Tree Boosting in ensemble.partial_dependence.
partial_dependence by Peter Prettenhofer. See Partial Dependence Plots for an example.

• The table of contents on the website has now been made expandable by Jaques Grobler.
• feature_selection.SelectPercentile now breaks ties deterministically instead of returning all
equally ranked features.
• feature_selection.SelectKBest and feature_selection.SelectPercentile are more
numerically stable since they use scores, rather than p-values, to rank results. This means that they might
sometimes select different features than they did previously.
• Ridge regression and ridge classification fitting with sparse_cg solver no longer has quadratic memory com-
plexity, by Lars Buitinck and Fabian Pedregosa.
• Ridge regression and ridge classification now support a new fast solver called lsqr, by Mathieu Blondel.
• Speed up of metrics.precision_recall_curve by Conrad Lee.
• Added support for reading/writing svmlight files with pairwise preference attribute (qid in svmlight file format)
in datasets.dump_svmlight_file and datasets.load_svmlight_file by Fabian Pedregosa.
• Faster and more robust metrics.confusion_matrix and Clustering performance evaluation by Wei Li.
• cross_validation.cross_val_score now works with precomputed kernels and affinity matrices, by
Andreas Müller.
• LARS algorithm made more numerically stable with heuristics to drop regressors too correlated as well as to
stop the path when numerical noise becomes predominant, by Gael Varoquaux.
• Faster implementation of metrics.precision_recall_curve by Conrad Lee.
• New kernel metrics.chi2_kernel by Andreas Müller, often used in computer vision applications.
• Fix of longstanding bug in naive_bayes.BernoulliNB fixed by Shaun Jackman.
• Implemented predict_proba in multiclass.OneVsRestClassifier, by Andrew Winterman.
• Improve consistency in gradient boosting: estimators ensemble.GradientBoostingRegressor and
ensemble.GradientBoostingClassifier use the estimator tree.DecisionTreeRegressor
instead of the tree._tree.Tree data structure by Arnaud Joly.
• Fixed a floating point exception in the decision trees module, by Seberg.
• Fix metrics.roc_curve fails when y_true has only one class by Wei Li.
• Add the metrics.mean_absolute_error function which computes the mean absolute error. The
metrics.mean_squared_error, metrics.mean_absolute_error and metrics.r2_score
metrics support multioutput by Arnaud Joly.
• Fixed class_weight support in svm.LinearSVC and linear_model.LogisticRegression by
Andreas Müller. The meaning of class_weight was reversed as erroneously higher weight meant less
positives of a given class in earlier releases.
• Improve narrative documentation and consistency in sklearn.metrics for regression and classification
metrics by Arnaud Joly.
• Fixed a bug in sklearn.svm.SVC when using csr-matrices with unsorted indices by Xinfan Meng and An-
dreas Müller.
• MiniBatchKMeans: Add random reassignment of cluster centers with little observations attached to them,
by Gael Varoquaux.
API changes summary
• Renamed all occurrences of n_atoms to n_components for consistency.

This applies to decomposition.DictionaryLearning, decomposition.

MiniBatchDictionaryLearning, decomposition.dict_learning, decomposition.

dict_learning_online.
• Renamed all occurrences of max_iters to max_iter for consistency. This applies to
semi_supervised.LabelPropagation and semi_supervised.label_propagation.
LabelSpreading.
• Renamed all occurrences of learn_rate to learning_rate for consistency in ensemble.
BaseGradientBoosting and ensemble.GradientBoostingRegressor.
• The module sklearn.linear_model.sparse is gone. Sparse matrix support was already integrated into
the “regular” linear models.
• sklearn.metrics.mean_square_error, which incorrectly returned the accumulated error, was re-
moved. Use mean_squared_error instead.
• Passing class_weight parameters to fit methods is no longer supported. Pass them to estimator construc-
tors instead.
• GMMs no longer have decode and rvs methods. Use the score, predict or sample methods instead.
• The solver fit option in Ridge regression and classification is now deprecated and will be removed in v0.14.
Use the constructor option instead.
• feature_extraction.text.DictVectorizer now returns sparse matrices in the CSR format, in-
stead of COO.
• Renamed k in cross_validation.KFold and cross_validation.StratifiedKFold to
n_folds, renamed n_bootstraps to n_iter in cross_validation.Bootstrap.
• Renamed all occurrences of n_iterations to n_iter for consistency. This applies to
cross_validation.ShuffleSplit, cross_validation.StratifiedShuffleSplit,
utils.randomized_range_finder and utils.randomized_svd.
• Replaced rho in linear_model.ElasticNet and linear_model.SGDClassifier by
l1_ratio. The rho parameter had different meanings; l1_ratio was introduced to avoid confu-
sion. It has the same meaning as previously rho in linear_model.ElasticNet and (1-rho) in
linear_model.SGDClassifier.
• linear_model.LassoLars and linear_model.Lars now store a list of paths in the case of multiple
targets, rather than an array of paths.
• The attribute gmm of hmm.GMMHMM was renamed to gmm_ to adhere more strictly with the API.
• cluster.spectral_embedding was moved to manifold.spectral_embedding.
• Renamed eig_tol in manifold.spectral_embedding, cluster.SpectralClustering to
eigen_tol, renamed mode to eigen_solver.
• Renamed mode in manifold.spectral_embedding and cluster.SpectralClustering to
eigen_solver.
• classes_ and n_classes_ attributes of tree.DecisionTreeClassifier and all derived ensemble
models are now flat in case of single output problems and nested in case of multi-output problems.
• The estimators_ attribute of ensemble.gradient_boosting.GradientBoostingRegressor
and ensemble.gradient_boosting.GradientBoostingClassifier is now an array of
:class:’tree.DecisionTreeRegressor’.
• Renamed chunk_size to batch_size in decomposition.MiniBatchDictionaryLearning
and decomposition.MiniBatchSparsePCA for consistency.
• svm.SVC and svm.NuSVC now provide a classes_ attribute and support arbitrary dtypes for labels y.
Also, the dtype returned by predict now reflects the dtype of y during fit (used to be np.float).

• Changed default test_size in cross_validation.train_test_split to None, added pos-

sibility to infer test_size from train_size in cross_validation.ShuffleSplit and
cross_validation.StratifiedShuffleSplit.
• Renamed function sklearn.metrics.zero_one to sklearn.metrics.zero_one_loss. Be
aware that the default behavior in sklearn.metrics.zero_one_loss is different from sklearn.
metrics.zero_one: normalize=False is changed to normalize=True.
• Renamed function metrics.zero_one_score to metrics.accuracy_score.
• datasets.make_circles now has the same number of inner and outer points.
• In the Naive Bayes classifiers, the class_prior parameter was moved from fit to __init__.
People

• 143 Arnaud Joly
• 106 Wei Li
• 65 Vlad Niculae
• 30 Rob Zinkov
• 19 Aymeric Masurelle
• 18 Andrew Winterman
• 16 Christian Osendorfer
• 14 Daniel Nouri
• 13 syhw
• 12 Satrajit Ghosh
• 10 Corey Lynch
• 10 Kyle Beauchamp
• 9 Brian Cheung

• 9 Immanuel Bayer
• 9 mr.Shu
• 8 Conrad Lee
• 8 James Bergstra
• 7 Tadej Janež
• 6 Brian Cajes
• 6 Jake Vanderplas
• 6 Michael
• 6 Noel Dawe
• 6 Tiago Nunes
• 6 cow
• 5 Anze
• 5 Shiqiao Du
• 4 Christian Jauvin
• 4 Jacques Kvam
• 4 Richard T. Guy
• 4 Robert Layton
• 3 Alexandre Abraham
• 3 Doug Coleman
• 3 Scott Dickerson
• 2 ApproximateIdentity
• 2 Mark Veronda
• 2 Matti Lyra
• 2 Xinfan Meng
• 1 Alejandro Weinstein
• 1 Alexandre Passos
• 1 Christoph Deil
• 1 Eugene Nizhibitsky
• 1 Kenneth C. Arnold
• 1 Luis Pedro Coelho
• 1 Miroslav Batchkarov
• 1 Pavel
• 1 Sebastian Berg
• 1 Shaun Jackman

• 1 Subhodeep Moitra
• 1 bob
• 1 dengemann
• 1 emanuele
• 1 x006
1.15.14 Version 0.12.1
October 8, 2012
The 0.12.1 release is a bug-fix release with no additional features, but is instead a set of bug fixes
Changelog
• Improved numerical stability in spectral embedding by Gael Varoquaux

• Doctest under windows 64bit by Gael Varoquaux
• Documentation fixes for elastic net by Andreas Müller and Alexandre Gramfort
• Proper behavior with fortran-ordered NumPy arrays by Gael Varoquaux
• Make GridSearchCV work with non-CSR sparse matrix by Lars Buitinck
• Fix parallel computing in MDS by Gael Varoquaux
• Fix Unicode support in count vectorizer by Andreas Müller
• Fix MinCovDet breaking with X.shape = (3, 1) by Virgile Fritsch
• Fix clone of SGD objects by Peter Prettenhofer
• Stabilize GMM by Virgile Fritsch
People
• 5 Lars Buitinck
• 1 Gilles Louppe
1.15.15 Version 0.12
September 4, 2012

Changelog
• Various speed improvements of the decision trees module, by Gilles Louppe.

• ensemble.GradientBoostingRegressor and ensemble.GradientBoostingClassifier
now support feature subsampling via the max_features argument, by Peter Prettenhofer.
• Added Huber and Quantile loss functions to ensemble.GradientBoostingRegressor, by Peter Pret-
tenhofer.
• Decision trees and forests of randomized trees now support multi-output classification and regression problems,
by Gilles Louppe.
• Added preprocessing.LabelEncoder, a simple utility class to normalize labels or transform non-
numerical labels, by Mathieu Blondel.
• Added the epsilon-insensitive loss and the ability to make probabilistic predictions with the modified huber loss
in Stochastic Gradient Descent, by Mathieu Blondel.
• Added Multi-dimensional Scaling (MDS), by Nelle Varoquaux.
• SVMlight file format loader now detects compressed (gzip/bzip2) files and decompresses them on the fly, by
Lars Buitinck.
• SVMlight file format serializer now preserves double precision floating point values, by Olivier Grisel.
• A common testing framework for all estimators was added, by Andreas Müller.
• Understandable error messages for estimators that do not accept sparse input by Gael Varoquaux
• Speedups in hierarchical clustering by Gael Varoquaux. In particular building the tree now supports early
stopping. This is useful when the number of clusters is not small compared to the number of samples.
• Add MultiTaskLasso and MultiTaskElasticNet for joint feature selection, by Alexandre Gramfort.
• Added metrics.auc_score and metrics.average_precision_score convenience functions by
Andreas Müller.
• Improved sparse matrix support in the Feature selection module by Andreas Müller.
• New word boundaries-aware character n-gram analyzer for the Text feature extraction module by @kernc.
• Fixed bug in spectral clustering that led to single point clusters by Andreas Müller.
• In feature_extraction.text.CountVectorizer, added an option to ignore infrequent words,
min_df by Andreas Müller.
• Add support for multiple targets in some linear models (ElasticNet, Lasso and OrthogonalMatchingPursuit) by
Vlad Niculae and Alexandre Gramfort.
• Fixes in decomposition.ProbabilisticPCA score function by Wei Li.
• Fixed feature importance computation in Gradient Tree Boosting.
API changes summary
• The old scikits.learn package has disappeared; all code should import from sklearn instead, which
was introduced in 0.9.
• In metrics.roc_curve, the thresholds array is now returned with it’s order reversed, in order to keep
it consistent with the order of the returned fpr and tpr.
• In hmm objects, like hmm.GaussianHMM, hmm.MultinomialHMM, etc., all parameters must be passed to
the object when initialising it and not through fit. Now fit will only accept the data as an input parameter.

• For all SVM classes, a faulty behavior of gamma was fixed. Previously, the default gamma value was only
computed the first time fit was called and then stored. It is now recalculated on every call to fit.
• All Base classes are now abstract meta classes so that they can not be instantiated.
• cluster.ward_tree now also returns the parent array. This is necessary for early-stopping in which case
the tree is not completely built.
• In feature_extraction.text.CountVectorizer the parameters min_n and max_n were joined to
the parameter n_gram_range to enable grid-searching both at once.
• In feature_extraction.text.CountVectorizer, words that appear only in one document are now
ignored by default. To reproduce the previous behavior, set min_df=1.
• Fixed API inconsistency: linear_model.SGDClassifier.predict_proba now returns 2d array
when fit on two classes.
• Fixed API inconsistency: discriminant_analysis.QuadraticDiscriminantAnalysis.
decision_function and discriminant_analysis.LinearDiscriminantAnalysis.
decision_function now return 1d arrays when fit on two classes.
• Grid of alphas used for fitting linear_model.LassoCV and linear_model.ElasticNetCV is now
stored in the attribute alphas_ rather than overriding the init parameter alphas.
• Linear models when alpha is estimated by cross-validation store the estimated value in the alpha_ attribute
rather than just alpha or best_alpha.
• ensemble.GradientBoostingClassifier now supports ensemble.
GradientBoostingClassifier.staged_predict_proba, and ensemble.
GradientBoostingClassifier.staged_predict.
• svm.sparse.SVC and other sparse SVM classes are now deprecated. The all classes in the Support Vector
Machines module now automatically select the sparse or dense representation base on the input.
• All clustering algorithms now interpret the array X given to fit as input data, in particular cluster.
SpectralClustering and cluster.AffinityPropagation which previously expected affinity ma-
trices.
• For clustering algorithms that take the desired number of clusters as a parameter, this parameter is now called
n_clusters.
People

• 52 Vlad Niculae
• 30 Alexis Mignon

• 30 Immanuel Bayer
• 16 Subhodeep Moitra
• 12 @kernc
• 9 Daniel Duckworth
• 9 Robert Layton
• 7 Marko Burjek
• 5 Nicolas Pinto
• 4 Alexandre Abraham
• 3 Brian Holt
• 3 Edouard Duchesnay
• 3 Florian Hoenig
• 3 flyingimmidev
• 2 Francois Savard
• 2 Hannes Schulz
• 2 Peter Welinder
• 2 Wei Li
• 1 Alex Companioni
• 1 Brandyn A. White
• 1 Bussonnier Matthias
• 1 Charles-Pierre Astolfi
• 1 Dan O’Huiginn
• 1 David Cournapeau
• 1 Keith Goodman
• 1 Ludwig Schwardt
• 1 Olivier Hervieu
• 1 Sergio Medina
• 1 Shiqiao Du
• 1 Tim Sheerman-Chase
• 1 buguen

1.15.16 Version 0.11
May 7, 2012
Changelog
Highlights
• Gradient boosted regression trees (Gradient Tree Boosting) for classification and regression by Peter Pretten-
hofer and Scott White .
• Simple dict-based feature loader with support for categorical variables (feature_extraction.
DictVectorizer) by Lars Buitinck.
• Added Matthews correlation coefficient (metrics.matthews_corrcoef) and added macro and micro av-
erage options to metrics.precision_score, metrics.recall_score and metrics.f1_score
by Satrajit Ghosh.
• Out of Bag Estimates of generalization error for Ensemble methods by Andreas Müller.
• Randomized sparse linear models for feature selection, by Alexandre Gramfort and Gael Varoquaux
• Label Propagation for semi-supervised learning, by Clay Woolam. Note the semi-supervised API is still work
in progress, and may change.
• Added BIC/AIC model selection to classical Gaussian mixture models and unified the API with the remainder
of scikit-learn, by Bertrand Thirion
• Added sklearn.cross_validation.StratifiedShuffleSplit, which is a sklearn.
cross_validation.ShuffleSplit with balanced splits, by Yannick Schwartz.
• sklearn.neighbors.NearestCentroid classifier added, along with a shrink_threshold param-
eter, which implements shrunken centroid classification, by Robert Layton.
Other changes
• Merged dense and sparse implementations of Stochastic Gradient Descent module and exposed utility extension
types for sequential datasets seq_dataset and weight vectors weight_vector by Peter Prettenhofer.
• Added partial_fit (support for online/minibatch learning) and warm_start to the Stochastic Gradient De-
scent module by Mathieu Blondel.
• Dense and sparse implementations of Support Vector Machines classes and linear_model.
LogisticRegression merged by Lars Buitinck.
• Regressors can now be used as base estimator in the Multiclass and multilabel algorithms module by Mathieu
Blondel.
• Added n_jobs option to metrics.pairwise.pairwise_distances and metrics.pairwise.
pairwise_kernels for parallel computation, by Mathieu Blondel.
• K-means can now be run in parallel, using the n_jobs argument to either K-means or KMeans, by Robert
Layton.
• Improved Cross-validation: evaluating estimator performance and Tuning the hyper-parameters of an estima-
tor documentation and introduced the new cross_validation.train_test_split helper function by
Olivier Grisel

• svm.SVC members coef_ and intercept_ changed sign for consistency with decision_function;
for kernel==linear, coef_ was fixed in the one-vs-one case, by Andreas Müller.
• Performance improvements to efficient leave-one-out cross-validated Ridge regression, esp. for the
n_samples > n_features case, in linear_model.RidgeCV , by Reuben Fletcher-Costin.
• Refactoring and simplification of the Text feature extraction API and fixed a bug that caused possible negative
IDF, by Olivier Grisel.
• Beam pruning option in _BaseHMM module has been removed since it is difficult to Cythonize. If you are
interested in contributing a Cython version, you can use the python version in the git history as a reference.
• Classes in Nearest Neighbors now support arbitrary Minkowski metric for nearest neighbors searches. The
metric can be specified by argument p.
API changes summary
• covariance.EllipticEnvelop is now deprecated - Please use covariance.EllipticEnvelope

instead.
• NeighborsClassifier and NeighborsRegressor are gone in the module Nearest Neighbors. Use
the classes KNeighborsClassifier, RadiusNeighborsClassifier, KNeighborsRegressor
and/or RadiusNeighborsRegressor instead.
• Sparse classes in the Stochastic Gradient Descent module are now deprecated.
• In mixture.GMM, mixture.DPGMM and mixture.VBGMM, parameters must be passed to an object when
initialising it and not through fit. Now fit will only accept the data as an input parameter.
• methods rvs and decode in GMM module are now deprecated. sample and score or predict should be
used instead.
• attribute _scores and _pvalues in univariate feature selection objects are now deprecated. scores_ or
pvalues_ should be used instead.
• In LogisticRegression, LinearSVC, SVC and NuSVC, the class_weight parameter is now an ini-
tialization parameter, not a parameter to fit. This makes grid searches over this parameter possible.
• LFW data is now always shape (n_samples, n_features) to be consistent with the Olivetti faces
dataset. Use images and pairs attribute to access the natural images shapes instead.
• In svm.LinearSVC, the meaning of the multi_class parameter changed. Options now are 'ovr' and
'crammer_singer', with 'ovr' being the default. This does not change the default behavior but hopefully
is less confusing.
• Class feature_selection.text.Vectorizer is deprecated and replaced by
feature_selection.text.TfidfVectorizer.
• The preprocessor / analyzer nested structure for text feature extraction has been removed. All those features are
now directly passed as flat constructor arguments to feature_selection.text.TfidfVectorizer
and feature_selection.text.CountVectorizer, in particular the following parameters are now
used:
• analyzer can be 'word' or 'char' to switch the default analysis scheme, or use a specific python callable
(as previously).
• tokenizer and preprocessor have been introduced to make it still possible to customize those steps with
the new API.
• input explicitly control how to interpret the sequence passed to fit and predict: filenames, file objects or
direct (byte or Unicode) strings.

• charset decoding is explicit and strict by default.

• the vocabulary, fitted or not is now stored in the vocabulary_ attribute to be consistent with the project
conventions.
• Class feature_selection.text.TfidfVectorizer now derives directly from
feature_selection.text.CountVectorizer to make grid search trivial.
• methods rvs in _BaseHMM module are now deprecated. sample should be used instead.
• Beam pruning option in _BaseHMM module is removed since it is difficult to be Cythonized. If you are inter-
ested, you can look in the history codes by git.
• The SVMlight format loader now supports files with both zero-based and one-based column indices, since both
occur “in the wild”.
• Arguments in class ShuffleSplit are now consistent with StratifiedShuffleSplit. Arguments
test_fraction and train_fraction are deprecated and renamed to test_size and train_size
and can accept both float and int.
• Arguments in class Bootstrap are now consistent with StratifiedShuffleSplit. Arguments
n_test and n_train are deprecated and renamed to test_size and train_size and can accept both
float and int.
• Argument p added to classes in Nearest Neighbors to specify an arbitrary Minkowski metric for nearest neigh-
bors searches.
People

• 103 Clay Woolam
• 50 Bertrand Thirion
• 28 flyingimmidev
• 26 Shiqiao Du
• 17 David Marek
• 14 Vlad Niculae

• 9 fcostin
• 7 Nick Wilson
• 5 Adrien Gaidon
• 5 Nicolas Pinto
• 4 David Warde-Farley
• 5 Emmanuelle Gouillart
• 3 Joonas Sillanpää
• 3 Paolo Losi
• 2 Charles McCarthy
• 2 Roy Hyunjin Han
• 2 Scott White
• 2 ibayer
• 1 Brandyn White
• 1 Carlos Scheidegger
• 1 Claire Revillet
• 1 Conrad Lee
• 1 Meng Xinfan
• 1 Rob Zinkov
• 1 Shiqiao
• 1 Udi Weinsberg
• 1 Xinfan Meng
• 1 jansoe
• 1 Leon Palafox
1.15.17 Version 0.10
January 11, 2012

Changelog
• Python 2.5 compatibility was dropped; the minimum Python version needed to use scikit-learn is now 2.6.
• Sparse inverse covariance estimation using the graph Lasso, with associated cross-validated estimator, by Gael
Varoquaux
• New Tree module by Brian Holt, Peter Prettenhofer, Satrajit Ghosh and Gilles Louppe. The module comes with
complete documentation and examples.
• Fixed a bug in the RFE module by Gilles Louppe (issue #378).
• Fixed a memory leak in Support Vector Machines module by Brian Holt (issue #367).
• Faster tests by Fabian Pedregosa and others.
• Silhouette Coefficient cluster analysis evaluation metric added as sklearn.metrics.
silhouette_score by Robert Layton.
• Fixed a bug in K-means in the handling of the n_init parameter: the clustering algorithm used to be run
n_init times but the last solution was retained instead of the best solution by Olivier Grisel.
• Minor refactoring in Stochastic Gradient Descent module; consolidated dense and sparse predict methods; En-
hanced test time performance by converting model parameters to fortran-style arrays after fitting (only multi-
class).
• Adjusted Mutual Information metric added as sklearn.metrics.adjusted_mutual_info_score
by Robert Layton.
• Models like SVC/SVR/LinearSVC/LogisticRegression from libsvm/liblinear now support scaling of C regular-
ization parameter by the number of samples by Alexandre Gramfort.
• New Ensemble Methods module by Gilles Louppe and Brian Holt. The module comes with the random forest
algorithm and the extra-trees method, along with documentation and examples.
• Novelty and Outlier Detection: outlier and novelty detection, by Virgile Fritsch.
• Kernel Approximation: a transform implementing kernel approximation for fast SGD on non-linear kernels by
Andreas Müller.
• Fixed a bug due to atom swapping in Orthogonal Matching Pursuit (OMP) by Vlad Niculae.
• Sparse coding with a precomputed dictionary by Vlad Niculae.
• Mini Batch K-Means performance improvements by Olivier Grisel.
• K-means support for sparse matrices by Mathieu Blondel.
• Improved documentation for developers and for the sklearn.utils module, by Jake Vanderplas.
• Vectorized 20newsgroups dataset loader (sklearn.datasets.fetch_20newsgroups_vectorized)
by Mathieu Blondel.
• Multiclass and multilabel algorithms by Lars Buitinck.
• Utilities for fast computation of mean and variance for sparse matrices by Mathieu Blondel.
• Make sklearn.preprocessing.scale and sklearn.preprocessing.Scaler work on sparse
matrices by Olivier Grisel
• Feature importances using decision trees and/or forest of trees, by Gilles Louppe.
• Parallel implementation of forests of randomized trees by Gilles Louppe.
• sklearn.cross_validation.ShuffleSplit can subsample the train sets as well as the test sets by
Olivier Grisel.

• Errors in the build of the documentation fixed by Andreas Müller.
API changes summary
Here are the code migration instructions when upgrading from scikit-learn version 0.9:
• Some estimators that may overwrite their inputs to save memory previously had overwrite_ parameters;
these have been replaced with copy_ parameters with exactly the opposite meaning.
This particularly affects some of the estimators in linear_model. The default behavior is still to copy
everything passed in.
• The SVMlight dataset loader sklearn.datasets.load_svmlight_file no longer supports loading
two files at once; use load_svmlight_files instead. Also, the (unused) buffer_mb parameter is gone.
• Sparse estimators in the Stochastic Gradient Descent module use dense parameter vector coef_ instead of
sparse_coef_. This significantly improves test time performance.
• The Covariance estimation module now has a robust estimator of covariance, the Minimum Covariance Deter-
minant estimator.
• Cluster evaluation metrics in metrics.cluster have been refactored but the changes are backwards compat-
ible. They have been moved to the metrics.cluster.supervised, along with metrics.cluster.
unsupervised which contains the Silhouette Coefficient.
• The permutation_test_score function now behaves the same way as cross_val_score (i.e. uses
the mean score across the folds.)
• Cross Validation generators now use integer indices (indices=True) by default instead of boolean masks.
This make it more intuitive to use with sparse matrix data.
• The functions used for sparse coding, sparse_encode and sparse_encode_parallel have been com-
bined into sklearn.decomposition.sparse_encode, and the shapes of the arrays have been trans-
posed for consistency with the matrix factorization setting, as opposed to the regression setting.
• Fixed an off-by-one error in the SVMlight/LibSVM file format handling; files generated using sklearn.
datasets.dump_svmlight_file should be re-generated. (They should continue to work, but acciden-
tally had one extra column of zeros prepended.)
• BaseDictionaryLearning class replaced by SparseCodingMixin.
• sklearn.utils.extmath.fast_svd has been renamed sklearn.utils.extmath.
randomized_svd and the default oversampling is now fixed to 10 additional random vectors instead
of doubling the number of components to extract. The new behavior follows the reference paper.
People
The following people contributed to scikit-learn since last release:

• 183 Brian Holt
• 73 Vlad Niculae

• 44 Noel Dawe
• 3 Kenneth C. Arnold
• 3 Shiqiao Du
• 2 Bala Subrahmanyam Varanasi
• 2 DraXus
• 1 Bogdan Trach
• 1 Juan Manuel Caicedo Carvajal
• 1 Nicolas Pinto
• 1 Tiziano Zito
• 1 Xinfan Meng
1.15.18 Version 0.9
September 21, 2011

scikit-learn 0.9 was released on September 2011, three months after the 0.8 release and includes the new modules
Manifold learning, The Dirichlet Process as well as several new algorithms and documentation improvements.
This release also includes the dictionary-learning work developed by Vlad Niculae as part of the Google Summer of
Code program.


Changelog
• New Manifold learning module by Jake Vanderplas and Fabian Pedregosa.

• New Dirichlet Process Gaussian Mixture Model by Alexandre Passos
• Nearest Neighbors module refactoring by Jake Vanderplas : general refactoring, support for sparse matrices in
input, speed and documentation improvements. See the next section for a full list of API changes.
• Improvements on the Feature selection module by Gilles Louppe : refactoring of the RFE classes, documenta-
tion rewrite, increased efficiency and minor API changes.
• Sparse principal components analysis (SparsePCA and MiniBatchSparsePCA) by Vlad Niculae, Gael Varo-
quaux and Alexandre Gramfort
• Printing an estimator now behaves independently of architectures and Python version thanks to Jean Kossaifi.
• Loader for libsvm/svmlight format by Mathieu Blondel and Lars Buitinck
• Documentation improvements: thumbnails in example gallery by Fabian Pedregosa.
• Important bugfixes in Support Vector Machines module (segfaults, bad performance) by Fabian Pedregosa.
• Added Multinomial Naive Bayes and Bernoulli Naive Bayes by Lars Buitinck
• Text feature extraction optimizations by Lars Buitinck
• Chi-Square feature selection (feature_selection.univariate_selection.chi2) by Lars Buit-
inck.
• Generated datasets module refactoring by Gilles Louppe
• Multiclass and multilabel algorithms by Mathieu Blondel
• Ball tree rewrite by Jake Vanderplas

• Implementation of DBSCAN algorithm by Robert Layton

• Kmeans predict and transform by Robert Layton
• Preprocessing module refactoring by Olivier Grisel
• Faster mean shift by Conrad Lee
• New Bootstrap, Random permutations cross-validation a.k.a. Shuffle & Split and various other improve-
ments in cross validation schemes by Olivier Grisel and Gael Varoquaux
• Adjusted Rand index and V-Measure clustering evaluation metrics by Olivier Grisel
• Added Orthogonal Matching Pursuit by Vlad Niculae
• Added 2D-patch extractor utilities in the Feature extraction module by Vlad Niculae
• Implementation of linear_model.LassoLarsCV (cross-validated Lasso solver using the Lars algorithm)
and linear_model.LassoLarsIC (BIC/AIC model selection in Lars) by Gael Varoquaux and Alexandre
Gramfort
• Scalability improvements to metrics.roc_curve by Olivier Hervieu
• Distance helper functions metrics.pairwise.pairwise_distances and metrics.pairwise.
pairwise_kernels by Robert Layton
• Mini-Batch K-Means by Nelle Varoquaux and Peter Prettenhofer.
• mldata utilities by Pietro Berkes.
• olivetti_faces by David Warde-Farley.
API changes summary
Here are the code migration instructions when upgrading from scikit-learn version 0.8:
• The scikits.learn package was renamed sklearn. There is still a scikits.learn package alias for
backward compatibility.
Third-party projects with a dependency on scikit-learn 0.9+ should upgrade their codebase. For instance, under
Linux / MacOSX just run (make a backup first!):
find -name "*.py" | xargs sed -i 's/\bscikits.learn\b/sklearn/g'
• Estimators no longer accept model parameters as fit arguments: instead all parameters must be only
be passed as constructor arguments or using the now public set_params method inherited from base.
BaseEstimator.
Some estimators can still accept keyword arguments on the fit but this is restricted to data-dependent values
(e.g. a Gram matrix or an affinity matrix that are precomputed from the X data matrix.
• The cross_val package has been renamed to cross_validation although there is also a cross_val
package alias in place for backward compatibility.
Third-party projects with a dependency on scikit-learn 0.9+ should upgrade their codebase. For instance, under
Linux / MacOSX just run (make a backup first!):
find -name "*.py" | xargs sed -i 's/\bcross_val\b/cross_validation/g'
• The score_func argument of the sklearn.cross_validation.cross_val_score function is

now expected to accept y_test and y_predicted as only arguments for classification and regression tasks
or X_test for unsupervised estimators.

• gamma parameter for support vector machine algorithms is set to 1 / n_features by default, instead of 1
/ n_samples.
• The sklearn.hmm has been marked as orphaned: it will be removed from scikit-learn in version 0.11 unless
someone steps up to contribute documentation, examples and fix lurking numerical stability issues.
• sklearn.neighbors has been made into a submodule. The two previously available estimators,
NeighborsClassifier and NeighborsRegressor have been marked as deprecated. Their function-
ality has been divided among five new classes: NearestNeighbors for unsupervised neighbors searches,
KNeighborsClassifier & RadiusNeighborsClassifier for supervised classification problems,
and KNeighborsRegressor & RadiusNeighborsRegressor for supervised regression problems.
• sklearn.ball_tree.BallTree has been moved to sklearn.neighbors.BallTree. Using the
former will generate a warning.
• sklearn.linear_model.LARS() and related classes (LassoLARS, LassoLARSCV, etc.) have been re-
named to sklearn.linear_model.Lars().
• All distance metrics and kernels in sklearn.metrics.pairwise now have a Y parameter, which by
default is None. If not given, the result is the distance (or kernel similarity) between each sample in Y. If given,
the result is the pairwise distance (or kernel similarity) between samples in X to Y.
• sklearn.metrics.pairwise.l1_distance is now called manhattan_distance, and by default
returns the pairwise distance. For the component wise distance, set the parameter sum_over_features to
False.
Backward compatibility package aliases and other deprecated classes and functions will be removed in version 0.11.
People
38 people contributed to this release.

• 387 Vlad Niculae
• 168 Fabian Pedregosa (INRIA, Parietal Team)
• 32 Jean Kossaifi
• 30 Conrad Lee
• 22 Pietro Berkes
• 18 andy

• 17 David Warde-Farley
• 12 Brian Holt
• 11 Robert
• 8 Amit Aides
• 6 Salvatore Masecchia
• 5 Paolo Losi
• 4 Vincent Schut
• 3 Alexis Metaireau
• 3 Bryan Silverthorn
• 2 Minwoo Jake Lee
• 1 Keith Goodman
• 1 Lucas Wiman
• 1 Nicolas Pinto
• 1 Thouis (Ray) Jones
1.15.19 Version 0.8
May 11, 2011

scikit-learn 0.8 was released on May 2011, one month after the first “international” scikit-learn coding sprint and is
marked by the inclusion of important modules: Hierarchical clustering, Cross decomposition, Non-negative matrix
factorization (NMF or NNMF), initial support for Python 3 and by important enhancements and bug fixes.
Changelog
Several new modules where introduced during this release:

• New Hierarchical clustering module by Vincent Michel, Bertrand Thirion, Alexandre Gramfort and Gael Varo-
quaux.
• Kernel PCA implementation by Mathieu Blondel
• labeled_faces_in_the_wild by Olivier Grisel.
• New Cross decomposition module by Edouard Duchesnay.
• Non-negative matrix factorization (NMF or NNMF) module Vlad Niculae
• Implementation of the Oracle Approximating Shrinkage algorithm by Virgile Fritsch in the Covariance estima-
tion module.
Some other modules benefited from significant improvements or cleanups.

• Initial support for Python 3: builds and imports cleanly, some modules are usable while others have failing tests
by Fabian Pedregosa.
• decomposition.PCA is now usable from the Pipeline object by Olivier Grisel.
• Guide How to optimize for speed by Olivier Grisel.
• Fixes for memory leaks in libsvm bindings, 64-bit safer BallTree by Lars Buitinck.
• bug and style fixing in K-means algorithm by Jan Schlüter.
• Add attribute converged to Gaussian Mixture Models by Vincent Schut.
• Implemented transform, predict_log_proba in discriminant_analysis.
LinearDiscriminantAnalysis By Mathieu Blondel.
• Refactoring in the Support Vector Machines module and bug fixes by Fabian Pedregosa, Gael Varoquaux and
Amit Aides.
• Refactored SGD module (removed code duplication, better variable naming), added interface for sample weight
by Peter Prettenhofer.
• Wrapped BallTree with Cython by Thouis (Ray) Jones.
• Added function svm.l1_min_c by Paolo Losi.
• Typos, doc style, etc. by Yaroslav Halchenko, Gael Varoquaux, Olivier Grisel, Yann Malet, Nicolas Pinto, Lars
Buitinck and Fabian Pedregosa.
People
People that made this release possible preceded by number of commits:

• 96 Vlad Niculae
• 32 Paolo Losi
• 22 Nicolas Pinto
– 7 Lars Buitinck
– 6 Vincent Michel
– 5 Bertrand Thirion
– 4 Thouis (Ray) Jones
– 4 Vincent Schut
– 3 Jan Schlüter
– 2 Julien Miotte

– 2 Matthieu Perrot
– 2 Yann Malet
– 2 Yaroslav Halchenko
– 1 Amit Aides
– 1 Andreas Müller
– 1 Feth Arezki
– 1 Meng Xinfan
1.15.20 Version 0.7
March 2, 2011
scikit-learn 0.7 was released in March 2011, roughly three months after the 0.6 release. This release is marked by the
speed improvements in existing algorithms like k-Nearest Neighbors and K-Means algorithm and by the inclusion of
an efficient algorithm for computing the Ridge Generalized Cross Validation solution. Unlike the preceding release,
no new modules where added to this release.
Changelog
• Performance improvements for Gaussian Mixture Model sampling [Jan Schlüter].

• Implementation of efficient leave-one-out cross-validated Ridge in linear_model.RidgeCV [Mathieu
Blondel]
• Better handling of collinearity and early stopping in linear_model.lars_path [Alexandre Gramfort and
Fabian Pedregosa].
• Fixes for liblinear ordering of labels and sign of coefficients [Dan Yamins, Paolo Losi, Mathieu Blondel and
Fabian Pedregosa].
• Performance improvements for Nearest Neighbors algorithm in high-dimensional spaces [Fabian Pedregosa].
• Performance improvements for cluster.KMeans [Gael Varoquaux and James Bergstra].
• Sanity checks for SVM-based classes [Mathieu Blondel].
• Refactoring of neighbors.NeighborsClassifier and neighbors.kneighbors_graph: added
different algorithms for the k-Nearest Neighbor Search and implemented a more stable algorithm for finding
barycenter weights. Also added some developer documentation for this module, see notes_neighbors for more
information [Fabian Pedregosa].
• Documentation improvements: Added pca.RandomizedPCA and linear_model.
LogisticRegression to the class reference. Also added references of matrices used for clustering
and other fixes [Gael Varoquaux, Fabian Pedregosa, Mathieu Blondel, Olivier Grisel, Virgile Fritsch ,
Emmanuelle Gouillart]
• Binded decision_function in classes that make use of liblinear, dense and sparse variants, like svm.
LinearSVC or linear_model.LogisticRegression [Fabian Pedregosa].
• Performance and API improvements to metrics.euclidean_distances and to pca.
RandomizedPCA [James Bergstra].
• Fix compilation issues under NetBSD [Kamel Ibn Hassen Derouiche]
• Allow input sequences of different lengths in hmm.GaussianHMM [Ron Weiss].

• Fix bug in affinity propagation caused by incorrect indexing [Xinfan Meng]
People

• 19 James Bergstra
• 14 Dan Yamins
• 4 Ron Weiss
• 2 Vincent Dubourg
• 1 Kamel Ibn Hassen Derouiche
• 1 Paolo Losi
• 1 VirgileFritsch
• 1 Xinfan Meng
1.15.21 Version 0.6
December 21, 2010

scikit-learn 0.6 was released on December 2010. It is marked by the inclusion of several new modules and a general
renaming of old ones. It is also marked by the inclusion of new example, including applications to real-world datasets.
Changelog
• New stochastic gradient descent module by Peter Prettenhofer. The module comes with complete documentation
and examples.
• Improved svm module: memory consumption has been reduced by 50%, heuristic to automatically set class
weights, possibility to assign weights to samples (see SVM: Weighted samples for an example).
• New Gaussian Processes module by Vincent Dubourg. This module also has great documenta-
tion and some very neat examples. See example_gaussian_process_plot_gp_regression.py or exam-
ple_gaussian_process_plot_gp_probabilistic_classification_after_regression.py for a taste of what can be done.
• It is now possible to use liblinear’s Multi-class SVC (option multi_class in svm.LinearSVC)
• New features and performance improvements of text feature extraction.

• Improved sparse matrix support, both in main classes (grid_search.GridSearchCV) as in modules

sklearn.svm.sparse and sklearn.linear_model.sparse.
• Lots of cool new examples and a new section that uses real-world datasets was created. These include: Faces
recognition example using eigenfaces and SVMs, Species distribution modeling, Libsvm GUI, Wikipedia princi-
pal eigenvector and others.
• Faster Least Angle Regression algorithm. It is now 2x faster than the R version on worst case and up to 10x
times faster on some cases.
• Faster coordinate descent algorithm. In particular, the full path version of lasso (linear_model.
lasso_path) is more than 200x times faster than before.
• It is now possible to get probability estimates from a linear_model.LogisticRegression model.
• module renaming: the glm module has been renamed to linear_model, the gmm module has been included into
the more general mixture model and the sgd module has been included in linear_model.
• Lots of bug fixes and documentation improvements.
People

• 33 Vincent Dubourg
• 21 Ron Weiss
• 3 Anne-Laure Fouque
• 2 Ronan Amicel
• 1 Christian Osendorfer
1.15.22 Version 0.5
October 11, 2010
Changelog
New classes
• Support for sparse matrices in some classifiers of modules svm and linear_model (see svm.
sparse.SVC, svm.sparse.SVR, svm.sparse.LinearSVC, linear_model.sparse.Lasso,
linear_model.sparse.ElasticNet)

• New pipeline.Pipeline object to compose different estimators.

• Recursive Feature Elimination routines in module Feature selection.
• Addition of various classes capable of cross validation in the linear_model module (linear_model.
LassoCV , linear_model.ElasticNetCV , etc.).
• New, more efficient LARS algorithm implementation. The Lasso variant of the algorithm is also implemented.
See linear_model.lars_path, linear_model.Lars and linear_model.LassoLars.
• New Hidden Markov Models module (see classes hmm.GaussianHMM, hmm.MultinomialHMM, hmm.
GMMHMM)
• New module feature_extraction (see class reference)
• New FastICA algorithm in module sklearn.fastica
Documentation
• Improved documentation for many modules, now separating narrative documentation from the class reference.
As an example, see documentation for the SVM module and the complete class reference.
Fixes
• API changes: adhere variable names to PEP-8, give more meaningful names.
• Fixes for svm module to run on a shared memory context (multiprocessing).
• It is again possible to generate latex (and thus PDF) from the sphinx docs.
Examples
• new examples using some of the mlcomp datasets: sphx_glr_auto_examples_mlcomp_sparse_document_classif

py (since removed) and Classification of text documents using sparse features
• Many more examples. See here the full list of examples.
External dependencies
• Joblib is now a dependency of this package, although it is shipped with (sklearn.externals.joblib).
Removed modules
• Module ann (Artificial Neural Networks) has been removed from the distribution. Users wanting this sort of
algorithms should take a look into pybrain.
Misc
• New sphinx theme for the web page.

Authors
The following is a list of authors for this release, preceded by number of commits:
• 40 Vincent Michel
• 38 Ron Weiss
• 23 Matthieu Perrot
• 9 VirgileFritsch
• 1 Ariel Rokem
• 1 Matthieu Brucher
1.15.23 Version 0.4
August 26, 2010
Changelog
Major changes in this release include:

• Coordinate Descent algorithm (Lasso, ElasticNet) refactoring & speed improvements (roughly 100x times
faster).
• Coordinate Descent Refactoring (and bug fixing) for consistency with R’s package GLMNET.
• New metrics module.
• New GMM module contributed by Ron Weiss.
• Implementation of the LARS algorithm (without Lasso variant for now).
• feature_selection module redesign.
• Migration to GIT as version control system.
• Removal of obsolete attrselect module.
• Rename of private compiled extensions (added underscore).
• Removal of legacy unmaintained code.
• Documentation improvements (both docstring and rst).
• Improvement of the build system to (optionally) link with MKL. Also, provide a lite BLAS implementation in
case no system-wide BLAS is found.

• Lots of new examples.

• Many, many bug fixes . . .
Authors
The committer list for this release is the following (preceded by number of commits):
• 2 Vincent Michel
• 1 Chris Filo Gorgolewski
1.15.24 Earlier versions
Earlier versions included contributions by Fred Mailhot, David Cooke, David Huard, Dave Morrill, Ed Schofield,
Travis Oliphant, Pearu Peterson.

CHAPTER
TWO
SCIKIT-LEARN TUTORIALS
2.1 An introduction to machine learning with scikit-learn
Section contents
In this section, we introduce the machine learning vocabulary that we use throughout scikit-learn and give a simple
learning example.
2.1.1 Machine learning: the problem setting
In general, a learning problem considers a set of n samples of data and then tries to predict properties of unknown data.
If each sample is more than a single number and, for instance, a multi-dimensional entry (aka multivariate data), it is
said to have several attributes or features.
Learning problems fall into a few categories:
• supervised learning, in which the data comes with additional attributes that we want to predict (Click here to go
to the scikit-learn supervised learning page).This problem can be either:
– classification: samples belong to two or more classes and we want to learn from already labeled data how
to predict the class of unlabeled data. An example of a classification problem would be handwritten digit
recognition, in which the aim is to assign each input vector to one of a finite number of discrete categories.
Another way to think of classification is as a discrete (as opposed to continuous) form of supervised
learning where one has a limited number of categories and for each of the n samples provided, one is to
try to label them with the correct category or class.
– regression: if the desired output consists of one or more continuous variables, then the task is called
regression. An example of a regression problem would be the prediction of the length of a salmon as a
function of its age and weight.
• unsupervised learning, in which the training data consists of a set of input vectors x without any corresponding
target values. The goal in such problems may be to discover groups of similar examples within the data, where
it is called clustering, or to determine the distribution of data within the input space, known as density estima-
tion, or to project the data from a high-dimensional space down to two or three dimensions for the purpose of
visualization (Click here to go to the Scikit-Learn unsupervised learning page).
129
Training set and testing set
Machine learning is about learning some properties of a data set and then testing those properties against another
data set. A common practice in machine learning is to evaluate an algorithm by splitting a data set into two. We
call one of those sets the training set, on which we learn some properties; we call the other set the testing set, on
which we test the learned properties.
2.1.2 Loading an example dataset
scikit-learn comes with a few standard datasets, for instance the iris and digits datasets for classification and the boston
house prices dataset for regression.
In the following, we start a Python interpreter from our shell and then load the iris and digits datasets. Our
notational convention is that $ denotes the shell prompt while >>> denotes the Python interpreter prompt:
$ python
>>> from sklearn import datasets
>>> iris = datasets.load_iris()
>>> digits = datasets.load_digits()
A dataset is a dictionary-like object that holds all the data and some metadata about the data. This data is stored in
the .data member, which is a n_samples, n_features array. In the case of supervised problem, one or more
response variables are stored in the .target member. More details on the different datasets can be found in the
dedicated section.
For instance, in the case of the digits dataset, digits.data gives access to the features that can be used to classify
the digits samples:
>>> print(digits.data)
[[ 0. 0. 5. ... 0. 0. 0.]
[ 0. 0. 0. ... 10. 0. 0.]
[ 0. 0. 0. ... 16. 9. 0.]
...
[ 0. 0. 1. ... 6. 0. 0.]
[ 0. 0. 2. ... 12. 0. 0.]
[ 0. 0. 10. ... 12. 1. 0.]]
and digits.target gives the ground truth for the digit dataset, that is the number corresponding to each digit
image that we are trying to learn:
>>> digits.target
array([0, 1, 2, ..., 8, 9, 8])
Shape of the data arrays
The data is always a 2D array, shape (n_samples, n_features), although the original data may have had a
different shape. In the case of the digits, each original sample is an image of shape (8, 8) and can be accessed
using:
>>> digits.images[0]
array([[ 0., 0., 5., 13., 9., 1., 0., 0.],
[ 0., 0., 13., 15., 10., 15., 5., 0.],
[ 0., 3., 15., 2., 0., 11., 8., 0.],
[ 0., 4., 12., 0., 0., 8., 8., 0.],
[ 0., 5., 8., 0., 0., 9., 8., 0.],
[ 0., 4., 11., 0., 1., 12., 7., 0.],
[ 0., 2., 14., 5., 10., 12., 0., 0.],
130 [ 0., 0., 6., 13., 10., 0., 0., 0.]]) Chapter 2. scikit-learn Tutorials
The simple example on this dataset illustrates how starting from the original problem one can shape the data for
consumption in scikit-learn.
Loading from external datasets
To load from an external dataset, please refer to loading external datasets.
2.1.3 Learning and predicting
In the case of the digits dataset, the task is to predict, given an image, which digit it represents. We are given samples
of each of the 10 possible classes (the digits zero through nine) on which we fit an estimator to be able to predict the
classes to which unseen samples belong.
In scikit-learn, an estimator for classification is a Python object that implements the methods fit(X, y) and
predict(T).
An example of an estimator is the class sklearn.svm.SVC, which implements support vector classification. The
estimator’s constructor takes as arguments the model’s parameters.
For now, we will consider the estimator as a black box:
>>> from sklearn import svm

>>> clf = svm.SVC(gamma=0.001, C=100.)
Choosing the parameters of the model
In this example, we set the value of gamma manually. To find good values for these parameters, we can use tools
such as grid search and cross validation.
The clf (for classifier) estimator instance is first fitted to the model; that is, it must learn from the model. This is
done by passing our training set to the fit method. For the training set, we’ll use all the images from our dataset,
except for the last image, which we’ll reserve for our predicting. We select the training set with the [:-1] Python
syntax, which produces a new array that contains all but the last item from digits.data:
>>> clf.fit(digits.data[:-1], digits.target[:-1])

SVC(C=100.0, cache_size=200, class_weight=None, coef0=0.0,
decision_function_shape='ovr', degree=3, gamma=0.001, kernel='rbf',
max_iter=-1, probability=False, random_state=None, shrinking=True,
tol=0.001, verbose=False)
Now you can predict new values. In this case, you’ll predict using the last image from digits.data. By predicting,
you’ll determine the image from the training set that best matches the last image.
>>> clf.predict(digits.data[-1:])
array([8])
2.1. An introduction to machine learning with scikit-learn 131

The corresponding image is: As you can see, it is a challenging task: after all, the images
are of poor resolution. Do you agree with the classifier?
A complete example of this classification problem is available as an example that you can run and study: Recognizing
hand-written digits.
2.1.4 Model persistence
It is possible to save a model in scikit-learn by using Python’s built-in persistence model, pickle:

>>> clf = svm.SVC(gamma='scale')
>>> X, y = iris.data, iris.target
>>> clf.fit(X, y)
decision_function_shape='ovr', degree=3, gamma='scale', kernel='rbf',
>>> import pickle

>>> s = pickle.dumps(clf)
>>> clf2 = pickle.loads(s)
>>> clf2.predict(X[0:1])
array([0])
>>> y[0]
0
In the specific case of scikit-learn, it may be more interesting to use joblib’s replacement for pickle (joblib.dump
& joblib.load), which is more efficient on big data but it can only pickle to the disk and not to a string:
>>> from joblib import dump, load

>>> dump(clf, 'filename.joblib')
Later, you can reload the pickled model (possibly in another Python process) with:
>>> clf = load('filename.joblib')
Note: joblib.dump and joblib.load functions also accept file-like object instead of filenames. More infor-
mation on data persistence with Joblib is available here.
Note that pickle has some security and maintainability issues. Please refer to section Model persistence for more
detailed information about model persistence with scikit-learn.
132 Chapter 2. scikit-learn Tutorials

2.1.5 Conventions
scikit-learn estimators follow certain rules to make their behavior more predictive. These are described in more detail
in the Glossary of Common Terms and API Elements.
Type casting
Unless otherwise specified, input will be cast to float64:

>>> from sklearn import random_projection
>>> rng = np.random.RandomState(0)

>>> X = rng.rand(10, 2000)
>>> X = np.array(X, dtype='float32')
>>> X.dtype
dtype('float32')
>>> transformer = random_projection.GaussianRandomProjection()

>>> X_new = transformer.fit_transform(X)
>>> X_new.dtype
dtype('float64')
In this example, X is float32, which is cast to float64 by fit_transform(X).

Regression targets are cast to float64 and classification targets are maintained:

>>> from sklearn.svm import SVC
>>> clf = SVC(gamma='scale')
>>> clf.fit(iris.data, iris.target)
>>> list(clf.predict(iris.data[:3]))
[0, 0, 0]
>>> clf.fit(iris.data, iris.target_names[iris.target])

>>> list(clf.predict(iris.data[:3]))
['setosa', 'setosa', 'setosa']
Here, the first predict() returns an integer array, since iris.target (an integer array) was used in fit. The
second predict() returns a string array, since iris.target_names was for fitting.
Refitting and updating parameters
Hyper-parameters of an estimator can be updated after it has been constructed via the set_params() method. Calling
fit() more than once will overwrite what was learned by any previous fit():
2.1. An introduction to machine learning with scikit-learn 133



>>> X = rng.rand(100, 10)
>>> y = rng.binomial(1, 0.5, 100)
>>> X_test = rng.rand(5, 10)
>>> clf = SVC()

>>> clf.set_params(kernel='linear').fit(X, y)
decision_function_shape='ovr', degree=3, gamma='auto_deprecated',
kernel='linear', max_iter=-1, probability=False, random_state=None,
shrinking=True, tol=0.001, verbose=False)
>>> clf.predict(X_test)
array([1, 0, 1, 1, 0])
>>> clf.set_params(kernel='rbf', gamma='scale').fit(X, y)

>>> clf.predict(X_test)
array([0, 0, 0, 1, 0])
Here, the default kernel rbf is first changed to linear via SVC.set_params() after the estimator has been
constructed, and changed back to rbf to refit the estimator and to make a second prediction.
Multiclass vs. multilabel fitting
When using multiclass classifiers, the learning and prediction task that is performed is dependent on the
format of the target data fit upon:

>>> from sklearn.multiclass import OneVsRestClassifier
>>> from sklearn.preprocessing import LabelBinarizer
>>> X = [[1, 2], [2, 4], [4, 5], [3, 2], [3, 1]]
>>> y = [0, 0, 1, 1, 2]
>>> classif = OneVsRestClassifier(estimator=SVC(gamma='scale',

... random_state=0))
>>> classif.fit(X, y).predict(X)
array([0, 0, 1, 1, 2])
In the above case, the classifier is fit on a 1d array of multiclass labels and the predict() method therefore provides
corresponding multiclass predictions. It is also possible to fit upon a 2d array of binary label indicators:
>>> y = LabelBinarizer().fit_transform(y)
array([[1, 0, 0],
[1, 0, 0],
[0, 1, 0],
[0, 0, 0],
[0, 0, 0]])

Here, the classifier is fit() on a 2d binary label representation of y, using the LabelBinarizer. In this case
predict() returns a 2d array representing the corresponding multilabel predictions.
Note that the fourth and fifth instances returned all zeroes, indicating that they matched none of the three labels fit
upon. With multilabel outputs, it is similarly possible for an instance to be assigned multiple labels:
>>> from sklearn.preprocessing import MultiLabelBinarizer
>>> y = [[0, 1], [0, 2], [1, 3], [0, 2, 3], [2, 4]]
>>> y = MultiLabelBinarizer().fit_transform(y)
array([[1, 1, 0, 0, 0],
[1, 0, 1, 0, 0],
[0, 1, 0, 1, 0],
[1, 0, 1, 0, 0],
[1, 0, 1, 0, 0]])
In this case, the classifier is fit upon instances each assigned multiple labels. The MultiLabelBinarizer is
used to binarize the 2d array of multilabels to fit upon. As a result, predict() returns a 2d array with multiple
predicted labels for each instance.
2.2 A tutorial on statistical-learning for scientific data processing
Statistical learning
Machine learning is a technique with a growing importance, as the size of the datasets experimental sciences are fac-
ing is rapidly growing. Problems it tackles range from building a prediction function linking different observations,
to classifying observations, or learning the structure in an unlabeled dataset.
This tutorial will explore statistical learning, the use of machine learning techniques with the goal of statistical
inference: drawing conclusions on the data at hand.
Scikit-learn is a Python module integrating classic machine learning algorithms in the tightly-knit world of scientific
Python packages (NumPy, SciPy, matplotlib).
2.2.1 Statistical learning: the setting and the estimator object in scikit-learn
Datasets
Scikit-learn deals with learning information from one or more datasets that are represented as 2D arrays. They can be
understood as a list of multi-dimensional observations. We say that the first axis of these arrays is the samples axis,
while the second is the features axis.
A simple example shipped with scikit-learn: iris dataset

>>> data = iris.data
>>> data.shape
(150, 4)
It is made of 150 observations of irises, each described by 4 features: their sepal and petal length and width, as
detailed in iris.DESCR.
2.2. A tutorial on statistical-learning for scientific data processing 135

When the data is not initially in the (n_samples, n_features) shape, it needs to be preprocessed in order to
be used by scikit-learn.
An example of reshaping data would be the digits dataset
The digits dataset is made of 1797 8x8 images of hand-written digits

>>> digits.images.shape
(1797, 8, 8)
>>> import matplotlib.pyplot as plt
>>> plt.imshow(digits.images[-1], cmap=plt.cm.gray_r)
<matplotlib.image.AxesImage object at ...>
To use this dataset with scikit-learn, we transform each 8x8 image into a feature vector of length 64
>>> data = digits.images.reshape((digits.images.shape[0], -1))
Estimators objects
Fitting data: the main API implemented by scikit-learn is that of the estimator. An estimator is any object that learns
from data; it may be a classification, regression or clustering algorithm or a transformer that extracts/filters useful
features from raw data.
All estimator objects expose a fit method that takes a dataset (usually a 2-d array):
>>> estimator.fit(data)
Estimator parameters: All the parameters of an estimator can be set when it is instantiated or by modifying the
corresponding attribute:
>>> estimator = Estimator(param1=1, param2=2)

>>> estimator.param1
1
Estimated parameters: When data is fitted with an estimator, parameters are estimated from the data at hand. All the
estimated parameters are attributes of the estimator object ending by an underscore:
>>> estimator.estimated_param_

2.2.2 Supervised learning: predicting an output variable from high-dimensional ob-

servations
The problem solved in supervised learning
Supervised learning consists in learning the link between two datasets: the observed data X and an external variable
y that we are trying to predict, usually called “target” or “labels”. Most often, y is a 1D array of length n_samples.
All supervised estimators in scikit-learn implement a fit(X, y) method to fit the model and a predict(X)
method that, given unlabeled observations X, returns the predicted labels y.
Vocabulary: classification and regression
If the prediction task is to classify the observations in a set of finite labels, in other words to “name” the objects
observed, the task is said to be a classification task. On the other hand, if the goal is to predict a continuous target
variable, it is said to be a regression task.
When doing classification in scikit-learn, y is a vector of integers or strings.
Note: See the Introduction to machine learning with scikit-learn Tutorial for a quick run-through on the basic
machine learning vocabulary used within scikit-learn.
Nearest neighbor and the curse of dimensionality
Classifying irises:

The iris dataset is a

classification task consisting in identifying 3 different types of irises (Setosa, Versicolour, and Virginica) from their
petal and sepal length and width:
>>> iris_X = iris.data
>>> iris_y = iris.target
>>> np.unique(iris_y)
array([0, 1, 2])
k-Nearest neighbors classifier
The simplest possible classifier is the nearest neighbor: given a new observation X_test, find in the training set (i.e.
the data used to train the estimator) the observation with the closest feature vector. (Please see the Nearest Neighbors
section of the online Scikit-learn documentation for more information about this type of classifier.)
Training set and testing set
While experimenting with any learning algorithm, it is important not to test the prediction of an estimator on the
data used to fit the estimator as this would not be evaluating the performance of the estimator on new data. This is
why datasets are often split into train and test data.

KNN (k nearest neighbors) classification example:
>>> # Split iris data in train and test data

>>> # A random permutation, to split the data randomly
>>> np.random.seed(0)
>>> indices = np.random.permutation(len(iris_X))
>>> iris_X_train = iris_X[indices[:-10]]
>>> iris_y_train = iris_y[indices[:-10]]
>>> iris_X_test = iris_X[indices[-10:]]
>>> iris_y_test = iris_y[indices[-10:]]
>>> # Create and fit a nearest-neighbor classifier
>>> from sklearn.neighbors import KNeighborsClassifier
>>> knn = KNeighborsClassifier()
>>> knn.fit(iris_X_train, iris_y_train)
KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',
metric_params=None, n_jobs=None, n_neighbors=5, p=2,
weights='uniform')
>>> knn.predict(iris_X_test)
array([1, 2, 1, 0, 0, 0, 2, 1, 2, 0])
>>> iris_y_test
array([1, 1, 1, 0, 0, 0, 2, 1, 2, 0])
The curse of dimensionality
For an estimator to be effective, you need the distance between neighboring points to be less than some value 𝑑, which
depends on the problem. In one dimension, this requires on average 𝑛 ∼ 1/𝑑 points. In the context of the above 𝑘-NN
example, if the data is described by just one feature with values ranging from 0 to 1 and with 𝑛 training observations,
then new data will be no further away than 1/𝑛. Therefore, the nearest neighbor decision rule will be efficient as soon
as 1/𝑛 is small compared to the scale of between-class feature variations.
If the number of features is 𝑝, you now require 𝑛 ∼ 1/𝑑𝑝 points. Let’s say that we require 10 points in one dimension:
now 10𝑝 points are required in 𝑝 dimensions to pave the [0, 1] space. As 𝑝 becomes large, the number of training points
required for a good estimator grows exponentially.
For example, if each point is just a single number (8 bytes), then an effective 𝑘-NN estimator in a paltry 𝑝 ∼ 20
dimensions would require more training data than the current estimated size of the entire internet (±1000 Exabytes or

so).
This is called the curse of dimensionality and is a core problem that machine learning addresses.
Linear model: from regression to sparsity
Diabetes dataset
The diabetes dataset consists of 10 physiological variables (age, sex, weight, blood pressure) measure on 442
patients, and an indication of disease progression after one year:
>>> diabetes = datasets.load_diabetes()
>>> diabetes_X_train = diabetes.data[:-20]
>>> diabetes_X_test = diabetes.data[-20:]
>>> diabetes_y_train = diabetes.target[:-20]
>>> diabetes_y_test = diabetes.target[-20:]
The task at hand is to predict disease progression from physiological variables.
Linear regression
LinearRegression, in its simplest form, fits a linear model to the data set by adjusting a set
of parameters in order to make the sum of the squared residuals of the model as small as possible.
Linear models: 𝑦 = 𝑋𝛽 + 𝜖
• 𝑋: data
• 𝑦: target variable
• 𝛽: Coefficients
• 𝜖: Observation noise
>>> from sklearn import linear_model

>>> regr = linear_model.LinearRegression()
>>> regr.fit(diabetes_X_train, diabetes_y_train)
...
LinearRegression(copy_X=True, fit_intercept=True, n_jobs=None,
normalize=False)
>>> print(regr.coef_)
[ 0.30349955 -237.63931533 510.53060544 327.73698041 -814.13170937
492.81458798 102.84845219 184.60648906 743.51961675 76.09517222]
>>> # The mean square error

>>> np.mean((regr.predict(diabetes_X_test) - diabetes_y_test)**2)

...
2004.56760268...
>>> # Explained variance score: 1 is perfect prediction

>>> # and 0 means that there is no linear relationship
>>> # between X and y.
>>> regr.score(diabetes_X_test, diabetes_y_test)
0.5850753022690...
Shrinkage
If there are few data points per dimension, noise in the observations induces high variance:
>>> X = np.c_[ .5, 1].T

>>> y = [.5, 1]
>>> test = np.c_[ 0, 2].T
>>> regr = linear_model.LinearRegression()

>>> plt.figure()
>>> for _ in range(6):
... this_X = .1 * np.random.normal(size=(2, 1)) + X
... regr.fit(this_X, y)
... plt.plot(test, regr.predict(test))
... plt.scatter(this_X, y, s=3)
A solution in high-dimensional statistical learning is to shrink the regression coefficients to zero: any
two randomly chosen set of observations are likely to be uncorrelated. This is called Ridge regression:

>>> regr = linear_model.Ridge(alpha=.1)
>>> plt.figure()
>>> for _ in range(6):
... this_X = .1 * np.random.normal(size=(2, 1)) + X
... regr.fit(this_X, y)
... plt.plot(test, regr.predict(test))
... plt.scatter(this_X, y, s=3)
This is an example of bias/variance tradeoff: the larger the ridge alpha parameter, the higher the bias and the lower
the variance.
We can choose alpha to minimize left out error, this time using the diabetes dataset rather than our synthetic data:
>>> from __future__ import print_function

>>> alphas = np.logspace(-4, -1, 6)
>>> print([regr.set_params(alpha=alpha)
... .fit(diabetes_X_train, diabetes_y_train)
... .score(diabetes_X_test, diabetes_y_test)
... for alpha in alphas])
...
[0.5851110683883..., 0.5852073015444..., 0.5854677540698...,
0.5855512036503..., 0.5830717085554..., 0.57058999437...]
Note: Capturing in the fitted parameters noise that prevents the model to generalize to new data is called overfitting.
The bias introduced by the ridge regression is called a regularization.
Sparsity
Fitting only features 1 and 2

Note: A representation of the full diabetes dataset would involve 11 dimensions (10 feature dimensions and one of
the target variable). It is hard to develop an intuition on such representation, but it may be useful to keep in mind that
it would be a fairly empty space.
We can see that, although feature 2 has a strong coefficient on the full model, it conveys little information on y when
considered with feature 1.
To improve the conditioning of the problem (i.e. mitigating the The curse of dimensionality), it would be interesting
to select only the informative features and set non-informative ones, like feature 2 to 0. Ridge regression will decrease
their contribution, but not set them to zero. Another penalization approach, called Lasso (least absolute shrinkage and
selection operator), can set some coefficients to zero. Such methods are called sparse method and sparsity can be
seen as an application of Occam’s razor: prefer simpler models.
>>> regr = linear_model.Lasso()

>>> scores = [regr.set_params(alpha=alpha)
... .fit(diabetes_X_train, diabetes_y_train)
... .score(diabetes_X_test, diabetes_y_test)
... for alpha in alphas]
>>> best_alpha = alphas[scores.index(max(scores))]
>>> regr.alpha = best_alpha
>>> regr.fit(diabetes_X_train, diabetes_y_train)
Lasso(alpha=0.025118864315095794, copy_X=True, fit_intercept=True,
max_iter=1000, normalize=False, positive=False, precompute=False,
random_state=None, selection='cyclic', tol=0.0001, warm_start=False)
[ 0. -212.43764548 517.19478111 313.77959962 -160.8303982 -0.
-187.19554705 69.38229038 508.66011217 71.84239008]

Different algorithms for the same problem
Different algorithms can be used to solve the same mathematical problem. For instance the Lasso object in scikit-
learn solves the lasso regression problem using a coordinate descent method, that is efficient on large datasets.
However, scikit-learn also provides the LassoLars object using the LARS algorithm, which is very efficient for
problems in which the weight vector estimated is very sparse (i.e. problems with very few observations).
Classification
For classification, as in the labeling iris task, linear regression is not

the right approach as it will give too much weight to data far from the decision frontier. A linear approach is to fit a
sigmoid function or logistic function:
1
𝑦 = sigmoid(𝑋𝛽 − offset) + 𝜖 = +𝜖
1 + exp(−𝑋𝛽 + offset)
>>> log = linear_model.LogisticRegression(solver='lbfgs', C=1e5,

... multi_class='multinomial')
>>> log.fit(iris_X_train, iris_y_train)
LogisticRegression(C=100000.0, class_weight=None, dual=False,
fit_intercept=True, intercept_scaling=1, max_iter=100,
multi_class='multinomial', n_jobs=None, penalty='l2', random_state=None,
solver='lbfgs', tol=0.0001, verbose=0, warm_start=False)
This is known as LogisticRegression.

Multiclass classification
If you have several classes to predict, an option often used is to fit one-versus-all classifiers and then use a voting
heuristic for the final decision.
Shrinkage and sparsity with logistic regression
The C parameter controls the amount of regularization in the LogisticRegression object: a large value
for C results in less regularization. penalty="l2" gives Shrinkage (i.e. non-sparse coefficients), while
penalty="l1" gives Sparsity.
Exercise
Try classifying the digits dataset with nearest neighbors and a linear model. Leave out the last 10% and test
prediction performance on these observations.
from sklearn import datasets, neighbors, linear_model
digits = datasets.load_digits()
X_digits = digits.data / digits.data.max()
y_digits = digits.target
Solution: ../../auto_examples/exercises/plot_digits_classification_exercise.py
Support vector machines (SVMs)
Linear SVMs
Support Vector Machines belong to the discriminant model family: they try to find a combination of samples to build
a plane maximizing the margin between the two classes. Regularization is set by the C parameter: a small value for C
means the margin is calculated using many or all of the observations around the separating line (more regularization);
a large value for C means the margin is calculated on observations close to the separating line (less regularization).
Unregularized SVM Regularized SVM (default)

Example:
• Plot different SVM classifiers in the iris dataset
SVMs can be used in regression –SVR (Support Vector Regression)–, or in classification –SVC (Support Vector Clas-
sification).

>>> svc = svm.SVC(kernel='linear')
>>> svc.fit(iris_X_train, iris_y_train)
kernel='linear', max_iter=-1, probability=False, random_state=None,
Warning: Normalizing data

For many estimators, including the SVMs, having datasets with unit standard deviation for each feature is important
to get good prediction.
Using kernels
Classes are not always linearly separable in feature space. The solution is to build a decision function that is not linear
but may be polynomial instead. This is done using the kernel trick that can be seen as creating a decision energy by
positioning kernels on observations:
Linear kernel Polynomial kernel
>>> svc = svm.SVC(kernel='linear') >>> svc = svm.SVC(kernel='poly',

... degree=3)
>>> # degree: polynomial degree

RBF kernel (Radial Basis Function)
>>> svc = svm.SVC(kernel='rbf')

>>> # gamma: inverse of size of
>>> # radial kernel
Interactive example
See the SVM GUI to download svm_gui.py; add data points of both classes with right and left button, fit the
model and change parameters and data.

Exercise
Try classifying classes 1 and 2 from the iris dataset with SVMs, with the 2 first features. Leave out 10% of each
class and test prediction performance on these observations.
Warning: the classes are ordered, do not leave out the last 10%, you would be testing on only one class.
Hint: You can use the decision_function method on a grid to get intuitions.
iris = datasets.load_iris()
X = iris.data
y = iris.target
X = X[y != 0, :2]
y = y[y != 0]
Solution: ../../auto_examples/exercises/plot_iris_exercise.py
2.2.3 Model selection: choosing estimators and their parameters
Score, and cross-validated scores
As we have seen, every estimator exposes a score method that can judge the quality of the fit (or the prediction) on
new data. Bigger is better.
>>> from sklearn import datasets, svm

>>> X_digits = digits.data
>>> y_digits = digits.target
>>> svc = svm.SVC(C=1, kernel='linear')
>>> svc.fit(X_digits[:-100], y_digits[:-100]).score(X_digits[-100:], y_digits[-100:])
0.98
To get a better measure of prediction accuracy (which we can use as a proxy for goodness of fit of the model), we can
successively split the data in folds that we use for training and testing:

>>> X_folds = np.array_split(X_digits, 3)
>>> y_folds = np.array_split(y_digits, 3)
>>> scores = list()
>>> for k in range(3):
... # We use 'list' to copy, in order to 'pop' later on
... X_train = list(X_folds)
... X_test = X_train.pop(k)
... X_train = np.concatenate(X_train)
... y_train = list(y_folds)
... y_test = y_train.pop(k)
... y_train = np.concatenate(y_train)
... scores.append(svc.fit(X_train, y_train).score(X_test, y_test))
>>> print(scores)
[0.934..., 0.956..., 0.939...]
This is called a KFold cross-validation.

Cross-validation generators
Scikit-learn has a collection of classes which can be used to generate lists of train/test indices for popular cross-
validation strategies.
They expose a split method which accepts the input dataset to be split and yields the train/test set indices for each
iteration of the chosen cross-validation strategy.
This example shows an example usage of the split method.
>>> from sklearn.model_selection import KFold, cross_val_score

>>> X = ["a", "a", "a", "b", "b", "c", "c", "c", "c", "c"]
>>> k_fold = KFold(n_splits=5)
>>> for train_indices, test_indices in k_fold.split(X):
... print('Train: %s | test: %s' % (train_indices, test_indices))
Train: [2 3 4 5 6 7 8 9] | test: [0 1]
Train: [0 1 4 5 6 7 8 9] | test: [2 3]
Train: [0 1 2 3 6 7 8 9] | test: [4 5]
Train: [0 1 2 3 4 5 8 9] | test: [6 7]
Train: [0 1 2 3 4 5 6 7] | test: [8 9]
The cross-validation can then be performed easily:
>>> [svc.fit(X_digits[train], y_digits[train]).score(X_digits[test], y_digits[test])

... for train, test in k_fold.split(X_digits)]
[0.963..., 0.922..., 0.963..., 0.963..., 0.930...]
The cross-validation score can be directly calculated using the cross_val_score helper. Given an estimator, the
cross-validation object and the input dataset, the cross_val_score splits the data repeatedly into a training and a
testing set, trains the estimator using the training set and computes the scores based on the testing set for each iteration
of cross-validation.
By default the estimator’s score method is used to compute the individual scores.
Refer the metrics module to learn more on the available scoring methods.
>>> cross_val_score(svc, X_digits, y_digits, cv=k_fold, n_jobs=-1)

array([0.96388889, 0.92222222, 0.9637883 , 0.9637883 , 0.93036212])
n_jobs=-1 means that the computation will be dispatched on all the CPUs of the computer.
Alternatively, the scoring argument can be provided to specify an alternative scoring method.
>>> cross_val_score(svc, X_digits, y_digits, cv=k_fold,

... scoring='precision_macro')
array([0.96578289, 0.92708922, 0.96681476, 0.96362897, 0.93192644])
Cross-validation generators
KFold (n_splits, shuffle, ran- StratifiedKFold (n_splits, GroupKFold (n_splits)

dom_state) shuffle, random_state)
Splits it into K folds, trains on K-1 Same as K-Fold but preserves the Ensures that the same group is not in
and then tests on the left-out. class distribution within each fold. both testing and training sets.

ShuffleSplit (n_splits, StratifiedShuffleSplit GroupShuffleSplit

test_size, train_size, ran-
dom_state)
Generates train/test indices based Same as shuffle split but preserves the Ensures that the same group is not
on random permutation. class distribution within each iteration. in both testing and training sets.
LeaveOneGroupOut () LeavePGroupsOut (n_groups) LeaveOneOut ()

Takes a group array to group observations. Leave P groups out. Leave one observation out.
LeavePOut (p) PredefinedSplit

Leave P observations out. Generates train/test indices based on predefined splits.
Exercise
On the digits dataset, plot the cross-validation

score of a SVC estimator with an linear kernel as a function of parameter C (use a logarithmic grid of points,
from 1 to 10).
import numpy as np
from sklearn.model_selection import cross_val_score
from sklearn import datasets, svm
X = digits.data
y = digits.target
svc = svm.SVC(kernel='linear')
C_s = np.logspace(-10, 0, 10)
Solution: Cross-validation on Digits Dataset Exercise

Grid-search and cross-validated estimators
Grid-search
scikit-learn provides an object that, given data, computes the score during the fit of an estimator on a parameter grid and
chooses the parameters to maximize the cross-validation score. This object takes an estimator during the construction
and exposes an estimator API:
>>> from sklearn.model_selection import GridSearchCV, cross_val_score

>>> Cs = np.logspace(-6, -1, 10)
>>> clf = GridSearchCV(estimator=svc, param_grid=dict(C=Cs),
... n_jobs=-1)
>>> clf.fit(X_digits[:1000], y_digits[:1000])
GridSearchCV(cv=None,...
>>> clf.best_score_
0.925...
>>> clf.best_estimator_.C
0.0077...
>>> # Prediction performance on test set is not as good as on train set

>>> clf.score(X_digits[1000:], y_digits[1000:])
0.943...
By default, the GridSearchCV uses a 3-fold cross-validation. However, if it detects that a classifier is passed, rather
than a regressor, it uses a stratified 3-fold. The default will change to a 5-fold cross-validation in version 0.22.
Nested cross-validation
>>> cross_val_score(clf, X_digits, y_digits)

array([0.938..., 0.963..., 0.944...])
Two cross-validation loops are performed in parallel: one by the GridSearchCV estimator to set gamma and the
other one by cross_val_score to measure the prediction performance of the estimator. The resulting scores
are unbiased estimates of the prediction score on new data.
Warning: You cannot nest objects with parallel computing (n_jobs different than 1).
Cross-validated estimators
Cross-validation to set a parameter can be done more efficiently on an algorithm-by-algorithm basis. This is why, for
certain estimators, scikit-learn exposes Cross-validation: evaluating estimator performance estimators that set their
parameter automatically by cross-validation:
>>> from sklearn import linear_model, datasets

>>> lasso = linear_model.LassoCV(cv=3)
>>> X_diabetes = diabetes.data
>>> y_diabetes = diabetes.target
>>> lasso.fit(X_diabetes, y_diabetes)
LassoCV(alphas=None, copy_X=True, cv=3, eps=0.001, fit_intercept=True,
max_iter=1000, n_alphas=100, n_jobs=None, normalize=False,
positive=False, precompute='auto', random_state=None,

selection='cyclic', tol=0.0001, verbose=False)

>>> # The estimator chose automatically its lambda:
>>> lasso.alpha_
0.01229...
These estimators are called similarly to their counterparts, with ‘CV’ appended to their name.
Exercise
On the diabetes dataset, find the optimal regularization parameter alpha.

Bonus: How much can you trust the selection of alpha?
from sklearn import datasets

from sklearn.linear_model import LassoCV
from sklearn.linear_model import Lasso
from sklearn.model_selection import KFold
from sklearn.model_selection import GridSearchCV
diabetes = datasets.load_diabetes()
Solution: Cross-validation on diabetes Dataset Exercise
2.2.4 Unsupervised learning: seeking representations of the data
Clustering: grouping observations together
The problem solved in clustering
Given the iris dataset, if we knew that there were 3 types of iris, but did not have access to a taxonomist to label
them: we could try a clustering task: split the observations into well-separated group called clusters.
K-means clustering
Note that there exist a lot of different clustering criteria and associated algorithms. The simplest clustering algorithm
is K-means.

>>> from sklearn import cluster, datasets

>>> X_iris = iris.data
>>> y_iris = iris.target
>>> k_means = cluster.KMeans(n_clusters=3)

>>> k_means.fit(X_iris)
KMeans(algorithm='auto', copy_x=True, init='k-means++', ...
>>> print(k_means.labels_[::10])
[1 1 1 1 1 0 0 0 0 0 2 2 2 2 2]
>>> print(y_iris[::10])
[0 0 0 0 0 1 1 1 1 1 2 2 2 2 2]
Warning: There is absolutely no guarantee of recovering a ground truth. First, choosing the right number of
clusters is hard. Second, the algorithm is sensitive to initialization, and can fall into local minima, although scikit-
learn employs several tricks to mitigate this issue.
Bad initialization 8 clusters Ground truth
Don’t over-interpret clustering results
Application example: vector quantization
Clustering in general and KMeans, in particular, can be seen as a way of choosing a small number of exemplars to
compress the information. The problem is sometimes known as vector quantization. For instance, this can be used
to posterize an image:
>>> import scipy as sp
>>> try:
... face = sp.face(gray=True)
... except AttributeError:
... from scipy import misc
... face = misc.face(gray=True)
>>> X = face.reshape((-1, 1)) # We need an (n_sample, n_feature) array
>>> k_means = cluster.KMeans(n_clusters=5, n_init=1)
>>> k_means.fit(X)
KMeans(algorithm='auto', copy_x=True, init='k-means++', ...
>>> values = k_means.cluster_centers_.squeeze()
>>> labels = k_means.labels_
>>> face_compressed = np.choose(labels, values)
>>> face_compressed.shape = face.shape

Raw image K-means quantization Equal bins Image histogram
Hierarchical agglomerative clustering: Ward
A Hierarchical clustering method is a type of cluster analysis that aims to build a hierarchy of clusters. In general, the
various approaches of this technique are either:
• Agglomerative - bottom-up approaches: each observation starts in its own cluster, and clusters are iteratively
merged in such a way to minimize a linkage criterion. This approach is particularly interesting when the clus-
ters of interest are made of only a few observations. When the number of clusters is large, it is much more
computationally efficient than k-means.
• Divisive - top-down approaches: all observations start in one cluster, which is iteratively split as one moves
down the hierarchy. For estimating large numbers of clusters, this approach is both slow (due to all observations
starting as one cluster, which it splits recursively) and statistically ill-posed.
Connectivity-constrained clustering
With agglomerative clustering, it is possible to specify which samples can be clustered together by giving a connec-
tivity graph. Graphs in scikit-learn are represented by their adjacency matrix. Often, a sparse matrix is used. This
can be useful, for instance, to retrieve connected regions (sometimes also referred to as connected components) when
clustering an image:
import matplotlib.pyplot as plt
from skimage.data import coins

from skimage.transform import rescale
from sklearn.feature_extraction.image import grid_to_graph

from sklearn.cluster import AgglomerativeClustering
# #############################################################################
# Generate data

orig_coins = coins()
# Resize it to 20% of the original size to speed up the processing

# Applying a Gaussian filter for smoothing prior to down-scaling
# reduces aliasing artifacts.
smoothened_coins = gaussian_filter(orig_coins, sigma=2)
rescaled_coins = rescale(smoothened_coins, 0.2, mode="reflect")
X = np.reshape(rescaled_coins, (-1, 1))
# #############################################################################
# Define the structure A of the data. Pixels connected to their neighbors.
connectivity = grid_to_graph(*rescaled_coins.shape)
Feature agglomeration
We have seen that sparsity could be used to mitigate the curse of dimensionality, i.e an insufficient amount of ob-
servations compared to the number of features. Another approach is to merge together similar features: feature
agglomeration. This approach can be implemented by clustering in the feature direction, in other words clustering
the transposed data.

>>> images = digits.images
>>> X = np.reshape(images, (len(images), -1))
>>> connectivity = grid_to_graph(*images[0].shape)
>>> agglo = cluster.FeatureAgglomeration(connectivity=connectivity,

... n_clusters=32)
>>> agglo.fit(X)
FeatureAgglomeration(affinity='euclidean', compute_full_tree='auto',...
>>> X_reduced = agglo.transform(X)
>>> X_approx = agglo.inverse_transform(X_reduced)

>>> images_approx = np.reshape(X_approx, images.shape)
transform and inverse_transform methods
Some estimators expose a transform method, for instance to reduce the dimensionality of the dataset.

Decompositions: from a signal to components and loadings
Components and loadings
If X is our multivariate data, then the problem that we are trying to solve is to rewrite it on a different observational
basis: we want to learn loadings L and a set of components C such that X = L C. Different criteria exist to choose
the components
Principal component analysis: PCA
Principal component analysis (PCA) selects the successive components that explain the maximum variance in the
signal.
The point cloud spanned by the observations above is very flat in one direction: one of the three univariate features
can almost be exactly computed using the other two. PCA finds the directions in which the data is not flat
When used to transform data, PCA can reduce the dimensionality of the data by projecting on a principal subspace.
>>> # Create a signal with only 2 useful dimensions

>>> x1 = np.random.normal(size=100)
>>> x2 = np.random.normal(size=100)
>>> x3 = x1 + x2
>>> X = np.c_[x1, x2, x3]
>>> from sklearn import decomposition

>>> pca = decomposition.PCA()
>>> pca.fit(X)
PCA(copy=True, iterated_power='auto', n_components=None, random_state=None,
svd_solver='auto', tol=0.0, whiten=False)
>>> print(pca.explained_variance_)
[ 2.18565811e+00 1.19346747e+00 8.43026679e-32]
>>> # As we can see, only the 2 first components are useful

>>> pca.n_components = 2
>>> X_reduced = pca.fit_transform(X)
>>> X_reduced.shape
(100, 2)

Independent Component Analysis: ICA
Independent component analysis (ICA) selects components so that the distribution of their loadings carries
a maximum amount of independent information. It is able to recover non-Gaussian independent signals:
>>> # Generate sample data

>>> from scipy import signal
>>> time = np.linspace(0, 10, 2000)
>>> s1 = np.sin(2 * time) # Signal 1 : sinusoidal signal
>>> s2 = np.sign(np.sin(3 * time)) # Signal 2 : square signal
>>> s3 = signal.sawtooth(2 * np.pi * time) # Signal 3: saw tooth signal
>>> S = np.c_[s1, s2, s3]
>>> S += 0.2 * np.random.normal(size=S.shape) # Add noise
>>> S /= S.std(axis=0) # Standardize data
>>> # Mix data
>>> A = np.array([[1, 1, 1], [0.5, 2, 1], [1.5, 1, 2]]) # Mixing matrix
>>> X = np.dot(S, A.T) # Generate observations
>>> # Compute ICA

>>> ica = decomposition.FastICA()
>>> S_ = ica.fit_transform(X) # Get the estimated sources
>>> A_ = ica.mixing_.T
>>> np.allclose(X, np.dot(S_, A_) + ica.mean_)
True

2.2.5 Putting it all together
Pipelining
We have seen that some estimators can transform data and that some estimators can predict variables. We can also
create combined estimators:
import numpy as np
import pandas as pd

from sklearn.decomposition import PCA
from sklearn.linear_model import SGDClassifier
from sklearn.pipeline import Pipeline
# Define a pipeline to search for the best combination of PCA truncation

# and classifier regularization.
logistic = SGDClassifier(loss='log', penalty='l2', early_stopping=True,
max_iter=10000, tol=1e-5, random_state=0)
pca = PCA()
pipe = Pipeline(steps=[('pca', pca), ('logistic', logistic)])
X_digits = digits.data
# Parameters of pipelines can be set using ‘__’ separated parameter names:

param_grid = {
'pca__n_components': [5, 20, 30, 40, 50, 64],
'logistic__alpha': np.logspace(-4, 4, 5),

}
search = GridSearchCV(pipe, param_grid, iid=False, cv=5,
return_train_score=False)
search.fit(X_digits, y_digits)
print("Best parameter (CV score=%0.3f):" % search.best_score_)
print(search.best_params_)
# Plot the PCA spectrum

pca.fit(X_digits)
fig, (ax0, ax1) = plt.subplots(nrows=2, sharex=True, figsize=(6, 6))

ax0.plot(pca.explained_variance_ratio_, linewidth=2)
ax0.set_ylabel('PCA explained variance')
ax0.axvline(search.best_estimator_.named_steps['pca'].n_components,
Face recognition with eigenfaces
The dataset used in this example is a preprocessed excerpt of the “Labeled Faces in the Wild”, also known as LFW:
http://vis-www.cs.umass.edu/lfw/lfw-funneled.tgz (233MB)
"""
===================================================
Faces recognition example using eigenfaces and SVMs
===================================================
The dataset used in this example is a preprocessed excerpt of the

"Labeled Faces in the Wild", aka LFW_:
.. _LFW: http://vis-www.cs.umass.edu/lfw/
Expected results for the top 5 most represented people in the dataset:
================== ============ ======= ========== =======

precision recall f1-score support
================== ============ ======= ========== =======
Ariel Sharon 0.67 0.92 0.77 13
Colin Powell 0.75 0.78 0.76 60
Donald Rumsfeld 0.78 0.67 0.72 27
George W Bush 0.86 0.86 0.86 146
Gerhard Schroeder 0.76 0.76 0.76 25
Hugo Chavez 0.67 0.67 0.67 15
Tony Blair 0.81 0.69 0.75 36
avg / total 0.80 0.80 0.80 322

================== ============ ======= ========== =======
"""
from __future__ import print_function
from time import time

import logging

from sklearn.model_selection import train_test_split

from sklearn.datasets import fetch_lfw_people
from sklearn.metrics import classification_report
from sklearn.metrics import confusion_matrix
from sklearn.svm import SVC
print(__doc__)
# Display progress logs on stdout

logging.basicConfig(level=logging.INFO, format='%(asctime)s %(message)s')
# #############################################################################
# Download the data, if not already on disk and load it as numpy arrays
lfw_people = fetch_lfw_people(min_faces_per_person=70, resize=0.4)
# introspect the images arrays to find the shapes (for plotting)

n_samples, h, w = lfw_people.images.shape
# for machine learning we use the 2 data directly (as relative pixel
# positions info is ignored by this model)
X = lfw_people.data
n_features = X.shape[1]
# the label to predict is the id of the person

y = lfw_people.target
target_names = lfw_people.target_names
n_classes = target_names.shape[0]
print("Total dataset size:")

print("n_samples: %d" % n_samples)
print("n_features: %d" % n_features)
print("n_classes: %d" % n_classes)
# #############################################################################
# Split into a training set and a test set using a stratified k fold
# split into a training and testing set

X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.25, random_state=42)
# #############################################################################
# Compute a PCA (eigenfaces) on the face dataset (treated as unlabeled
# dataset): unsupervised feature extraction / dimensionality reduction
n_components = 150
print("Extracting the top %d eigenfaces from %d faces"

% (n_components, X_train.shape[0]))
t0 = time()
pca = PCA(n_components=n_components, svd_solver='randomized',
whiten=True).fit(X_train)
print("done in %0.3fs" % (time() - t0))

eigenfaces = pca.components_.reshape((n_components, h, w))
print("Projecting the input data on the eigenfaces orthonormal basis")

t0 = time()
X_train_pca = pca.transform(X_train)
X_test_pca = pca.transform(X_test)
# #############################################################################
# Train a SVM classification model
print("Fitting the classifier to the training set")

t0 = time()
param_grid = {'C': [1e3, 5e3, 1e4, 5e4, 1e5],
'gamma': [0.0001, 0.0005, 0.001, 0.005, 0.01, 0.1], }
clf = GridSearchCV(SVC(kernel='rbf', class_weight='balanced'),
param_grid, cv=5)
clf = clf.fit(X_train_pca, y_train)
print("Best estimator found by grid search:")
print(clf.best_estimator_)
# #############################################################################
# Quantitative evaluation of the model quality on the test set
print("Predicting people's names on the test set")

t0 = time()
y_pred = clf.predict(X_test_pca)
print(classification_report(y_test, y_pred, target_names=target_names))

print(confusion_matrix(y_test, y_pred, labels=range(n_classes)))
# #############################################################################
# Qualitative evaluation of the predictions using matplotlib
def plot_gallery(images, titles, h, w, n_row=3, n_col=4):

"""Helper function to plot a gallery of portraits"""
plt.figure(figsize=(1.8 * n_col, 2.4 * n_row))
plt.subplots_adjust(bottom=0, left=.01, right=.99, top=.90, hspace=.35)
for i in range(n_row * n_col):
plt.subplot(n_row, n_col, i + 1)
plt.imshow(images[i].reshape((h, w)), cmap=plt.cm.gray)
plt.title(titles[i], size=12)
plt.xticks(())
plt.yticks(())
# plot the result of the prediction on a portion of the test set
def title(y_pred, y_test, target_names, i):

pred_name = target_names[y_pred[i]].rsplit(' ', 1)[-1]
true_name = target_names[y_test[i]].rsplit(' ', 1)[-1]
return 'predicted: %s\ntrue: %s' % (pred_name, true_name)

prediction_titles = [title(y_pred, y_test, target_names, i)

for i in range(y_pred.shape[0])]
plot_gallery(X_test, prediction_titles, h, w)
# plot the gallery of the most significative eigenfaces
eigenface_titles = ["eigenface %d" % i for i in range(eigenfaces.shape[0])]

plot_gallery(eigenfaces, eigenface_titles, h, w)
plt.show()
Prediction Eigenfaces
Gerhard_Schroeder 0.91 0.75 0.82 28

Donald_Rumsfeld 0.84 0.82 0.83 33
Tony_Blair 0.65 0.82 0.73 34
Colin_Powell 0.78 0.88 0.83 58
George_W_Bush 0.93 0.86 0.90 129
avg / total 0.86 0.84 0.85 282
Open problem: Stock Market Structure
Can we predict the variation in stock prices for Google over a given time frame?
Learning a graph structure

2.2.6 Finding help
The project mailing list
If you encounter a bug with scikit-learn or something that needs clarification in the docstring or the online
documentation, please feel free to ask on the Mailing List
Q&A communities with Machine Learning practitioners
Quora.com Quora has a topic for Machine Learning related questions that also features some
interesting discussions: https://www.quora.com/topic/Machine-Learning
Stack Exchange The Stack Exchange family of sites hosts multiple subdomains for Machine
Learning questions.
– _’An excellent free online course for Machine Learning taught by Professor Andrew Ng of Stanford’: https://www.
coursera.org/learn/machine-learning
– _’Another excellent free online course that takes a more general approach to Artificial Intelligence’:
https://www.udacity.com/course/intro-to-artificial-intelligence–cs271
2.3 Working With Text Data
The goal of this guide is to explore some of the main scikit-learn tools on a single practical task: analyzing a
collection of text documents (newsgroups posts) on twenty different topics.
In this section we will see how to:
• load the file contents and the categories
• extract feature vectors suitable for machine learning
• train a linear model to perform categorization
• use a grid search strategy to find a good configuration of both the feature extraction components and the classifier
2.3.1 Tutorial setup
To get started with this tutorial, you must first install scikit-learn and all of its required dependencies.
Please refer to the installation instructions page for more information and for system-specific instructions.
The source of this tutorial can be found within your scikit-learn folder:
scikit-learn/doc/tutorial/text_analytics/
The source can also be found on Github.

The tutorial folder should contain the following sub-folders:
• *.rst files - the source of the tutorial document written with sphinx
• data - folder to put the datasets used during the tutorial
• skeletons - sample incomplete scripts for the exercises
• solutions - solutions of the exercises
2.3. Working With Text Data 163

You can already copy the skeletons into a new folder somewhere on your hard-drive named
sklearn_tut_workspace where you will edit your own files for the exercises while keeping the original
skeletons intact:
% cp -r skeletons work_directory/sklearn_tut_workspace
Machine learning algorithms need data. Go to each $TUTORIAL_HOME/data sub-folder and run the
fetch_data.py script from there (after having read them first).
For instance:
% cd $TUTORIAL_HOME/data/languages
% less fetch_data.py
% python fetch_data.py
2.3.2 Loading the 20 newsgroups dataset
The dataset is called “Twenty Newsgroups”. Here is the official description, quoted from the website:
The 20 Newsgroups data set is a collection of approximately 20,000 newsgroup documents, partitioned
(nearly) evenly across 20 different newsgroups. To the best of our knowledge, it was originally collected
by Ken Lang, probably for his paper “Newsweeder: Learning to filter netnews,” though he does not explic-
itly mention this collection. The 20 newsgroups collection has become a popular data set for experiments
in text applications of machine learning techniques, such as text classification and text clustering.
In the following we will use the built-in dataset loader for 20 newsgroups from scikit-learn. Alternatively, it is possible
to download the dataset manually from the website and use the sklearn.datasets.load_files function by
pointing it to the 20news-bydate-train sub-folder of the uncompressed archive folder.
In order to get faster execution times for this first example we will work on a partial dataset with only 4 categories out
of the 20 available in the dataset:
>>> categories = ['alt.atheism', 'soc.religion.christian',

... 'comp.graphics', 'sci.med']
We can now load the list of files matching those categories as follows:
>>> from sklearn.datasets import fetch_20newsgroups

>>> twenty_train = fetch_20newsgroups(subset='train',
... categories=categories, shuffle=True, random_state=42)
The returned dataset is a scikit-learn “bunch”: a simple holder object with fields that can be both accessed
as python dict keys or object attributes for convenience, for instance the target_names holds the list of the
requested category names:
>>> twenty_train.target_names
['alt.atheism', 'comp.graphics', 'sci.med', 'soc.religion.christian']
The files themselves are loaded in memory in the data attribute. For reference the filenames are also available:
>>> len(twenty_train.data)
2257
>>> len(twenty_train.filenames)
2257
Let’s print the first lines of the first loaded file:

>>> print("\n".join(twenty_train.data[0].split("\n")[:3]))
From: sd345@city.ac.uk (Michael Collier)
Subject: Converting images to HP LaserJet III?
Nntp-Posting-Host: hampton
>>> print(twenty_train.target_names[twenty_train.target[0]])
comp.graphics
Supervised learning algorithms will require a category label for each document in the training set. In this case the cat-
egory is the name of the newsgroup which also happens to be the name of the folder holding the individual documents.
For speed and space efficiency reasons scikit-learn loads the target attribute as an array of integers that corre-
sponds to the index of the category name in the target_names list. The category integer id of each sample is stored
in the target attribute:
>>> twenty_train.target[:10]
array([1, 1, 3, 3, 3, 3, 3, 2, 2, 2])
It is possible to get back the category names as follows:
>>> for t in twenty_train.target[:10]:

... print(twenty_train.target_names[t])
...
comp.graphics
comp.graphics
soc.religion.christian
sci.med
sci.med
sci.med
You might have noticed that the samples were shuffled randomly when we called fetch_20newsgroups(...,
shuffle=True, random_state=42): this is useful if you wish to select only a subset of samples to quickly
train a model and get a first idea of the results before re-training on the complete dataset later.
2.3.3 Extracting features from text files
In order to perform machine learning on text documents, we first need to turn the text content into numerical feature
vectors.
Bags of words
The most intuitive way to do so is to use a bags of words representation:

1. Assign a fixed integer id to each word occurring in any document of the training set (for instance by building a
dictionary from words to integer indices).
2. For each document #i, count the number of occurrences of each word w and store it in X[i, j] as the value
of feature #j where j is the index of word w in the dictionary.
The bags of words representation implies that n_features is the number of distinct words in the corpus: this
number is typically larger than 100,000.

If n_samples == 10000, storing X as a NumPy array of type float32 would require 10000 x 100000 x 4 bytes =
4GB in RAM which is barely manageable on today’s computers.
Fortunately, most values in X will be zeros since for a given document less than a few thousand distinct words will
be used. For this reason we say that bags of words are typically high-dimensional sparse datasets. We can save a lot
of memory by only storing the non-zero parts of the feature vectors in memory.
scipy.sparse matrices are data structures that do exactly this, and scikit-learn has built-in support for these
structures.
Tokenizing text with scikit-learn
Text preprocessing, tokenizing and filtering of stopwords are all included in CountVectorizer, which builds a
dictionary of features and transforms documents to feature vectors:
>>> from sklearn.feature_extraction.text import CountVectorizer

>>> count_vect = CountVectorizer()
>>> X_train_counts = count_vect.fit_transform(twenty_train.data)
>>> X_train_counts.shape
(2257, 35788)
CountVectorizer supports counts of N-grams of words or consecutive characters. Once fitted, the vectorizer has
built a dictionary of feature indices:
>>> count_vect.vocabulary_.get(u'algorithm')
4690
The index value of a word in the vocabulary is linked to its frequency in the whole training corpus.
From occurrences to frequencies
Occurrence count is a good start but there is an issue: longer documents will have higher average count values than
shorter documents, even though they might talk about the same topics.
To avoid these potential discrepancies it suffices to divide the number of occurrences of each word in a document by
the total number of words in the document: these new features are called tf for Term Frequencies.
Another refinement on top of tf is to downscale weights for words that occur in many documents in the corpus and are
therefore less informative than those that occur only in a smaller portion of the corpus.
This downscaling is called tf–idf for “Term Frequency times Inverse Document Frequency”.
Both tf and tf–idf can be computed as follows using TfidfTransformer:
>>> from sklearn.feature_extraction.text import TfidfTransformer

>>> tf_transformer = TfidfTransformer(use_idf=False).fit(X_train_counts)
>>> X_train_tf = tf_transformer.transform(X_train_counts)
>>> X_train_tf.shape
(2257, 35788)
In the above example-code, we firstly use the fit(..) method to fit our estimator to the data and secondly the
transform(..) method to transform our count-matrix to a tf-idf representation. These two steps can be com-
bined to achieve the same end result faster by skipping redundant processing. This is done through using the
fit_transform(..) method as shown below, and as mentioned in the note in the previous section:

>>> tfidf_transformer = TfidfTransformer()

>>> X_train_tfidf = tfidf_transformer.fit_transform(X_train_counts)
>>> X_train_tfidf.shape
(2257, 35788)
2.3.4 Training a classifier
Now that we have our features, we can train a classifier to try to predict the category of a post. Let’s start with a
naïve Bayes classifier, which provides a nice baseline for this task. scikit-learn includes several variants of this
classifier; the one most suitable for word counts is the multinomial variant:
>>> from sklearn.naive_bayes import MultinomialNB

>>> clf = MultinomialNB().fit(X_train_tfidf, twenty_train.target)
To try to predict the outcome on a new document we need to extract the features using almost the same feature extract-
ing chain as before. The difference is that we call transform instead of fit_transform on the transformers,
since they have already been fit to the training set:
>>> docs_new = ['God is love', 'OpenGL on the GPU is fast']

>>> X_new_counts = count_vect.transform(docs_new)
>>> X_new_tfidf = tfidf_transformer.transform(X_new_counts)
>>> predicted = clf.predict(X_new_tfidf)
>>> for doc, category in zip(docs_new, predicted):

... print('%r => %s' % (doc, twenty_train.target_names[category]))
...
'God is love' => soc.religion.christian
'OpenGL on the GPU is fast' => comp.graphics
2.3.5 Building a pipeline
In order to make the vectorizer => transformer => classifier easier to work with, scikit-learn provides a
Pipeline class that behaves like a compound classifier:
>>> from sklearn.pipeline import Pipeline

>>> text_clf = Pipeline([
... ('vect', CountVectorizer()),
... ('tfidf', TfidfTransformer()),
... ('clf', MultinomialNB()),
... ])
The names vect, tfidf and clf (classifier) are arbitrary. We will use them to perform grid search for suitable
hyperparameters below. We can now train the model with a single command:
>>> text_clf.fit(twenty_train.data, twenty_train.target)

Pipeline(...)
2.3.6 Evaluation of the performance on the test set
Evaluating the predictive accuracy of the model is equally easy:


>>> twenty_test = fetch_20newsgroups(subset='test',
... categories=categories, shuffle=True, random_state=42)
>>> docs_test = twenty_test.data
>>> predicted = text_clf.predict(docs_test)
>>> np.mean(predicted == twenty_test.target)
0.8348...
We achieved 83.5% accuracy. Let’s see if we can do better with a linear support vector machine (SVM), which is
widely regarded as one of the best text classification algorithms (although it’s also a bit slower than naïve Bayes). We
can change the learner by simply plugging a different classifier object into our pipeline:
>>> from sklearn.linear_model import SGDClassifier

>>> text_clf = Pipeline([
... ('vect', CountVectorizer()),
... ('tfidf', TfidfTransformer()),
... ('clf', SGDClassifier(loss='hinge', penalty='l2',
... alpha=1e-3, random_state=42,
... max_iter=5, tol=None)),
... ])
>>> text_clf.fit(twenty_train.data, twenty_train.target)

Pipeline(...)
>>> predicted = text_clf.predict(docs_test)
>>> np.mean(predicted == twenty_test.target)
0.9127...
We achieved 91.3% accuracy using the SVM. scikit-learn provides further utilities for more detailed perfor-
mance analysis of the results:
>>> from sklearn import metrics

>>> print(metrics.classification_report(twenty_test.target, predicted,
... target_names=twenty_test.target_names))
...
alt.atheism 0.95 0.81 0.87 319

comp.graphics 0.88 0.97 0.92 389
sci.med 0.94 0.90 0.92 396
soc.religion.christian 0.90 0.95 0.93 398
micro avg 0.91 0.91 0.91 1502

macro avg 0.92 0.91 0.91 1502
weighted avg 0.92 0.91 0.91 1502
>>> metrics.confusion_matrix(twenty_test.target, predicted)

array([[258, 11, 15, 35],
[ 4, 379, 3, 3],
[ 5, 33, 355, 3],
[ 5, 10, 4, 379]])
As expected the confusion matrix shows that posts from the newsgroups on atheism and Christianity are more often
confused for one another than with computer graphics.

2.3.7 Parameter tuning using grid search
We’ve already encountered some parameters such as use_idf in the TfidfTransformer. Classifiers tend to have
many parameters as well; e.g., MultinomialNB includes a smoothing parameter alpha and SGDClassifier
has a penalty parameter alpha and configurable loss and penalty terms in the objective function (see the module
documentation, or use the Python help function to get a description of these).
Instead of tweaking the parameters of the various components of the chain, it is possible to run an exhaustive search of
the best parameters on a grid of possible values. We try out all classifiers on either words or bigrams, with or without
idf, and with a penalty parameter of either 0.01 or 0.001 for the linear SVM:
>>> from sklearn.model_selection import GridSearchCV

>>> parameters = {
... 'vect__ngram_range': [(1, 1), (1, 2)],
... 'tfidf__use_idf': (True, False),
... 'clf__alpha': (1e-2, 1e-3),
... }
Obviously, such an exhaustive search can be expensive. If we have multiple CPU cores at our disposal, we can tell
the grid searcher to try these eight parameter combinations in parallel with the n_jobs parameter. If we give this
parameter a value of -1, grid search will detect how many cores are installed and use them all:
>>> gs_clf = GridSearchCV(text_clf, parameters, cv=5, iid=False, n_jobs=-1)
The grid search instance behaves like a normal scikit-learn model. Let’s perform the search on a smaller subset
of the training data to speed up the computation:
>>> gs_clf = gs_clf.fit(twenty_train.data[:400], twenty_train.target[:400])
The result of calling fit on a GridSearchCV object is a classifier that we can use to predict:
>>> twenty_train.target_names[gs_clf.predict(['God is love'])[0]]

'soc.religion.christian'
The object’s best_score_ and best_params_ attributes store the best mean score and the parameters setting
corresponding to that score:
>>> gs_clf.best_score_
0.9...
>>> for param_name in sorted(parameters.keys()):
... print("%s: %r" % (param_name, gs_clf.best_params_[param_name]))
...
clf__alpha: 0.001
tfidf__use_idf: True
vect__ngram_range: (1, 2)
A more detailed summary of the search is available at gs_clf.cv_results_.

The cv_results_ parameter can be easily imported into pandas as a DataFrame for further inspection.
Exercises
To do the exercises, copy the content of the ‘skeletons’ folder as a new folder named ‘workspace’:
% cp -r skeletons workspace

You can then edit the content of the workspace without fear of losing the original exercise instructions.
Then fire an ipython shell and run the work-in-progress script with:
[1] %run workspace/exercise_XX_script.py arg1 arg2 arg3
If an exception is triggered, use %debug to fire-up a post mortem ipdb session.

Refine the implementation and iterate until the exercise is solved.
For each exercise, the skeleton file provides all the necessary import statements, boilerplate code to load the
data and sample code to evaluate the predictive accuracy of the model.
2.3.8 Exercise 1: Language identification
• Write a text classification pipeline using a custom preprocessor and CharNGramAnalyzer using data from
Wikipedia articles as training set.
• Evaluate the performance on some held out test set.
ipython command line:
%run workspace/exercise_01_language_train_model.py data/languages/paragraphs/
2.3.9 Exercise 2: Sentiment Analysis on movie reviews
• Write a text classification pipeline to classify movie reviews as either positive or negative.
• Find a good set of parameters using grid search.
• Evaluate the performance on a held out test set.
ipython command line:
%run workspace/exercise_02_sentiment.py data/movie_reviews/txt_sentoken/
2.3.10 Exercise 3: CLI text classification utility
Using the results of the previous exercises and the cPickle module of the standard library, write a command line
utility that detects the language of some text provided on stdin and estimate the polarity (positive or negative) if the
text is written in English.
Bonus point if the utility is able to give a confidence level for its predictions.
2.3.11 Where to from here
Here are a few suggestions to help further your scikit-learn intuition upon the completion of this tutorial:
• Try playing around with the analyzer and token normalisation under CountVectorizer.
• If you don’t have labels, try using Clustering on your problem.
• If you have multiple labels per document, e.g categories, have a look at the Multiclass and multilabel section.
• Try using Truncated SVD for latent semantic analysis.

• Have a look at using Out-of-core Classification to learn from data that would not fit into the computer main
memory.
• Have a look at the Hashing Vectorizer as a memory efficient alternative to CountVectorizer.
2.4 Choosing the right estimator
Often the hardest part of solving a machine learning problem can be finding the right estimator for the job.
Different estimators are better suited for different types of data and different problems.
The flowchart below is designed to give users a bit of a rough guide on how to approach problems with regard to which
estimators to try on your data.
Click on any estimator in the chart below to see its documentation.
2.5 External Resources, Videos and Talks
For written tutorials, see the Tutorial section of the documentation.
2.5.1 New to Scientific Python?
For those that are still new to the scientific Python ecosystem, we highly recommend the Python Scientific Lecture
Notes. This will help you find your footing a bit and will definitely improve your scikit-learn experience. A basic
understanding of NumPy arrays is recommended to make the most of scikit-learn.
2.5.2 External Tutorials
There are several online tutorials available which are geared toward specific subject areas:
• Machine Learning for NeuroImaging in Python
• Machine Learning for Astronomical Data Analysis
2.5.3 Videos
• An introduction to scikit-learn Part I and Part II at Scipy 2013 by Gael Varoquaux, Jake Vanderplas and Olivier
Grisel. Notebooks on github.
• Introduction to scikit-learn by Gael Varoquaux at ICML 2010
A three minute video from a very early stage of scikit-learn, explaining the basic idea and approach
we are following.
• Introduction to statistical learning with scikit-learn by Gael Varoquaux at SciPy 2011
An extensive tutorial, consisting of four sessions of one hour. The tutorial covers the basics of ma-
chine learning, many algorithms and how to apply them using scikit-learn. The material correspond-
ing is now in the scikit-learn documentation section A tutorial on statistical-learning for scientific
data processing.
• Statistical Learning for Text Classification with scikit-learn and NLTK (and slides) by Olivier Grisel at PyCon
2011
2.4. Choosing the right estimator 171

Thirty minute introduction to text classification. Explains how to use NLTK and scikit-learn to solve
real-world text classification tasks and compares against cloud-based solutions.
• Introduction to Interactive Predictive Analytics in Python with scikit-learn by Olivier Grisel at PyCon 2012
3-hours long introduction to prediction tasks using scikit-learn.
• scikit-learn - Machine Learning in Python by Jake Vanderplas at the 2012 PyData workshop at Google
Interactive demonstration of some scikit-learn features. 75 minutes.
• scikit-learn tutorial by Jake Vanderplas at PyData NYC 2012
Presentation using the online tutorial, 45 minutes.
Note: Doctest Mode

The code-examples in the above tutorials are written in a python-console format. If you wish to easily execute these
examples in IPython, use:
%doctest_mode
in the IPython-console. You can then simply copy and paste the examples directly into IPython without having to
worry about removing the >>> manually.

CHAPTER
THREE
USER GUIDE
3.1 Supervised learning
3.1.1 Generalized Linear Models
The following are a set of methods intended for regression in which the target value is expected to be a linear combi-
nation of the input variables. In mathematical notion, if 𝑦ˆ is the predicted value.
𝑦ˆ(𝑤, 𝑥) = 𝑤0 + 𝑤1 𝑥1 + ... + 𝑤𝑝 𝑥𝑝
Across the module, we designate the vector 𝑤 = (𝑤1 , ..., 𝑤𝑝 ) as coef_ and 𝑤0 as intercept_.
To perform classification with generalized linear models, see Logistic regression.
Ordinary Least Squares
LinearRegression fits a linear model with coefficients 𝑤 = (𝑤1 , ..., 𝑤𝑝 ) to minimize the residual sum of squares
between the observed responses in the dataset, and the responses predicted by the linear approximation. Mathemati-
cally it solves a problem of the form:
2
min ||𝑋𝑤 − 𝑦||2
𝑤
LinearRegression will take in its fit method arrays X, y and will store the coefficients 𝑤 of the linear model
in its coef_ member:
173

>>> reg = linear_model.LinearRegression()
>>> reg.fit([[0, 0], [1, 1], [2, 2]], [0, 1, 2])
...
LinearRegression(copy_X=True, fit_intercept=True, n_jobs=None,
normalize=False)
>>> reg.coef_
array([0.5, 0.5])
However, coefficient estimates for Ordinary Least Squares rely on the independence of the model terms. When terms
are correlated and the columns of the design matrix 𝑋 have an approximate linear dependence, the design matrix
becomes close to singular and as a result, the least-squares estimate becomes highly sensitive to random errors in the
observed response, producing a large variance. This situation of multicollinearity can arise, for example, when data
are collected without an experimental design.
Examples:
• Linear Regression Example
Ordinary Least Squares Complexity
This method computes the least squares solution using a singular value decomposition of X. If X is a matrix of size (n,
p) this method has a cost of 𝑂(𝑛𝑝2 ), assuming that 𝑛 ≥ 𝑝.
Ridge Regression
Ridge regression addresses some of the problems of Ordinary Least Squares by imposing a penalty on the size of
coefficients. The ridge coefficients minimize a penalized residual sum of squares,
2 2
min ||𝑋𝑤 − 𝑦||2 + 𝛼||𝑤||2
𝑤
Here, 𝛼 ≥ 0 is a complexity parameter that controls the amount of shrinkage: the larger the value of 𝛼, the greater the
amount of shrinkage and thus the coefficients become more robust to collinearity.
As with other linear models, Ridge will take in its fit method arrays X, y and will store the coefficients 𝑤 of the
linear model in its coef_ member:
174 Chapter 3. User Guide


>>> reg = linear_model.Ridge(alpha=.5)
>>> reg.fit([[0, 0], [0, 0], [1, 1]], [0, .1, 1])
Ridge(alpha=0.5, copy_X=True, fit_intercept=True, max_iter=None,
normalize=False, random_state=None, solver='auto', tol=0.001)
>>> reg.coef_
array([0.34545455, 0.34545455])
>>> reg.intercept_
0.13636...
Examples:
• Plot Ridge coefficients as a function of the regularization

• Classification of text documents using sparse features
Ridge Complexity
This method has the same order of complexity than an Ordinary Least Squares.
Setting the regularization parameter: generalized Cross-Validation
RidgeCV implements ridge regression with built-in cross-validation of the alpha parameter. The object works in
the same way as GridSearchCV except that it defaults to Generalized Cross-Validation (GCV), an efficient form of
leave-one-out cross-validation:

>>> reg = linear_model.RidgeCV(alphas=[0.1, 1.0, 10.0], cv=3)
>>> reg.fit([[0, 0], [0, 0], [1, 1]], [0, .1, 1])
RidgeCV(alphas=[0.1, 1.0, 10.0], cv=3, fit_intercept=True, scoring=None,
normalize=False)
>>> reg.alpha_
0.1
References
• “Notes on Regularized Least Squares”, Rifkin & Lippert (technical report, course slides).
Lasso
The Lasso is a linear model that estimates sparse coefficients. It is useful in some contexts due to its tendency
to prefer solutions with fewer parameter values, effectively reducing the number of variables upon which the given
solution is dependent. For this reason, the Lasso and its variants are fundamental to the field of compressed sensing.
Under certain conditions, it can recover the exact set of non-zero weights (see Compressive sensing: tomography
reconstruction with L1 prior (Lasso)).
Mathematically, it consists of a linear model trained with ℓ1 prior as regularizer. The objective function to minimize
3.1. Supervised learning 175

is:
1
min ||𝑋𝑤 − 𝑦||22 + 𝛼||𝑤||1
𝑤 2𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠
The lasso estimate thus solves the minimization of the least-squares penalty with 𝛼||𝑤||1 added, where 𝛼 is a constant
and ||𝑤||1 is the ℓ1 -norm of the parameter vector.
The implementation in the class Lasso uses coordinate descent as the algorithm to fit the coefficients. See Least
Angle Regression for another implementation:

>>> reg = linear_model.Lasso(alpha=0.1)
>>> reg.fit([[0, 0], [1, 1]], [0, 1])
Lasso(alpha=0.1, copy_X=True, fit_intercept=True, max_iter=1000,
normalize=False, positive=False, precompute=False, random_state=None,
selection='cyclic', tol=0.0001, warm_start=False)
>>> reg.predict([[1, 1]])
array([0.8])
Also useful for lower-level tasks is the function lasso_path that computes the coefficients along the full path of
possible values.
Examples:
• Lasso and Elastic Net for Sparse Signals

• Compressive sensing: tomography reconstruction with L1 prior (Lasso)
Note: Feature selection with Lasso

As the Lasso regression yields sparse models, it can thus be used to perform feature selection, as detailed in L1-based
feature selection.
The following two references explain the iterations used in the coordinate descent solver of scikit-learn, as well as the
duality gap computation used for convergence control.
References
• “Regularization Path For Generalized linear Models by Coordinate Descent”, Friedman, Hastie & Tibshirani,
J Stat Softw, 2010 (Paper).
• “An Interior-Point Method for Large-Scale L1-Regularized Least Squares,” S. J. Kim, K. Koh, M. Lustig, S.
Boyd and D. Gorinevsky, in IEEE Journal of Selected Topics in Signal Processing, 2007 (Paper)
Setting regularization parameter
The alpha parameter controls the degree of sparsity of the coefficients estimated.

Using cross-validation
scikit-learn exposes objects that set the Lasso alpha parameter by cross-validation: LassoCV and LassoLarsCV .
LassoLarsCV is based on the Least Angle Regression algorithm explained below.
For high-dimensional datasets with many collinear regressors, LassoCV is most often preferable. However,
LassoLarsCV has the advantage of exploring more relevant values of alpha parameter, and if the number of samples
is very small compared to the number of features, it is often faster than LassoCV .
Information-criteria based model selection
Alternatively, the estimator LassoLarsIC proposes to use the Akaike information criterion (AIC) and the Bayes
Information criterion (BIC). It is a computationally cheaper alternative to find the optimal value of alpha as the regu-
larization path is computed only once instead of k+1 times when using k-fold cross-validation. However, such criteria
needs a proper estimation of the degrees of freedom of the solution, are derived for large samples (asymptotic results)
and assume the model is correct, i.e. that the data are actually generated by this model. They also tend to break when
the problem is badly conditioned (more features than samples).
Examples:
• Lasso model selection: Cross-Validation / AIC / BIC

Comparison with the regularization parameter of SVM
The equivalence between alpha and the regularization parameter of SVM, C is given by alpha = 1 / C or
alpha = 1 / (n_samples * C), depending on the estimator and the exact objective function optimized by
the model.
Multi-task Lasso
The MultiTaskLasso is a linear model that estimates sparse coefficients for multiple regression problems jointly:
y is a 2D array, of shape (n_samples, n_tasks). The constraint is that the selected features are the same for all
the regression problems, also called tasks.
The following figure compares the location of the non-zeros in W obtained with a simple Lasso or a MultiTaskLasso.
The Lasso estimates yields scattered non-zeros while the non-zeros of the MultiTaskLasso are full columns.
Fitting a time-series model, imposing that any active feature be active at all times.
Examples:
• Joint feature selection with multi-task Lasso
Mathematically, it consists of a linear model trained with a mixed ℓ1 ℓ2 prior as regularizer. The objective function to

minimize is:
1
min ||𝑋𝑊 − 𝑌 ||2𝐹 𝑟𝑜 + 𝛼||𝑊 ||21
𝑤 2𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠
where 𝐹 𝑟𝑜 indicates the Frobenius norm:

√︃∑︁
||𝐴||𝐹 𝑟𝑜 = 𝑎2𝑖𝑗
𝑖𝑗
and ℓ1 ℓ2 reads:
∑︁ √︃∑︁
||𝐴||21 = 𝑎2𝑖𝑗
𝑖 𝑗
The implementation in the class MultiTaskLasso uses coordinate descent as the algorithm to fit the coefficients.
Elastic Net
ElasticNet is a linear regression model trained with L1 and L2 prior as regularizer. This combination allows for
learning a sparse model where few of the weights are non-zero like Lasso, while still maintaining the regularization
properties of Ridge. We control the convex combination of L1 and L2 using the l1_ratio parameter.
Elastic-net is useful when there are multiple features which are correlated with one another. Lasso is likely to pick one
of these at random, while elastic-net is likely to pick both.
A practical advantage of trading-off between Lasso and Ridge is it allows Elastic-Net to inherit some of Ridge’s
stability under rotation.
The objective function to minimize is in this case
1 𝛼(1 − 𝜌)
min ||𝑋𝑤 − 𝑦||22 + 𝛼𝜌||𝑤||1 + ||𝑤||22
𝑤 2𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠 2
The class ElasticNetCV can be used to set the parameters alpha (𝛼) and l1_ratio (𝜌) by cross-validation.
Examples:


• Lasso and Elastic Net
The following two references explain the iterations used in the coordinate descent solver of scikit-learn, as well as the
duality gap computation used for convergence control.
References
• “Regularization Path For Generalized linear Models by Coordinate Descent”, Friedman, Hastie & Tibshirani,
J Stat Softw, 2010 (Paper).
• “An Interior-Point Method for Large-Scale L1-Regularized Least Squares,” S. J. Kim, K. Koh, M. Lustig, S.
Boyd and D. Gorinevsky, in IEEE Journal of Selected Topics in Signal Processing, 2007 (Paper)
Multi-task Elastic Net
The MultiTaskElasticNet is an elastic-net model that estimates sparse coefficients for multiple regression prob-
lems jointly: Y is a 2D array, of shape (n_samples, n_tasks). The constraint is that the selected features are
the same for all the regression problems, also called tasks.
Mathematically, it consists of a linear model trained with a mixed ℓ1 ℓ2 prior and ℓ2 prior as regularizer. The objective
function to minimize is:
1 𝛼(1 − 𝜌)
min ||𝑋𝑊 − 𝑌 ||2𝐹 𝑟𝑜 + 𝛼𝜌||𝑊 ||21 + ||𝑊 ||2𝐹 𝑟𝑜
𝑊 2𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠 2
The implementation in the class MultiTaskElasticNet uses coordinate descent as the algorithm to fit the coef-
ficients.
The class MultiTaskElasticNetCV can be used to set the parameters alpha (𝛼) and l1_ratio (𝜌) by cross-
validation.
Least Angle Regression
Least-angle regression (LARS) is a regression algorithm for high-dimensional data, developed by Bradley Efron,
Trevor Hastie, Iain Johnstone and Robert Tibshirani. LARS is similar to forward stepwise regression. At each step,
it finds the predictor most correlated with the response. When there are multiple predictors having equal correlation,
instead of continuing along the same predictor, it proceeds in a direction equiangular between the predictors.
The advantages of LARS are:
• It is numerically efficient in contexts where p >> n (i.e., when the number of dimensions is significantly greater
than the number of points)
• It is computationally just as fast as forward selection and has the same order of complexity as an ordinary least
squares.
• It produces a full piecewise linear solution path, which is useful in cross-validation or similar attempts to tune
the model.
• If two variables are almost equally correlated with the response, then their coefficients should increase at ap-
proximately the same rate. The algorithm thus behaves as intuition would expect, and also is more stable.
• It is easily modified to produce solutions for other estimators, like the Lasso.
The disadvantages of the LARS method include:

• Because LARS is based upon an iterative refitting of the residuals, it would appear to be especially sensitive to
the effects of noise. This problem is discussed in detail by Weisberg in the discussion section of the Efron et al.
(2004) Annals of Statistics article.
The LARS model can be used using estimator Lars, or its low-level implementation lars_path.
LARS Lasso
LassoLars is a lasso model implemented using the LARS algorithm, and unlike the implementation based on
coordinate_descent, this yields the exact solution, which is piecewise linear as a function of the norm of its coefficients.

>>> reg = linear_model.LassoLars(alpha=.1)
>>> reg.fit([[0, 0], [1, 1]], [0, 1])
LassoLars(alpha=0.1, copy_X=True, eps=..., fit_intercept=True,
fit_path=True, max_iter=500, normalize=True, positive=False,
precompute='auto', verbose=False)
>>> reg.coef_
array([0.717157..., 0. ])
Examples:
• Lasso path using LARS
The Lars algorithm provides the full path of the coefficients along the regularization parameter almost for free, thus a
common operation consist of retrieving the path with function lars_path
Mathematical formulation
The algorithm is similar to forward stepwise regression, but instead of including variables at each step, the estimated
parameters are increased in a direction equiangular to each one’s correlations with the residual.
Instead of giving a vector result, the LARS solution consists of a curve denoting the solution for each value of the
L1 norm of the parameter vector. The full coefficients path is stored in the array coef_path_, which has size
(n_features, max_features+1). The first column is always zero.

References:
• Original Algorithm is detailed in the paper Least Angle Regression by Hastie et al.
Orthogonal Matching Pursuit (OMP)
OrthogonalMatchingPursuit and orthogonal_mp implements the OMP algorithm for approximating the
fit of a linear model with constraints imposed on the number of non-zero coefficients (ie. the L 0 pseudo-norm).
Being a forward feature selection method like Least Angle Regression, orthogonal matching pursuit can approximate
the optimum solution vector with a fixed number of non-zero elements:
arg min ||𝑦 − 𝑋𝛾||22 subject to ||𝛾||0 ≤ 𝑛𝑛𝑜𝑛𝑧𝑒𝑟𝑜_𝑐𝑜𝑒𝑓 𝑠

𝛾
Alternatively, orthogonal matching pursuit can target a specific error instead of a specific number of non-zero coeffi-
cients. This can be expressed as:
arg min ||𝛾||0 subject to ||𝑦 − 𝑋𝛾||22 ≤ tol

𝛾
OMP is based on a greedy algorithm that includes at each step the atom most highly correlated with the current
residual. It is similar to the simpler matching pursuit (MP) method, but better in that at each iteration, the residual is
recomputed using an orthogonal projection on the space of the previously chosen dictionary elements.
Examples:
• Orthogonal Matching Pursuit
References:
• http://www.cs.technion.ac.il/~ronrubin/Publications/KSVD-OMP-v2.pdf
• Matching pursuits with time-frequency dictionaries, S. G. Mallat, Z. Zhang,
Bayesian Regression
Bayesian regression techniques can be used to include regularization parameters in the estimation procedure: the
regularization parameter is not set in a hard sense but tuned to the data at hand.
This can be done by introducing uninformative priors over the hyper parameters of the model. The ℓ2 regularization
used in Ridge Regression is equivalent to finding a maximum a posteriori estimation under a Gaussian prior over the
parameters 𝑤 with precision 𝜆−1 . Instead of setting lambda manually, it is possible to treat it as a random variable to
be estimated from the data.
To obtain a fully probabilistic model, the output 𝑦 is assumed to be Gaussian distributed around 𝑋𝑤:
𝑝(𝑦|𝑋, 𝑤, 𝛼) = 𝒩 (𝑦|𝑋𝑤, 𝛼)
Alpha is again treated as a random variable that is to be estimated from the data.
The advantages of Bayesian Regression are:

• It adapts to the data at hand.

• It can be used to include regularization parameters in the estimation procedure.
The disadvantages of Bayesian regression include:
• Inference of the model can be time consuming.
References
• A good introduction to Bayesian methods is given in C. Bishop: Pattern Recognition and Machine learning
• Original Algorithm is detailed in the book Bayesian learning for neural networks by Radford M. Neal
Bayesian Ridge Regression
BayesianRidge estimates a probabilistic model of the regression problem as described above. The prior for the
parameter 𝑤 is given by a spherical Gaussian:
𝑝(𝑤|𝜆) = 𝒩 (𝑤|0, 𝜆−1 I𝑝 )
The priors over 𝛼 and 𝜆 are chosen to be gamma distributions, the conjugate prior for the precision of the Gaussian.
The resulting model is called Bayesian Ridge Regression, and is similar to the classical Ridge. The parameters
𝑤, 𝛼 and 𝜆 are estimated jointly during the fit of the model. The remaining hyperparameters are the parameters of
the gamma priors over 𝛼 and 𝜆. These are usually chosen to be non-informative. The parameters are estimated by
maximizing the marginal log likelihood.
By default 𝛼1 = 𝛼2 = 𝜆1 = 𝜆2 = 10−6 .
Bayesian Ridge Regression is used for regression:

>>> X = [[0., 0.], [1., 1.], [2., 2.], [3., 3.]]
>>> Y = [0., 1., 2., 3.]
>>> reg = linear_model.BayesianRidge()
>>> reg.fit(X, Y)
BayesianRidge(alpha_1=1e-06, alpha_2=1e-06, compute_score=False, copy_X=True,

fit_intercept=True, lambda_1=1e-06, lambda_2=1e-06, n_iter=300,

normalize=False, tol=0.001, verbose=False)
After being fitted, the model can then be used to predict new values:
>>> reg.predict([[1, 0.]])

array([0.50000013])
The weights 𝑤 of the model can be access:
>>> reg.coef_
array([0.49999993, 0.49999993])
Due to the Bayesian framework, the weights found are slightly different to the ones found by Ordinary Least Squares.
However, Bayesian Ridge Regression is more robust to ill-posed problem.
Examples:
• Bayesian Ridge Regression
References
• More details can be found in the article Bayesian Interpolation by MacKay, David J. C.
Automatic Relevance Determination - ARD
ARDRegression is very similar to Bayesian Ridge Regression, but can lead to sparser weights 𝑤12 .
ARDRegression poses a different prior over 𝑤, by dropping the assumption of the Gaussian being spherical.
Instead, the distribution over 𝑤 is assumed to be an axis-parallel, elliptical Gaussian distribution.
This means each weight 𝑤𝑖 is drawn from a Gaussian distribution, centered on zero and with a precision 𝜆𝑖 :
𝑝(𝑤|𝜆) = 𝒩 (𝑤|0, 𝐴−1 )
with 𝑑𝑖𝑎𝑔 (𝐴) = 𝜆 = {𝜆1 , ..., 𝜆𝑝 }.

In contrast to Bayesian Ridge Regression, each coordinate of 𝑤𝑖 has its own standard deviation 𝜆𝑖 . The prior over all
𝜆𝑖 is chosen to be the same gamma distribution given by hyperparameters 𝜆1 and 𝜆2 .
ARD is also known in the literature as Sparse Bayesian Learning and Relevance Vector Machine34 .
Examples:
• Automatic Relevance Determination Regression (ARD)
1 Christopher M. Bishop: Pattern Recognition and Machine Learning, Chapter 7.2.1

2 David Wipf and Srikantan Nagarajan: A new view of automatic relevance determination
3 Michael E. Tipping: Sparse Bayesian Learning and the Relevance Vector Machine
4 Tristan Fletcher: Relevance Vector Machines explained

References:
Logistic regression
Logistic regression, despite its name, is a linear model for classification rather than regression. Logistic regression is
also known in the literature as logit regression, maximum-entropy classification (MaxEnt) or the log-linear classifier.
In this model, the probabilities describing the possible outcomes of a single trial are modeled using a logistic function.
The implementation of logistic regression in scikit-learn can be accessed from class LogisticRegression. This
implementation can fit binary, One-vs- Rest, or multinomial logistic regression with optional L2 or L1 regularization.
As an optimization problem, binary class L2 penalized logistic regression minimizes the following cost function:
𝑛
1 ∑︁
min 𝑤𝑇 𝑤 + 𝐶 log(exp(−𝑦𝑖 (𝑋𝑖𝑇 𝑤 + 𝑐)) + 1).
𝑤,𝑐 2
𝑖=1
Similarly, L1 regularized logistic regression solves the following optimization problem

𝑛
∑︁
min ‖𝑤‖1 + 𝐶 log(exp(−𝑦𝑖 (𝑋𝑖𝑇 𝑤 + 𝑐)) + 1).
𝑤,𝑐
𝑖=1
Note that, in this notation, it’s assumed that the observation 𝑦𝑖 takes values in the set −1, 1 at trial 𝑖.
The solvers implemented in the class LogisticRegression are “liblinear”, “newton-cg”, “lbfgs”, “sag” and
“saga”:
The solver “liblinear” uses a coordinate descent (CD) algorithm, and relies on the excellent C++ LIBLINEAR library,
which is shipped with scikit-learn. However, the CD algorithm implemented in liblinear cannot learn a true multino-
mial (multiclass) model; instead, the optimization problem is decomposed in a “one-vs-rest” fashion so separate binary
classifiers are trained for all classes. This happens under the hood, so LogisticRegression instances using this
solver behave as multiclass classifiers. For L1 penalization sklearn.svm.l1_min_c allows to calculate the lower
bound for C in order to get a non “null” (all feature weights to zero) model.
The “lbfgs”, “sag” and “newton-cg” solvers only support L2 penalization and are found to converge faster for some
high dimensional data. Setting multi_class to “multinomial” with these solvers learns a true multinomial logistic
regression model5 , which means that its probability estimates should be better calibrated than the default “one-vs-
rest” setting.
5 Christopher M. Bishop: Pattern Recognition and Machine Learning, Chapter 4.3.4

The “sag” solver uses a Stochastic Average Gradient descent6 . It is faster than other solvers for large datasets, when
both the number of samples and the number of features are large.
The “saga” solver7 is a variant of “sag” that also supports the non-smooth penalty=”l1” option. This is therefore the
solver of choice for sparse multinomial logistic regression.
The “lbfgs” is an optimization algorithm that approximates the Broyden–Fletcher–Goldfarb–Shanno algorithm8 ,
which belongs to quasi-Newton methods. The “lbfgs” solver is recommended for use for small data-sets but for
larger datasets its performance suffers.9
The following table summarizes the penalties supported by each solver:
Solvers
Penalties ‘liblinear’ ‘lbfgs’ ‘newton-cg’ ‘sag’ ‘saga’
Multinomial + L2 penalty no yes yes yes yes
OVR + L2 penalty yes yes yes yes yes
Multinomial + L1 penalty no no no no yes
OVR + L1 penalty yes no no no yes
Behaviors
Penalize the intercept (bad) yes no no no no

Faster for large datasets no no no yes yes
Robust to unscaled datasets yes yes yes no no
The “lbfgs” solver is used by default for its robustness. For large datasets the “saga” solver is usually faster. For large
dataset, you may also consider using SGDClassifier with ‘log’ loss, which might be even faster but require more
tuning.
Examples:
• L1 Penalty and Sparsity in Logistic Regression

• Regularization path of L1- Logistic Regression
• Plot multinomial and One-vs-Rest Logistic Regression
• Multiclass sparse logisitic regression on newgroups20
• MNIST classfification using multinomial logistic + L1
Differences from liblinear:
There might be a difference in the scores obtained between LogisticRegression with solver=liblinear
or LinearSVC and the external liblinear library directly, when fit_intercept=False and the fit coef_
(or) the data to be predicted are zeroes. This is because for the sample(s) with decision_function zero,
LogisticRegression and LinearSVC predict the negative class, while liblinear predicts the positive class.
Note that a model with fit_intercept=False and having many samples with decision_function zero,
is likely to be a underfit, bad model and you are advised to set fit_intercept=True and increase the inter-
cept_scaling.
6Mark Schmidt, Nicolas Le Roux, and Francis Bach: Minimizing Finite Sums with the Stochastic Average Gradient.
7Aaron Defazio, Francis Bach, Simon Lacoste-Julien: SAGA: A Fast Incremental Gradient Method With Support for Non-Strongly Convex
Composite Objectives.
8 https://en.wikipedia.org/wiki/Broyden%E2%80%93Fletcher%E2%80%93Goldfarb%E2%80%93Shanno_algorithm
9 “Performance Evaluation of Lbfgs vs other solvers”

Note: Feature selection with sparse logistic regression

A logistic regression with L1 penalty yields sparse models, and can thus be used to perform feature selection, as
detailed in L1-based feature selection.
LogisticRegressionCV implements Logistic Regression with builtin cross-validation to find out the optimal C
parameter. “newton-cg”, “sag”, “saga” and “lbfgs” solvers are found to be faster for high-dimensional dense data, due
to warm-starting. For the multiclass case, if multi_class option is set to “ovr”, an optimal C is obtained for each class
and if the multi_class option is set to “multinomial”, an optimal C is obtained by minimizing the cross-entropy loss.
References:
Stochastic Gradient Descent - SGD
Stochastic gradient descent is a simple yet very efficient approach to fit linear models. It is particularly useful when the
number of samples (and the number of features) is very large. The partial_fit method allows online/out-of-core
learning.
The classes SGDClassifier and SGDRegressor provide functionality to fit linear models for classifica-
tion and regression using different (convex) loss functions and different penalties. E.g., with loss="log",
SGDClassifier fits a logistic regression model, while with loss="hinge" it fits a linear support vector ma-
chine (SVM).
References
• Stochastic Gradient Descent
Perceptron
The Perceptron is another simple classification algorithm suitable for large scale learning. By default:
• It does not require a learning rate.
• It is not regularized (penalized).
• It updates its model only on mistakes.
The last characteristic implies that the Perceptron is slightly faster to train than SGD with the hinge loss and that the
resulting models are sparser.
Passive Aggressive Algorithms
The passive-aggressive algorithms are a family of algorithms for large-scale learning. They are similar to the Per-
ceptron in that they do not require a learning rate. However, contrary to the Perceptron, they include a regularization
parameter C.
For classification, PassiveAggressiveClassifier can be used with loss='hinge' (PA-I) or
loss='squared_hinge' (PA-II). For regression, PassiveAggressiveRegressor can be used with
loss='epsilon_insensitive' (PA-I) or loss='squared_epsilon_insensitive' (PA-II).

References:
• “Online Passive-Aggressive Algorithms” K. Crammer, O. Dekel, J. Keshat, S. Shalev-Shwartz, Y. Singer -

JMLR 7 (2006)
Robustness regression: outliers and modeling errors
Robust regression is interested in fitting a regression model in the presence of corrupt data: either outliers, or error in
the model.
Different scenario and useful concepts
There are different things to keep in mind when dealing with data corrupted by outliers:
• Outliers in X or in y?
Outliers in the y direction Outliers in the X direction
• Fraction of outliers versus amplitude of error

The number of outlying points matters, but also how much they are outliers.
Small outliers Large outliers
An important notion of robust fitting is that of breakdown point: the fraction of data that can be outlying for the fit to
start missing the inlying data.
Note that in general, robust fitting in high-dimensional setting (large n_features) is very hard. The robust models here
will probably not work in these settings.
Trade-offs: which estimator?
Scikit-learn provides 3 robust regression estimators: RANSAC, Theil Sen and HuberRegressor
• HuberRegressor should be faster than RANSAC and Theil Sen unless the number of samples are
very large, i.e n_samples >> n_features. This is because RANSAC and Theil Sen fit on
smaller subsets of the data. However, both Theil Sen and RANSAC are unlikely to be as robust as
HuberRegressor for the default parameters.
• RANSAC is faster than Theil Sen and scales much better with the number of samples
• RANSAC will deal better with large outliers in the y direction (most common situation)
• Theil Sen will cope better with medium-size outliers in the X direction, but this property will
disappear in large dimensional settings.
When in doubt, use RANSAC
RANSAC: RANdom SAmple Consensus
RANSAC (RANdom SAmple Consensus) fits a model from random subsets of inliers from the complete data set.
RANSAC is a non-deterministic algorithm producing only a reasonable result with a certain probability, which is de-
pendent on the number of iterations (see max_trials parameter). It is typically used for linear and non-linear regression
problems and is especially popular in the fields of photogrammetric computer vision.
The algorithm splits the complete input sample data into a set of inliers, which may be subject to noise, and outliers,
which are e.g. caused by erroneous measurements or invalid hypotheses about the data. The resulting model is then
estimated only from the determined inliers.

Details of the algorithm
Each iteration performs the following steps:

1. Select min_samples random samples from the original data and check whether the set of data is valid (see
is_data_valid).
2. Fit a model to the random subset (base_estimator.fit) and check whether the estimated model is valid
(see is_model_valid).
3. Classify all data as inliers or outliers by calculating the residuals to the estimated model (base_estimator.
predict(X) - y) - all data samples with absolute residuals smaller than the residual_threshold are
considered as inliers.
4. Save fitted model as best model if number of inlier samples is maximal. In case the current estimated model has
the same number of inliers, it is only considered as the best model if it has better score.
These steps are performed either a maximum number of times (max_trials) or until one of the special stop criteria
are met (see stop_n_inliers and stop_score). The final model is estimated using all inlier samples (consensus
set) of the previously determined best model.
The is_data_valid and is_model_valid functions allow to identify and reject degenerate combinations of
random sub-samples. If the estimated model is not needed for identifying degenerate cases, is_data_valid should
be used as it is called prior to fitting the model and thus leading to better computational performance.
Examples:
• Robust linear model estimation using RANSAC

• Robust linear estimator fitting
References:
• https://en.wikipedia.org/wiki/RANSAC
• “Random Sample Consensus: A Paradigm for Model Fitting with Applications to Image Analysis and Auto-
mated Cartography” Martin A. Fischler and Robert C. Bolles - SRI International (1981)
• “Performance Evaluation of RANSAC Family” Sunglok Choi, Taemin Kim and Wonpil Yu - BMVC (2009)

Theil-Sen estimator: generalized-median-based estimator
The TheilSenRegressor estimator uses a generalization of the median in multiple dimensions. It is thus robust
to multivariate outliers. Note however that the robustness of the estimator decreases quickly with the dimensionality of
the problem. It looses its robustness properties and becomes no better than an ordinary least squares in high dimension.
Examples:
• Theil-Sen Regression
References:
• https://en.wikipedia.org/wiki/Theil%E2%80%93Sen_estimator
Theoretical considerations
TheilSenRegressor is comparable to the Ordinary Least Squares (OLS) in terms of asymptotic efficiency and as
an unbiased estimator. In contrast to OLS, Theil-Sen is a non-parametric method which means it makes no assumption
about the underlying distribution of the data. Since Theil-Sen is a median-based estimator, it is more robust against
corrupted data aka outliers. In univariate setting, Theil-Sen has a breakdown point of about 29.3% in case of a simple
linear regression which means that it can tolerate arbitrary corrupted data of up to 29.3%.
The implementation of TheilSenRegressor in scikit-learn follows a generalization to a multivariate linear re-

gression model10 using the spatial median which is a generalization of the median to multiple dimensions11 .
In terms of time and space complexity, Theil-Sen scales according to
(︂ )︂
𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠
𝑛𝑠𝑢𝑏𝑠𝑎𝑚𝑝𝑙𝑒𝑠
10 Xin Dang, Hanxiang Peng, Xueqin Wang and Heping Zhang: Theil-Sen Estimators in a Multiple Linear Regression Model.
11
20. Kärkkäinen and S. Äyrämö: On Computation of Spatial Median for Robust Data Mining.

which makes it infeasible to be applied exhaustively to problems with a large number of samples and features. There-
fore, the magnitude of a subpopulation can be chosen to limit the time and space complexity by considering only a
random subset of all possible combinations.
Examples:
References:
Huber Regression
The HuberRegressor is different to Ridge because it applies a linear loss to samples that are classified as outliers.
A sample is classified as an inlier if the absolute error of that sample is lesser than a certain threshold. It differs from
TheilSenRegressor and RANSACRegressor because it does not ignore the effect of the outliers but gives a
lesser weight to them.
The loss function that HuberRegressor minimizes is given by

𝑛 (︂ (︂ )︂ )︂
∑︁ 𝑋𝑖 𝑤 − 𝑦𝑖 2
min 𝜎 + 𝐻𝜖 𝜎 + 𝛼||𝑤||2
𝑤,𝜎
𝑖=1
𝜎
where
{︃
𝑧2, if |𝑧| < 𝜖,
𝐻𝜖 (𝑧) =
2𝜖|𝑧| − 𝜖2 , otherwise
It is advised to set the parameter epsilon to 1.35 to achieve 95% statistical efficiency.
Notes
The HuberRegressor differs from using SGDRegressor with loss set to huber in the following ways.

• HuberRegressor is scaling invariant. Once epsilon is set, scaling X and y down or up by different values
would produce the same robustness to outliers as before. as compared to SGDRegressor where epsilon
has to be set again when X and y are scaled.
• HuberRegressor should be more efficient to use on data with small number of samples while
SGDRegressor needs a number of passes on the training data to produce the same robustness.
Examples:
• HuberRegressor vs Ridge on dataset with strong outliers
References:
• Peter J. Huber, Elvezio M. Ronchetti: Robust Statistics, Concomitant scale estimates, pg 172
Also, this estimator is different from the R implementation of Robust Regression (http://www.ats.ucla.edu/stat/r/dae/
rreg.htm) because the R implementation does a weighted least squares implementation with weights given to each
sample on the basis of how much the residual is greater than a certain threshold.
Polynomial regression: extending linear models with basis functions
One common pattern within machine learning is to use linear models trained on nonlinear functions of the data. This
approach maintains the generally fast performance of linear methods, while allowing them to fit a much wider range
of data.
For example, a simple linear regression can be extended by constructing polynomial features from the coefficients.
In the standard linear regression case, you might have a model that looks like this for two-dimensional data:
𝑦ˆ(𝑤, 𝑥) = 𝑤0 + 𝑤1 𝑥1 + 𝑤2 𝑥2
If we want to fit a paraboloid to the data instead of a plane, we can combine the features in second-order polynomials,
so that the model looks like this:
𝑦ˆ(𝑤, 𝑥) = 𝑤0 + 𝑤1 𝑥1 + 𝑤2 𝑥2 + 𝑤3 𝑥1 𝑥2 + 𝑤4 𝑥21 + 𝑤5 𝑥22
The (sometimes surprising) observation is that this is still a linear model: to see this, imagine creating a new variable
𝑧 = [𝑥1 , 𝑥2 , 𝑥1 𝑥2 , 𝑥21 , 𝑥22 ]
With this re-labeling of the data, our problem can be written
𝑦ˆ(𝑤, 𝑥) = 𝑤0 + 𝑤1 𝑧1 + 𝑤2 𝑧2 + 𝑤3 𝑧3 + 𝑤4 𝑧4 + 𝑤5 𝑧5
We see that the resulting polynomial regression is in the same class of linear models we’d considered above (i.e. the
model is linear in 𝑤) and can be solved by the same techniques. By considering linear fits within a higher-dimensional
space built with these basis functions, the model has the flexibility to fit a much broader range of data.
Here is an example of applying this idea to one-dimensional data, using polynomial features of varying degrees:
This figure is created using the PolynomialFeatures preprocessor. This preprocessor transforms an input data
matrix into a new data matrix of a given degree. It can be used as follows:

>>> from sklearn.preprocessing import PolynomialFeatures

>>> X = np.arange(6).reshape(3, 2)
>>> X
array([[0, 1],
[2, 3],
[4, 5]])
>>> poly = PolynomialFeatures(degree=2)
>>> poly.fit_transform(X)
array([[ 1., 0., 1., 0., 0., 1.],
[ 1., 2., 3., 4., 6., 9.],
[ 1., 4., 5., 16., 20., 25.]])
The features of X have been transformed from [𝑥1 , 𝑥2 ] to [1, 𝑥1 , 𝑥2 , 𝑥21 , 𝑥1 𝑥2 , 𝑥22 ], and can now be used within any
linear model.
This sort of preprocessing can be streamlined with the Pipeline tools. A single object representing a simple polynomial
regression can be created and used as follows:

>>> from sklearn.linear_model import LinearRegression
>>> model = Pipeline([('poly', PolynomialFeatures(degree=3)),
... ('linear', LinearRegression(fit_intercept=False))])
>>> # fit to an order-3 polynomial data
>>> x = np.arange(5)
>>> y = 3 - 2 * x + x ** 2 - x ** 3
>>> model = model.fit(x[:, np.newaxis], y)
>>> model.named_steps['linear'].coef_
array([ 3., -2., 1., -1.])
The linear model trained on polynomial features is able to exactly recover the input polynomial coefficients.
In some cases it’s not necessary to include higher powers of any single feature, but only the so-called interaction
features that multiply together at most 𝑑 distinct features. These can be gotten from PolynomialFeatures with
the setting interaction_only=True.
For example, when dealing with boolean features, 𝑥𝑛𝑖 = 𝑥𝑖 for all 𝑛 and is therefore useless; but 𝑥𝑖 𝑥𝑗 represents the
conjunction of two booleans. This way, we can solve the XOR problem with a linear classifier:

>>> from sklearn.linear_model import Perceptron

>>> X = np.array([[0, 0], [0, 1], [1, 0], [1, 1]])
>>> y = X[:, 0] ^ X[:, 1]
>>> y
array([0, 1, 1, 0])
>>> X = PolynomialFeatures(interaction_only=True).fit_transform(X).astype(int)
>>> X
array([[1, 0, 0, 0],
[1, 0, 1, 0],
[1, 1, 0, 0],
[1, 1, 1, 1]])
>>> clf = Perceptron(fit_intercept=False, max_iter=10, tol=None,
... shuffle=False).fit(X, y)
And the classifier “predictions” are perfect:
>>> clf.predict(X)
array([0, 1, 1, 0])
>>> clf.score(X, y)
1.0
3.1.2 Linear and Quadratic Discriminant Analysis
Linear Discriminant Analysis (discriminant_analysis.LinearDiscriminantAnalysis) and

Quadratic Discriminant Analysis (discriminant_analysis.QuadraticDiscriminantAnalysis)
are two classic classifiers, with, as their names suggest, a linear and a quadratic decision surface, respectively.
These classifiers are attractive because they have closed-form solutions that can be easily computed, are inherently
multiclass, have proven to work well in practice, and have no hyperparameters to tune.

The plot shows decision boundaries for Linear Discriminant Analysis and Quadratic Discriminant Analysis. The
bottom row demonstrates that Linear Discriminant Analysis can only learn linear boundaries, while Quadratic Dis-
criminant Analysis can learn quadratic boundaries and is therefore more flexible.
Examples:
Linear and Quadratic Discriminant Analysis with covariance ellipsoid: Comparison of LDA and QDA on synthetic
data.
Dimensionality reduction using Linear Discriminant Analysis
discriminant_analysis.LinearDiscriminantAnalysis can be used to perform supervised dimen-

sionality reduction, by projecting the input data to a linear subspace consisting of the directions which maximize
the separation between classes (in a precise sense discussed in the mathematics section below). The dimension of the
output is necessarily less than the number of classes, so this is, in general, a rather strong dimensionality reduction,
and only makes sense in a multiclass setting.
This is implemented in discriminant_analysis.LinearDiscriminantAnalysis.transform. The
desired dimensionality can be set using the n_components constructor parameter. This parameter has no influence
on discriminant_analysis.LinearDiscriminantAnalysis.fit or discriminant_analysis.

LinearDiscriminantAnalysis.predict.
Examples:
Comparison of LDA and PCA 2D projection of Iris dataset: Comparison of LDA and PCA for dimensionality
reduction of the Iris dataset
Mathematical formulation of the LDA and QDA classifiers
Both LDA and QDA can be derived from simple probabilistic models which model the class conditional distribution
of the data 𝑃 (𝑋|𝑦 = 𝑘) for each class 𝑘. Predictions can then be obtained by using Bayes’ rule:
𝑃 (𝑋|𝑦 = 𝑘)𝑃 (𝑦 = 𝑘) 𝑃 (𝑋|𝑦 = 𝑘)𝑃 (𝑦 = 𝑘)

𝑃 (𝑦 = 𝑘|𝑋) = = ∑︀
𝑃 (𝑋) 𝑙 𝑃 (𝑋|𝑦 = 𝑙) · 𝑃 (𝑦 = 𝑙)
and we select the class 𝑘 which maximizes this conditional probability.

More specifically, for linear and quadratic discriminant analysis, 𝑃 (𝑋|𝑦) is modeled as a multivariate Gaussian distri-
bution with density:
(︂ )︂
1 1 𝑡 −1
𝑃 (𝑋|𝑦 = 𝑘) = exp − (𝑋 − 𝜇𝑘 ) Σ 𝑘 (𝑋 − 𝜇𝑘 )
(2𝜋)𝑑/2 |Σ𝑘 |1/2 2
where 𝑑 is the number of features.
To use this model as a classifier, we just need to estimate from the training data the class priors 𝑃 (𝑦 = 𝑘) (by the
proportion of instances of class 𝑘), the class means 𝜇𝑘 (by the empirical sample class means) and the covariance
matrices (either by the empirical sample class covariance matrices, or by a regularized estimator: see the section on
shrinkage below).
In the case of LDA, the Gaussians for each class are assumed to share the same covariance matrix: Σ𝑘 = Σ for
all 𝑘. This leads to linear decision surfaces, which can be seen by comparing the log-probability ratios log[𝑃 (𝑦 =
𝑘|𝑋)/𝑃 (𝑦 = 𝑙|𝑋)]:
(︂ )︂ (︂ )︂
𝑃 (𝑦 = 𝑘|𝑋) 𝑃 (𝑋|𝑦 = 𝑘)𝑃 (𝑦 = 𝑘)
log = log =0⇔
𝑃 (𝑦 = 𝑙|𝑋) 𝑃 (𝑋|𝑦 = 𝑙)𝑃 (𝑦 = 𝑙)
1 𝑃 (𝑦 = 𝑘)
(𝜇𝑘 − 𝜇𝑙 )𝑡 Σ−1 𝑋 = (𝜇𝑡𝑘 Σ−1 𝜇𝑘 − 𝜇𝑡𝑙 Σ−1 𝜇𝑙 ) − log
2 𝑃 (𝑦 = 𝑙)
In the case of QDA, there are no assumptions on the covariance matrices Σ𝑘 of the Gaussians, leading to quadratic
decision surfaces. See3 for more details.
Note: Relation with Gaussian Naive Bayes

If in the QDA model one assumes that the covariance matrices are diagonal, then the inputs are assumed to be con-
ditionally independent in each class, and the resulting classifier is equivalent to the Gaussian Naive Bayes classifier
naive_bayes.GaussianNB.
Mathematical formulation of LDA dimensionality reduction
To understand the use of LDA in dimensionality reduction, it is useful to start with a geometric reformulation of the
LDA classification rule explained above. We write 𝐾 for the total number of target classes. Since in LDA we assume
3 “The Elements of Statistical Learning”, Hastie T., Tibshirani R., Friedman J., Section 4.3, p.106-119, 2008.

that all classes have the same estimated covariance Σ, we can rescale the data so that this covariance is the identity:
𝑋 * = 𝐷−1/2 𝑈 𝑡 𝑋 with Σ = 𝑈 𝐷𝑈 𝑡
Then one can show that to classify a data point after scaling is equivalent to finding the estimated class mean 𝜇*𝑘 which
is closest to the data point in the Euclidean distance. But this can be done just as well after projecting on the 𝐾 − 1
affine subspace 𝐻𝐾 generated by all the 𝜇*𝑘 for all classes. This shows that, implicit in the LDA classifier, there is a
dimensionality reduction by linear projection onto a 𝐾 − 1 dimensional space.
We can reduce the dimension even more, to a chosen 𝐿, by projecting onto the linear subspace 𝐻𝐿 which max-
imizes the variance of the 𝜇*𝑘 after projection (in effect, we are doing a form of PCA for the transformed class
means 𝜇*𝑘 ). This 𝐿 corresponds to the n_components parameter used in the discriminant_analysis.
LinearDiscriminantAnalysis.transform method. See3 for more details.
Shrinkage
Shrinkage is a tool to improve estimation of covariance matrices in situations where the number of training sam-
ples is small compared to the number of features. In this scenario, the empirical sample covariance is a poor es-
timator. Shrinkage LDA can be used by setting the shrinkage parameter of the discriminant_analysis.
LinearDiscriminantAnalysis class to ‘auto’. This automatically determines the optimal shrinkage parameter
in an analytic way following the lemma introduced by Ledoit and Wolf4 . Note that currently shrinkage only works
when setting the solver parameter to ‘lsqr’ or ‘eigen’.
The shrinkage parameter can also be manually set between 0 and 1. In particular, a value of 0 corresponds to
no shrinkage (which means the empirical covariance matrix will be used) and a value of 1 corresponds to complete
shrinkage (which means that the diagonal matrix of variances will be used as an estimate for the covariance matrix).
Setting this parameter to a value between these two extrema will estimate a shrunk version of the covariance matrix.
4 Ledoit O, Wolf M. Honey, I Shrunk the Sample Covariance Matrix. The Journal of Portfolio Management 30(4), 110-119, 2004.

Estimation algorithms
The default solver is ‘svd’. It can perform both classification and transform, and it does not rely on the calculation
of the covariance matrix. This can be an advantage in situations where the number of features is large. However, the
‘svd’ solver cannot be used with shrinkage.
The ‘lsqr’ solver is an efficient algorithm that only works for classification. It supports shrinkage.
The ‘eigen’ solver is based on the optimization of the between class scatter to within class scatter ratio. It can be used
for both classification and transform, and it supports shrinkage. However, the ‘eigen’ solver needs to compute the
covariance matrix, so it might not be suitable for situations with a high number of features.
Examples:
Normal and Shrinkage Linear Discriminant Analysis for classification: Comparison of LDA classifiers with and
without shrinkage.
References:
3.1.3 Kernel ridge regression
Kernel ridge regression (KRR) [M2012] combines Ridge Regression (linear least squares with l2-norm regularization)
with the kernel trick. It thus learns a linear function in the space induced by the respective kernel and the data. For
non-linear kernels, this corresponds to a non-linear function in the original space.
The form of the model learned by KernelRidge is identical to support vector regression (SVR). However, different
loss functions are used: KRR uses squared error loss while support vector regression uses 𝜖-insensitive loss, both
combined with l2 regularization. In contrast to SVR, fitting KernelRidge can be done in closed-form and is typically
faster for medium-sized datasets. On the other hand, the learned model is non-sparse and thus slower than SVR, which
learns a sparse model for 𝜖 > 0, at prediction-time.
The following figure compares KernelRidge and SVR on an artificial dataset, which consists of a sinusoidal target
function and strong noise added to every fifth datapoint. The learned model of KernelRidge and SVR is plotted,
where both complexity/regularization and bandwidth of the RBF kernel have been optimized using grid-search. The
learned functions are very similar; however, fitting KernelRidge is approx. seven times faster than fitting SVR
(both with grid-search). However, prediction of 100000 target values is more than three times faster with SVR since it
has learned a sparse model using only approx. 1/3 of the 100 training datapoints as support vectors.
The next figure compares the time for fitting and prediction of KernelRidge and SVR for different sizes of the
training set. Fitting KernelRidge is faster than SVR for medium-sized training sets (less than 1000 samples);
however, for larger training sets SVR scales better. With regard to prediction time, SVR is faster than KernelRidge
for all sizes of the training set because of the learned sparse solution. Note that the degree of sparsity and thus the
prediction time depends on the parameters 𝜖 and 𝐶 of the SVR; 𝜖 = 0 would correspond to a dense model.
References:
3.1.4 Support Vector Machines
Support vector machines (SVMs) are a set of supervised learning methods used for classification, regression and
outliers detection.



The advantages of support vector machines are:

• Effective in high dimensional spaces.
• Still effective in cases where number of dimensions is greater than the number of samples.
• Uses a subset of training points in the decision function (called support vectors), so it is also memory efficient.
• Versatile: different Kernel functions can be specified for the decision function. Common kernels are provided,
but it is also possible to specify custom kernels.
The disadvantages of support vector machines include:
• If the number of features is much greater than the number of samples, avoid over-fitting in choosing Kernel
functions and regularization term is crucial.
• SVMs do not directly provide probability estimates, these are calculated using an expensive five-fold cross-
validation (see Scores and probabilities, below).
The support vector machines in scikit-learn support both dense (numpy.ndarray and convertible to that by numpy.
asarray) and sparse (any scipy.sparse) sample vectors as input. However, to use an SVM to make predictions
for sparse data, it must have been fit on such data. For optimal performance, use C-ordered numpy.ndarray (dense)
or scipy.sparse.csr_matrix (sparse) with dtype=float64.
Classification
SVC, NuSVC and LinearSVC are classes capable of performing multi-class classification on a dataset.

SVC and NuSVC are similar methods, but accept slightly different sets of parameters and have different mathematical
formulations (see section Mathematical formulation). On the other hand, LinearSVC is another implementation
of Support Vector Classification for the case of a linear kernel. Note that LinearSVC does not accept keyword
kernel, as this is assumed to be linear. It also lacks some of the members of SVC and NuSVC, like support_.
As other classifiers, SVC, NuSVC and LinearSVC take as input two arrays: an array X of size [n_samples,
n_features] holding the training samples, and an array y of class labels (strings or integers), size [n_samples]:

>>> X = [[0, 0], [1, 1]]
>>> y = [0, 1]
>>> clf.fit(X, y)
>>> clf.predict([[2., 2.]])

array([1])
SVMs decision function depends on some subset of the training data, called the support vectors. Some properties of
these support vectors can be found in members support_vectors_, support_ and n_support:
>>> # get support vectors

>>> clf.support_vectors_
array([[0., 0.],
[1., 1.]])
>>> # get indices of support vectors
>>> clf.support_
array([0, 1]...)
>>> # get number of support vectors for each class
>>> clf.n_support_
array([1, 1]...)
Multi-class classification
SVC and NuSVC implement the “one-against-one” approach (Knerr et al., 1990) for multiclass classifica-
tion. If n_class is the number of classes, then n_class * (n_class - 1) / 2 classifiers are con-
structed and each one trains data from two classes. To provide a consistent interface with other classifiers, the
decision_function_shape option allows to aggregate the results of the “one-against-one” classifiers to a deci-
sion function of shape (n_samples, n_classes):
>>> X = [[0], [1], [2], [3]]

>>> Y = [0, 1, 2, 3]
>>> clf = svm.SVC(gamma='scale', decision_function_shape='ovo')
>>> clf.fit(X, Y)
decision_function_shape='ovo', degree=3, gamma='scale', kernel='rbf',
>>> dec = clf.decision_function([[1]])
>>> dec.shape[1] # 4 classes: 4*3/2 = 6
6

>>> clf.decision_function_shape = "ovr"

>>> dec = clf.decision_function([[1]])
>>> dec.shape[1] # 4 classes
4
On the other hand, LinearSVC implements “one-vs-the-rest” multi-class strategy, thus training n_class models. If
there are only two classes, only one model is trained:
>>> lin_clf = svm.LinearSVC()

>>> lin_clf.fit(X, Y)
LinearSVC(C=1.0, class_weight=None, dual=True, fit_intercept=True,
intercept_scaling=1, loss='squared_hinge', max_iter=1000,
multi_class='ovr', penalty='l2', random_state=None, tol=0.0001,
verbose=0)
>>> dec = lin_clf.decision_function([[1]])
>>> dec.shape[1]
4
See Mathematical formulation for a complete description of the decision function.

Note that the LinearSVC also implements an alternative multi-class strategy, the so-called multi-class SVM formu-
lated by Crammer and Singer, by using the option multi_class='crammer_singer'. This method is consis-
tent, which is not true for one-vs-rest classification. In practice, one-vs-rest classification is usually preferred, since
the results are mostly similar, but the runtime is significantly less.
For “one-vs-rest” LinearSVC the attributes coef_ and intercept_ have the shape [n_class,
n_features] and [n_class] respectively. Each row of the coefficients corresponds to one of the n_class
many “one-vs-rest” classifiers and similar for the intercepts, in the order of the “one” class.
In the case of “one-vs-one” SVC, the layout of the attributes is a little more involved. In the case of having
a linear kernel, the attributes coef_ and intercept_ have the shape [n_class * (n_class - 1) /
2, n_features] and [n_class * (n_class - 1) / 2] respectively. This is similar to the layout for
LinearSVC described above, with each row now corresponding to a binary classifier. The order for classes 0 to n is
“0 vs 1”, “0 vs 2” , . . . “0 vs n”, “1 vs 2”, “1 vs 3”, “1 vs n”, . . . “n-1 vs n”.
The shape of dual_coef_ is [n_class-1, n_SV] with a somewhat hard to grasp layout. The columns corre-
spond to the support vectors involved in any of the n_class * (n_class - 1) / 2 “one-vs-one” classifiers.
Each of the support vectors is used in n_class - 1 classifiers. The n_class - 1 entries in each row correspond
to the dual coefficients for these classifiers.
This might be made more clear by an example:
Consider a three class problem with class 0 having three support vectors 𝑣00 , 𝑣01 , 𝑣02 and class 1 and 2 having two
support vectors 𝑣10 , 𝑣11 and 𝑣20 , 𝑣21 respectively. For each support vector 𝑣𝑖𝑗 , there are two dual coefficients. Let’s call
the coefficient of support vector 𝑣𝑖𝑗 in the classifier between classes 𝑖 and 𝑘 𝛼𝑖,𝑘 𝑗
. Then dual_coef_ looks like this:
0 0
𝛼0,1 𝛼0,2 Coefficients for SVs of class 0
1 1
𝛼0,1 𝛼0,2
2 2
𝛼0,1 𝛼0,2
0 0
1 1
𝛼1,0 𝛼1,2
0 0
1 1
𝛼2,0 𝛼2,1

Scores and probabilities
The decision_function method of SVC and NuSVC gives per-class scores for each sample (or a single score
per sample in the binary case). When the constructor option probability is set to True, class membership
probability estimates (from the methods predict_proba and predict_log_proba) are enabled. In the binary
case, the probabilities are calibrated using Platt scaling: logistic regression on the SVM’s scores, fit by an additional
cross-validation on the training data. In the multiclass case, this is extended as per Wu et al. (2004).
Needless to say, the cross-validation involved in Platt scaling is an expensive operation for large datasets. In addition,
the probability estimates may be inconsistent with the scores, in the sense that the “argmax” of the scores may not be
the argmax of the probabilities. (E.g., in binary classification, a sample may be labeled by predict as belonging
to a class that has probability <½ according to predict_proba.) Platt’s method is also known to have theoret-
ical issues. If confidence scores are required, but these do not have to be probabilities, then it is advisable to set
probability=False and use decision_function instead of predict_proba.
References:
• Wu, Lin and Weng, “Probability estimates for multi-class classification by pairwise coupling”, JMLR 5:975-
1005, 2004.
• Platt “Probabilistic outputs for SVMs and comparisons to regularized likelihood methods”.
Unbalanced problems
In problems where it is desired to give more importance to certain classes or certain individual samples keywords
class_weight and sample_weight can be used.
SVC (but not NuSVC) implement a keyword class_weight in the fit method. It’s a dictionary of the form
{class_label : value}, where value is a floating point number > 0 that sets the parameter C of class
class_label to C * value.

SVC, NuSVC, SVR, NuSVR and OneClassSVM implement also weights for individual samples in method fit
through keyword sample_weight. Similar to class_weight, these set the parameter C for the i-th example to
C * sample_weight[i].
Examples:
• Plot different SVM classifiers in the iris dataset,

• SVM: Maximum margin separating hyperplane,
• SVM: Separating hyperplane for unbalanced classes
• SVM-Anova: SVM with univariate feature selection,
• Non-linear SVM
• SVM: Weighted samples,
Regression
The method of Support Vector Classification can be extended to solve regression problems. This method is called
Support Vector Regression.
The model produced by support vector classification (as described above) depends only on a subset of the training
data, because the cost function for building the model does not care about training points that lie beyond the margin.
Analogously, the model produced by Support Vector Regression depends only on a subset of the training data, because
the cost function for building the model ignores any training data close to the model prediction.
There are three different implementations of Support Vector Regression: SVR, NuSVR and LinearSVR.
LinearSVR provides a faster implementation than SVR but only considers linear kernels, while NuSVR implements
a slightly different formulation than SVR and LinearSVR. See Implementation details for further details.
As with classification classes, the fit method will take as argument vectors X, y, only that in this case y is expected to
have floating point values instead of integer values:

>>> X = [[0, 0], [2, 2]]

>>> y = [0.5, 2.5]

>>> clf = svm.SVR()
>>> clf.fit(X, y)
SVR(C=1.0, cache_size=200, coef0=0.0, degree=3, epsilon=0.1,
gamma='auto_deprecated', kernel='rbf', max_iter=-1, shrinking=True,
>>> clf.predict([[1, 1]])
array([1.5])
Examples:
• Support Vector Regression (SVR) using linear and non-linear kernels
Density estimation, novelty detection
The class OneClassSVM implements a One-Class SVM which is used in outlier detection.
See Novelty and Outlier Detection for the description and usage of OneClassSVM.
Complexity
Support Vector Machines are powerful tools, but their compute and storage requirements increase rapidly with the
number of training vectors. The core of an SVM is a quadratic programming problem (QP), separating support
vectors from the rest of the training data. The QP solver used by this libsvm-based implementation scales between
𝑂(𝑛𝑓 𝑒𝑎𝑡𝑢𝑟𝑒𝑠 × 𝑛2𝑠𝑎𝑚𝑝𝑙𝑒𝑠 ) and 𝑂(𝑛𝑓 𝑒𝑎𝑡𝑢𝑟𝑒𝑠 × 𝑛3𝑠𝑎𝑚𝑝𝑙𝑒𝑠 ) depending on how efficiently the libsvm cache is used in
practice (dataset dependent). If the data is very sparse 𝑛𝑓 𝑒𝑎𝑡𝑢𝑟𝑒𝑠 should be replaced by the average number of non-
zero features in a sample vector.
Also note that for the linear case, the algorithm used in LinearSVC by the liblinear implementation is much more
efficient than its libsvm-based SVC counterpart and can scale almost linearly to millions of samples and/or features.
Tips on Practical Use
• Avoiding data copy: For SVC, SVR, NuSVC and NuSVR, if the data passed to certain methods is not C-ordered
contiguous, and double precision, it will be copied before calling the underlying C implementation. You can
check whether a given numpy array is C-contiguous by inspecting its flags attribute.
For LinearSVC (and LogisticRegression) any input passed as a numpy array will be copied and con-
verted to the liblinear internal sparse data representation (double precision floats and int32 indices of non-zero
components). If you want to fit a large-scale linear classifier without copying a dense numpy C-contiguous
double precision array as input we suggest to use the SGDClassifier class instead. The objective function
can be configured to be almost the same as the LinearSVC model.
• Kernel cache size: For SVC, SVR, NuSVC and NuSVR, the size of the kernel cache has a strong impact on run
times for larger problems. If you have enough RAM available, it is recommended to set cache_size to a
higher value than the default of 200(MB), such as 500(MB) or 1000(MB).
• Setting C: C is 1 by default and it’s a reasonable default choice. If you have a lot of noisy observations you
should decrease it. It corresponds to regularize more the estimation.
LinearSVC and :class‘LinearSVR‘ are less sensitive to C when it becomes large, and prediction results stop
improving after a certain threshold. Meanwhile, larger C values will take more time to train, sometimes up to
10 times longer, as shown by Fan et al. (2008)

• Support Vector Machine algorithms are not scale invariant, so it is highly recommended to scale your data.
For example, scale each attribute on the input vector X to [0,1] or [-1,+1], or standardize it to have mean 0
and variance 1. Note that the same scaling must be applied to the test vector to obtain meaningful results. See
section Preprocessing data for more details on scaling and normalization.
• Parameter nu in NuSVC/OneClassSVM /NuSVR approximates the fraction of training errors and support vec-
tors.
• In SVC, if data for classification are unbalanced (e.g. many positive and few negative), set
class_weight='balanced' and/or try different penalty parameters C.
• Randomness of the underlying implementations: The underlying implementations of SVC and NuSVC use
a random number generator only to shuffle the data for probability estimation (when probability is set to
True). This randomness can be controlled with the random_state parameter. If probability is set
to False these estimators are not random and random_state has no effect on the results. The underlying
OneClassSVM implementation is similar to the ones of SVC and NuSVC. As no probability estimation is
provided for OneClassSVM , it is not random.
The underlying LinearSVC implementation uses a random number generator to select features when fitting the
model with a dual coordinate descent (i.e when dual is set to True). It is thus not uncommon, to have slightly
different results for the same input data. If that happens, try with a smaller tol parameter. This randomness
can also be controlled with the random_state parameter. When dual is set to False the underlying
implementation of LinearSVC is not random and random_state has no effect on the results.
• Using L1 penalization as provided by LinearSVC(loss='l2', penalty='l1', dual=False)
yields a sparse solution, i.e. only a subset of feature weights is different from zero and contribute to the de-
cision function. Increasing C yields a more complex model (more feature are selected). The C value that yields
a “null” model (all weights equal to zero) can be calculated using l1_min_c.
References:
• Fan, Rong-En, et al., “LIBLINEAR: A library for large linear classification.”, Journal of machine learning
research 9.Aug (2008): 1871-1874.
Kernel functions
The kernel function can be any of the following:

• linear: ⟨𝑥, 𝑥′ ⟩.
• polynomial: (𝛾⟨𝑥, 𝑥′ ⟩ + 𝑟)𝑑 . 𝑑 is specified by keyword degree, 𝑟 by coef0.
• rbf: exp(−𝛾‖𝑥 − 𝑥′ ‖2 ). 𝛾 is specified by keyword gamma, must be greater than 0.
• sigmoid (tanh(𝛾⟨𝑥, 𝑥′ ⟩ + 𝑟)), where 𝑟 is specified by coef0.
Different kernels are specified by keyword kernel at initialization:
>>> linear_svc = svm.SVC(kernel='linear')

>>> linear_svc.kernel
'linear'
>>> rbf_svc = svm.SVC(kernel='rbf')
>>> rbf_svc.kernel
'rbf'

Custom Kernels
You can define your own kernels by either giving the kernel as a python function or by precomputing the Gram matrix.
Classifiers with custom kernels behave the same way as any other classifiers, except that:
• Field support_vectors_ is now empty, only indices of support vectors are stored in support_
• A reference (and not a copy) of the first argument in the fit() method is stored for future reference. If that
array changes between the use of fit() and predict() you will have unexpected results.
Using Python functions as kernels
You can also use your own defined kernels by passing a function to the keyword kernel in the constructor.
Your kernel must take as arguments two matrices of shape (n_samples_1, n_features), (n_samples_2,
n_features) and return a kernel matrix of shape (n_samples_1, n_samples_2).
The following code defines a linear kernel and creates a classifier instance that will use that kernel:

>>> def my_kernel(X, Y):
... return np.dot(X, Y.T)
...
>>> clf = svm.SVC(kernel=my_kernel)
Examples:
• SVM with custom kernel.
Using the Gram matrix
Set kernel='precomputed' and pass the Gram matrix instead of X in the fit method. At the moment, the kernel
values between all training vectors and the test vectors must be provided.

>>> X = np.array([[0, 0], [1, 1]])
>>> y = [0, 1]
>>> clf = svm.SVC(kernel='precomputed')
>>> # linear kernel computation
>>> gram = np.dot(X, X.T)
>>> clf.fit(gram, y)
kernel='precomputed', max_iter=-1, probability=False,
random_state=None, shrinking=True, tol=0.001, verbose=False)
>>> # predict on training examples
>>> clf.predict(gram)
array([0, 1])

Parameters of the RBF Kernel
When training an SVM with the Radial Basis Function (RBF) kernel, two parameters must be considered: C and
gamma. The parameter C, common to all SVM kernels, trades off misclassification of training examples against
simplicity of the decision surface. A low C makes the decision surface smooth, while a high C aims at classifying all
training examples correctly. gamma defines how much influence a single training example has. The larger gamma is,
the closer other examples must be to be affected.
Proper choice of C and gamma is critical to the SVM’s performance. One is advised to use sklearn.
model_selection.GridSearchCV with C and gamma spaced exponentially far apart to choose good values.
Examples:
• RBF SVM parameters
A support vector machine constructs a hyper-plane or set of hyper-planes in a high or infinite dimensional space, which
can be used for classification, regression or other tasks. Intuitively, a good separation is achieved by the hyper-plane
that has the largest distance to the nearest training data points of any class (so-called functional margin), since in
general the larger the margin the lower the generalization error of the classifier.

SVC
Given training vectors 𝑥𝑖 ∈ R𝑝 , i=1,. . . , n, in two classes, and a vector 𝑦 ∈ {1, −1}𝑛 , SVC solves the following primal
problem:
𝑛
1 ∑︁
min 𝑤𝑇 𝑤 + 𝐶 𝜁𝑖
𝑤,𝑏,𝜁 2
𝑖=1
subject to 𝑦𝑖 (𝑤𝑇 𝜑(𝑥𝑖 ) + 𝑏) ≥ 1 − 𝜁𝑖 ,
𝜁𝑖 ≥ 0, 𝑖 = 1, ..., 𝑛
Its dual is
1
min 𝛼𝑇 𝑄𝛼 − 𝑒𝑇 𝛼
𝛼 2
subject to 𝑦 𝑇 𝛼 = 0
0 ≤ 𝛼𝑖 ≤ 𝐶, 𝑖 = 1, ..., 𝑛
where 𝑒 is the vector of all ones, 𝐶 > 0 is the upper bound, 𝑄 is an 𝑛 by 𝑛 positive semidefinite matrix, 𝑄𝑖𝑗 ≡
𝑦𝑖 𝑦𝑗 𝐾(𝑥𝑖 , 𝑥𝑗 ), where 𝐾(𝑥𝑖 , 𝑥𝑗 ) = 𝜑(𝑥𝑖 )𝑇 𝜑(𝑥𝑗 ) is the kernel. Here training vectors are implicitly mapped into a
higher (maybe infinite) dimensional space by the function 𝜑.
The decision function is:
𝑛
∑︁
sgn( 𝑦𝑖 𝛼𝑖 𝐾(𝑥𝑖 , 𝑥) + 𝜌)
𝑖=1
Note: While SVM models derived from libsvm and liblinear use C as regularization parameter, most other estimators
use alpha. The exact equivalence between the amount of regularization of two models depends on the exact objective
function optimized by the model. For example, when the estimator used is sklearn.linear_model.Ridge
1
regression, the relation between them is given as 𝐶 = 𝑎𝑙𝑝ℎ𝑎 .
This parameters can be accessed through the members dual_coef_ which holds the product 𝑦𝑖 𝛼𝑖 ,
support_vectors_ which holds the support vectors, and intercept_ which holds the independent term 𝜌
:
References:
• “Automatic Capacity Tuning of Very Large VC-dimension Classifiers”, I. Guyon, B. Boser, V. Vapnik -
Advances in neural information processing 1993.
• “Support-vector networks”, C. Cortes, V. Vapnik - Machine Learning, 20, 273-297 (1995).
NuSVC
We introduce a new parameter 𝜈 which controls the number of support vectors and training errors. The parameter
𝜈 ∈ (0, 1] is an upper bound on the fraction of training errors and a lower bound of the fraction of support vectors.
It can be shown that the 𝜈-SVC formulation is a reparameterization of the 𝐶-SVC and therefore mathematically
equivalent.

SVR
Given training vectors 𝑥𝑖 ∈ R𝑝 , i=1,. . . , n, and a vector 𝑦 ∈ R𝑛 𝜀-SVR solves the following primal problem:
𝑛
1 ∑︁
min * 𝑤𝑇 𝑤 + 𝐶 (𝜁𝑖 + 𝜁𝑖* )
𝑤,𝑏,𝜁,𝜁 2
𝑖=1
subject to 𝑦𝑖 − 𝑤𝑇 𝜑(𝑥𝑖 ) − 𝑏 ≤ 𝜀 + 𝜁𝑖 ,
𝑤𝑇 𝜑(𝑥𝑖 ) + 𝑏 − 𝑦𝑖 ≤ 𝜀 + 𝜁𝑖* ,
𝜁𝑖 , 𝜁𝑖* ≥ 0, 𝑖 = 1, ..., 𝑛
Its dual is
1
min (𝛼 − 𝛼* )𝑇 𝑄(𝛼 − 𝛼* ) + 𝜀𝑒𝑇 (𝛼 + 𝛼* ) − 𝑦 𝑇 (𝛼 − 𝛼* )
2
𝛼,𝛼*
subject to 𝑒𝑇 (𝛼 − 𝛼* ) = 0
0 ≤ 𝛼𝑖 , 𝛼𝑖* ≤ 𝐶, 𝑖 = 1, ..., 𝑛
where 𝑒 is the vector of all ones, 𝐶 > 0 is the upper bound, 𝑄 is an 𝑛 by 𝑛 positive semidefinite matrix, 𝑄𝑖𝑗 ≡
𝐾(𝑥𝑖 , 𝑥𝑗 ) = 𝜑(𝑥𝑖 )𝑇 𝜑(𝑥𝑗 ) is the kernel. Here training vectors are implicitly mapped into a higher (maybe infinite)
dimensional space by the function 𝜑.
The decision function is:
𝑛
∑︁
(𝛼𝑖 − 𝛼𝑖* )𝐾(𝑥𝑖 , 𝑥) + 𝜌
𝑖=1
These parameters can be accessed through the members dual_coef_ which holds the difference 𝛼𝑖 − 𝛼𝑖* ,
support_vectors_ which holds the support vectors, and intercept_ which holds the independent term 𝜌
References:
• “A Tutorial on Support Vector Regression”, Alex J. Smola, Bernhard Schölkopf - Statistics and Computing
archive Volume 14 Issue 3, August 2004, p. 199-222.
Implementation details
Internally, we use libsvm and liblinear to handle all computations. These libraries are wrapped using C and Cython.
References:
For a description of the implementation and details of the algorithms used, please refer to
• LIBSVM: A Library for Support Vector Machines.
• LIBLINEAR – A Library for Large Linear Classification.
3.1.5 Stochastic Gradient Descent
Stochastic Gradient Descent (SGD) is a simple yet very efficient approach to discriminative learning of linear clas-
sifiers under convex loss functions such as (linear) Support Vector Machines and Logistic Regression. Even though

SGD has been around in the machine learning community for a long time, it has received a considerable amount of
attention just recently in the context of large-scale learning.
SGD has been successfully applied to large-scale and sparse machine learning problems often encountered in text
classification and natural language processing. Given that the data is sparse, the classifiers in this module easily scale
to problems with more than 10^5 training examples and more than 10^5 features.
The advantages of Stochastic Gradient Descent are:
• Efficiency.
• Ease of implementation (lots of opportunities for code tuning).
The disadvantages of Stochastic Gradient Descent include:
• SGD requires a number of hyperparameters such as the regularization parameter and the number of iterations.
• SGD is sensitive to feature scaling.
Classification
Warning: Make sure you permute (shuffle) your training data before fitting the model or use shuffle=True
to shuffle after each iteration.
The class SGDClassifier implements a plain stochastic gradient descent learning routine which supports different
loss functions and penalties for classification.
As other classifiers, SGD has to be fitted with two arrays: an array X of size [n_samples, n_features] holding the
training samples, and an array Y of size [n_samples] holding the target values (class labels) for the training samples:

>>> X = [[0., 0.], [1., 1.]]

>>> y = [0, 1]
>>> clf = SGDClassifier(loss="hinge", penalty="l2", max_iter=5)
>>> clf.fit(X, y)
SGDClassifier(alpha=0.0001, average=False, class_weight=None,
early_stopping=False, epsilon=0.1, eta0=0.0, fit_intercept=True,
l1_ratio=0.15, learning_rate='optimal', loss='hinge', max_iter=5,
n_iter=None, n_iter_no_change=5, n_jobs=None, penalty='l2',
power_t=0.5, random_state=None, shuffle=True, tol=None,
validation_fraction=0.1, verbose=0, warm_start=False)
>>> clf.predict([[2., 2.]])

array([1])
SGD fits a linear model to the training data. The member coef_ holds the model parameters:
>>> clf.coef_
array([[9.9..., 9.9...]])
Member intercept_ holds the intercept (aka offset or bias):
>>> clf.intercept_
array([-9.9...])
Whether or not the model should use an intercept, i.e. a biased hyperplane, is controlled by the parameter
fit_intercept.
To get the signed distance to the hyperplane use SGDClassifier.decision_function:
>>> clf.decision_function([[2., 2.]])

array([29.6...])
The concrete loss function can be set via the loss parameter. SGDClassifier supports the following loss func-
tions:
• loss="hinge": (soft-margin) linear Support Vector Machine,
• loss="modified_huber": smoothed hinge loss,
• loss="log": logistic regression,
• and all regression losses below.
The first two loss functions are lazy, they only update the model parameters if an example violates the margin con-
straint, which makes training very efficient and may result in sparser models, even when L2 penalty is used.
Using loss="log" or loss="modified_huber" enables the predict_proba method, which gives a vector
of probability estimates 𝑃 (𝑦|𝑥) per sample 𝑥:
>>> clf = SGDClassifier(loss="log", max_iter=5).fit(X, y)

>>> clf.predict_proba([[1., 1.]])
array([[0.00..., 0.99...]])
The concrete penalty can be set via the penalty parameter. SGD supports the following penalties:
• penalty="l2": L2 norm penalty on coef_.
• penalty="l1": L1 norm penalty on coef_.

• penalty="elasticnet": Convex combination of L2 and L1; (1 - l1_ratio) * L2 +

l1_ratio * L1.
The default setting is penalty="l2". The L1 penalty leads to sparse solutions, driving most coefficients to zero.
The Elastic Net solves some deficiencies of the L1 penalty in the presence of highly correlated attributes. The param-
eter l1_ratio controls the convex combination of L1 and L2 penalty.
SGDClassifier supports multi-class classification by combining multiple binary classifiers in a “one versus all”
(OVA) scheme. For each of the 𝐾 classes, a binary classifier is learned that discriminates between that and all other
𝐾 − 1 classes. At testing time, we compute the confidence score (i.e. the signed distances to the hyperplane) for each
classifier and choose the class with the highest confidence. The Figure below illustrates the OVA approach on the iris
dataset. The dashed lines represent the three OVA classifiers; the background colors show the decision surface induced
by the three classifiers.
In the case of multi-class classification coef_ is a two-dimensional array of shape=[n_classes,

n_features] and intercept_ is a one-dimensional array of shape=[n_classes]. The i-th row of coef_
holds the weight vector of the OVA classifier for the i-th class; classes are indexed in ascending order (see at-
tribute classes_). Note that, in principle, since they allow to create a probability model, loss="log" and
loss="modified_huber" are more suitable for one-vs-all classification.
SGDClassifier supports both weighted classes and weighted instances via the fit parameters class_weight
and sample_weight. See the examples below and the doc string of SGDClassifier.fit for further informa-
tion.
Examples:
• SGD: Maximum margin separating hyperplane,

• Plot multi-class SGD on the iris dataset
• SGD: Weighted samples
• Comparing various online solvers

• SVM: Separating hyperplane for unbalanced classes (See the Note)
SGDClassifier supports averaged SGD (ASGD). Averaging can be enabled by setting àverage=True`.
ASGD works by averaging the coefficients of the plain SGD over each iteration over a sample. When using ASGD
the learning rate can be larger and even constant leading on some datasets to a speed up in training time.
For classification with a logistic loss, another variant of SGD with an averaging strategy is available with Stochastic
Average Gradient (SAG) algorithm, available as a solver in LogisticRegression.
Regression
The class SGDRegressor implements a plain stochastic gradient descent learning routine which supports different
loss functions and penalties to fit linear regression models. SGDRegressor is well suited for regression prob-
lems with a large number of training samples (> 10.000), for other problems we recommend Ridge, Lasso, or
ElasticNet.
The concrete loss function can be set via the loss parameter. SGDRegressor supports the following loss functions:
• loss="squared_loss": Ordinary least squares,
• loss="huber": Huber loss for robust regression,
• loss="epsilon_insensitive": linear Support Vector Regression.
The Huber and epsilon-insensitive loss functions can be used for robust regression. The width of the insensitive region
has to be specified via the parameter epsilon. This parameter depends on the scale of the target variables.
SGDRegressor supports averaged SGD as SGDClassifier. Averaging can be enabled by setting
àverage=True`.
For regression with a squared loss and a l2 penalty, another variant of SGD with an averaging strategy is available with
Stochastic Average Gradient (SAG) algorithm, available as a solver in Ridge.
Stochastic Gradient Descent for sparse data
Note: The sparse implementation produces slightly different results than the dense implementation due to a shrunk
learning rate for the intercept.
There is built-in support for sparse data given in any matrix in a format supported by scipy.sparse. For maximum
efficiency, however, use the CSR matrix format as defined in scipy.sparse.csr_matrix.
Examples:
Complexity
The major advantage of SGD is its efficiency, which is basically linear in the number of training examples. If X is a
matrix of size (n, p) training has a cost of 𝑂(𝑘𝑛¯
𝑝), where k is the number of iterations (epochs) and 𝑝¯ is the average
number of non-zero attributes per sample.
Recent theoretical results, however, show that the runtime to get some desired optimization accuracy does not increase
as the training set size increases.

Stopping criterion
The classes SGDClassifier and SGDRegressor provide two criteria to stop the algorithm when a given level of
convergence is reached:
• With early_stopping=True, the input data is split into a training set and a validation set. The model
is then fitted on the training set, and the stopping criterion is based on the prediction score computed on the
validation set. The size of the validation set can be changed with the parameter validation_fraction.
• With early_stopping=False, the model is fitted on the entire input data and the stopping criterion is
based on the objective function computed on the input data.
In both cases, the criterion is evaluated once by epoch, and the algorithm stops when the criterion does not improve
n_iter_no_change times in a row. The improvement is evaluated with a tolerance tol, and the algorithm stops
in any case after a maximum number of iteration max_iter.
• Stochastic Gradient Descent is sensitive to feature scaling, so it is highly recommended to scale your data. For
example, scale each attribute on the input vector X to [0,1] or [-1,+1], or standardize it to have mean 0 and
variance 1. Note that the same scaling must be applied to the test vector to obtain meaningful results. This can
be easily done using StandardScaler:
from sklearn.preprocessing import StandardScaler
scaler = StandardScaler()
scaler.fit(X_train) # Don't cheat - fit only on training data
X_train = scaler.transform(X_train)
X_test = scaler.transform(X_test) # apply same transformation to test data
If your attributes have an intrinsic scale (e.g. word frequencies or indicator features) scaling is not needed.
• Finding a reasonable regularization term 𝛼 is best done using GridSearchCV, usually in the range 10.
0**-np.arange(1,7).
• Empirically, we found that SGD converges after observing approx. 10^6 training samples. Thus, a reasonable
first guess for the number of iterations is max_iter = np.ceil(10**6 / n), where n is the size of the
training set.
• If you apply SGD to features extracted using PCA we found that it is often wise to scale the feature values by
some constant c such that the average L2 norm of the training data equals one.
• We found that Averaged SGD works best with a larger number of features and a higher eta0
References:
• “Efficient BackProp” Y. LeCun, L. Bottou, G. Orr, K. Müller - In Neural Networks: Tricks of the Trade 1998.
Given a set of training examples (𝑥1 , 𝑦1 ), . . . , (𝑥𝑛 , 𝑦𝑛 ) where 𝑥𝑖 ∈ R𝑚 and 𝑦𝑖 ∈ {−1, 1}, our goal is to learn a linear
scoring function 𝑓 (𝑥) = 𝑤𝑇 𝑥 + 𝑏 with model parameters 𝑤 ∈ R𝑚 and intercept 𝑏 ∈ R. In order to make predictions,
we simply look at the sign of 𝑓 (𝑥). A common choice to find the model parameters is by minimizing the regularized
training error given by
𝑛
1 ∑︁
𝐸(𝑤, 𝑏) = 𝐿(𝑦𝑖 , 𝑓 (𝑥𝑖 )) + 𝛼𝑅(𝑤)
𝑛 𝑖=1

where 𝐿 is a loss function that measures model (mis)fit and 𝑅 is a regularization term (aka penalty) that penalizes
model complexity; 𝛼 > 0 is a non-negative hyperparameter.
Different choices for 𝐿 entail different classifiers such as
• Hinge: (soft-margin) Support Vector Machines.
• Log: Logistic Regression.
• Least-Squares: Ridge Regression.
• Epsilon-Insensitive: (soft-margin) Support Vector Regression.
All of the above loss functions can be regarded as an upper bound on the misclassification error (Zero-one loss) as
shown in the Figure below.
Popular choices for the regularization term 𝑅 include:

∑︀𝑛
• L2 norm: 𝑅(𝑤) := 12 𝑖=1 𝑤𝑖2 ,
∑︀𝑛
• L1 norm: 𝑅(𝑤) := 𝑖=1 |𝑤𝑖 |, which leads to sparse solutions.
∑︀𝑛 ∑︀𝑛
• Elastic Net: 𝑅(𝑤) := 𝜌2 𝑖=1 𝑤𝑖2 + (1 − 𝜌) 𝑖=1 |𝑤𝑖 |, a convex combination of L2 and L1, where 𝜌 is given
by 1 - l1_ratio.
The Figure below shows the contours of the different regularization terms in the parameter space when 𝑅(𝑤) = 1.
SGD
Stochastic gradient descent is an optimization method for unconstrained optimization problems. In contrast to (batch)
gradient descent, SGD approximates the true gradient of 𝐸(𝑤, 𝑏) by considering a single training example at a time.
The class SGDClassifier implements a first-order SGD learning routine. The algorithm iterates over the training


examples and for each example updates the model parameters according to the update rule given by
𝜕𝑅(𝑤) 𝜕𝐿(𝑤𝑇 𝑥𝑖 + 𝑏, 𝑦𝑖 )
𝑤 ← 𝑤 − 𝜂(𝛼 + )
𝜕𝑤 𝜕𝑤
where 𝜂 is the learning rate which controls the step-size in the parameter space. The intercept 𝑏 is updated similarly
but without regularization.
The learning rate 𝜂 can be either constant or gradually decaying. For classification, the default learning rate schedule
(learning_rate='optimal') is given by
1
𝜂 (𝑡) =
𝛼(𝑡0 + 𝑡)
where 𝑡 is the time step (there are a total of n_samples * n_iter time steps), 𝑡0 is determined based on a heuristic
proposed by Léon Bottou such that the expected initial updates are comparable with the expected size of the weights
(this assuming that the norm of the training samples is approx. 1). The exact definition can be found in _init_t in
BaseSGD.
For regression the default learning rate schedule is inverse scaling (learning_rate='invscaling'), given by
𝑒𝑡𝑎0
𝜂 (𝑡) = 𝑝𝑜𝑤𝑒𝑟_𝑡
𝑡
where 𝑒𝑡𝑎0 and 𝑝𝑜𝑤𝑒𝑟_𝑡 are hyperparameters chosen by the user via eta0 and power_t, resp.
For a constant learning rate use learning_rate='constant' and use eta0 to specify the learning rate.
For an adaptively decreasing learning rate, use learning_rate='adaptive' and use eta0 to specify the start-
ing learning rate. When the stopping criterion is reached, the learning rate is divided by 5, and the algorithm does not
stop. The algorithm stops when the learning rate goes below 1e-6.
The model parameters can be accessed through the members coef_ and intercept_:
• Member coef_ holds the weights 𝑤
• Member intercept_ holds 𝑏
References:
• “Solving large scale linear prediction problems using stochastic gradient descent algorithms” T. Zhang - In
Proceedings of ICML ‘04.
• “Regularization and variable selection via the elastic net” H. Zou, T. Hastie - Journal of the Royal Statistical
Society Series B, 67 (2), 301-320.
• “Towards Optimal One Pass Large Scale Learning with Averaged Stochastic Gradient Descent” Xu, Wei
The implementation of SGD is influenced by the Stochastic Gradient SVM of Léon Bottou. Similar to SvmSGD,
the weight vector is represented as the product of a scalar and a vector which allows an efficient weight update in
the case of L2 regularization. In the case of sparse feature vectors, the intercept is updated with a smaller learning
rate (multiplied by 0.01) to account for the fact that it is updated more frequently. Training examples are picked up
sequentially and the learning rate is lowered after each observed example. We adopted the learning rate schedule from
Shalev-Shwartz et al. 2007. For multi-class classification, a “one versus all” approach is used. We use the truncated
gradient algorithm proposed by Tsuruoka et al. 2009 for L1 regularization (and the Elastic Net). The code is written
in Cython.

References:
• “Stochastic Gradient Descent” L. Bottou - Website, 2010.

• “The Tradeoffs of Large Scale Machine Learning” L. Bottou - Website, 2011.
• “Pegasos: Primal estimated sub-gradient solver for svm” S. Shalev-Shwartz, Y. Singer, N. Srebro - In Pro-
ceedings of ICML ‘07.
• “Stochastic gradient descent training for l1-regularized log-linear models with cumulative penalty” Y. Tsu-
ruoka, J. Tsujii, S. Ananiadou - In Proceedings of the AFNLP/ACL ‘09.
3.1.6 Nearest Neighbors
sklearn.neighbors provides functionality for unsupervised and supervised neighbors-based learning methods.
Unsupervised nearest neighbors is the foundation of many other learning methods, notably manifold learning and
spectral clustering. Supervised neighbors-based learning comes in two flavors: classification for data with discrete
labels, and regression for data with continuous labels.
The principle behind nearest neighbor methods is to find a predefined number of training samples closest in distance
to the new point, and predict the label from these. The number of samples can be a user-defined constant (k-nearest
neighbor learning), or vary based on the local density of points (radius-based neighbor learning). The distance can,
in general, be any metric measure: standard Euclidean distance is the most common choice. Neighbors-based meth-
ods are known as non-generalizing machine learning methods, since they simply “remember” all of its training data
(possibly transformed into a fast indexing structure such as a Ball Tree or KD Tree).
Despite its simplicity, nearest neighbors has been successful in a large number of classification and regression prob-
lems, including handwritten digits and satellite image scenes. Being a non-parametric method, it is often successful in
classification situations where the decision boundary is very irregular.
The classes in sklearn.neighbors can handle either NumPy arrays or scipy.sparse matrices as input. For dense
matrices, a large number of possible distance metrics are supported. For sparse matrices, arbitrary Minkowski metrics
are supported for searches.
There are many learning routines which rely on nearest neighbors at their core. One example is kernel density estima-
tion, discussed in the density estimation section.
Unsupervised Nearest Neighbors
NearestNeighbors implements unsupervised nearest neighbors learning. It acts as a uniform interface to three
different nearest neighbors algorithms: BallTree, KDTree, and a brute-force algorithm based on routines in
sklearn.metrics.pairwise. The choice of neighbors search algorithm is controlled through the keyword
'algorithm', which must be one of ['auto', 'ball_tree', 'kd_tree', 'brute']. When the de-
fault value 'auto' is passed, the algorithm attempts to determine the best approach from the training data. For a
discussion of the strengths and weaknesses of each option, see Nearest Neighbor Algorithms.
Warning: Regarding the Nearest Neighbors algorithms, if two neighbors 𝑘 + 1 and 𝑘 have identical
distances but different labels, the result will depend on the ordering of the training data.

Finding the Nearest Neighbors
For the simple task of finding the nearest neighbors between two sets of data, the unsupervised algorithms within
sklearn.neighbors can be used:
>>> from sklearn.neighbors import NearestNeighbors

>>> X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]])
>>> nbrs = NearestNeighbors(n_neighbors=2, algorithm='ball_tree').fit(X)
>>> distances, indices = nbrs.kneighbors(X)
>>> indices
array([[0, 1],
[1, 0],
[2, 1],
[3, 4],
[4, 3],
[5, 4]]...)
>>> distances
array([[0. , 1. ],
[0. , 1. ],
[0. , 1.41421356],
[0. , 1. ],
[0. , 1. ],
[0. , 1.41421356]])
Because the query set matches the training set, the nearest neighbor of each point is the point itself, at a distance of
zero.
It is also possible to efficiently produce a sparse graph showing the connections between neighboring points:
>>> nbrs.kneighbors_graph(X).toarray()
array([[1., 1., 0., 0., 0., 0.],
[1., 1., 0., 0., 0., 0.],
[0., 1., 1., 0., 0., 0.],
[0., 0., 0., 1., 1., 0.],
[0., 0., 0., 1., 1., 0.],
[0., 0., 0., 0., 1., 1.]])
The dataset is structured such that points nearby in index order are nearby in parameter space, leading to an ap-
proximately block-diagonal matrix of K-nearest neighbors. Such a sparse graph is useful in a variety of cir-
cumstances which make use of spatial relationships between points for unsupervised learning: in particular,
see sklearn.manifold.Isomap, sklearn.manifold.LocallyLinearEmbedding, and sklearn.
cluster.SpectralClustering.
KDTree and BallTree Classes
Alternatively, one can use the KDTree or BallTree classes directly to find nearest neighbors. This is the function-
ality wrapped by the NearestNeighbors class used above. The Ball Tree and KD Tree have the same interface;
we’ll show an example of using the KD Tree here:
>>> from sklearn.neighbors import KDTree

>>> X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]])
>>> kdt = KDTree(X, leaf_size=30, metric='euclidean')
>>> kdt.query(X, k=2, return_distance=False)
array([[0, 1],

[1, 0],
[2, 1],
[3, 4],
[4, 3],
[5, 4]]...)
Refer to the KDTree and BallTree class documentation for more information on the options available for nearest
neighbors searches, including specification of query strategies, distance metrics, etc. For a list of available metrics,
see the documentation of the DistanceMetric class.
Nearest Neighbors Classification
Neighbors-based classification is a type of instance-based learning or non-generalizing learning: it does not attempt
to construct a general internal model, but simply stores instances of the training data. Classification is computed from
a simple majority vote of the nearest neighbors of each point: a query point is assigned the data class which has the
most representatives within the nearest neighbors of the point.
scikit-learn implements two different nearest neighbors classifiers: KNeighborsClassifier implements learn-
ing based on the 𝑘 nearest neighbors of each query point, where 𝑘 is an integer value specified by the user.
RadiusNeighborsClassifier implements learning based on the number of neighbors within a fixed radius
𝑟 of each training point, where 𝑟 is a floating-point value specified by the user.
The 𝑘-neighbors classification in KNeighborsClassifier is the most commonly used technique. The optimal
choice of the value 𝑘 is highly data-dependent: in general a larger 𝑘 suppresses the effects of noise, but makes the
classification boundaries less distinct.
In cases where the data is not uniformly sampled, radius-based neighbors classification in
RadiusNeighborsClassifier can be a better choice. The user specifies a fixed radius 𝑟, such that
points in sparser neighborhoods use fewer nearest neighbors for the classification. For high-dimensional parameter
spaces, this method becomes less effective due to the so-called “curse of dimensionality”.
The basic nearest neighbors classification uses uniform weights: that is, the value assigned to a query point is computed
from a simple majority vote of the nearest neighbors. Under some circumstances, it is better to weight the neighbors
such that nearer neighbors contribute more to the fit. This can be accomplished through the weights keyword. The
default value, weights = 'uniform', assigns uniform weights to each neighbor. weights = 'distance'
assigns weights proportional to the inverse of the distance from the query point. Alternatively, a user-defined function
of the distance can be supplied to compute the weights.

Examples:
• Nearest Neighbors Classification: an example of classification using nearest neighbors.
Nearest Neighbors Regression
Neighbors-based regression can be used in cases where the data labels are continuous rather than discrete variables.
The label assigned to a query point is computed based on the mean of the labels of its nearest neighbors.
scikit-learn implements two different neighbors regressors: KNeighborsRegressor implements learning
based on the 𝑘 nearest neighbors of each query point, where 𝑘 is an integer value specified by the user.
RadiusNeighborsRegressor implements learning based on the neighbors within a fixed radius 𝑟 of the query
point, where 𝑟 is a floating-point value specified by the user.
The basic nearest neighbors regression uses uniform weights: that is, each point in the local neighborhood contributes
uniformly to the classification of a query point. Under some circumstances, it can be advantageous to weight points
such that nearby points contribute more to the regression than faraway points. This can be accomplished through the
weights keyword. The default value, weights = 'uniform', assigns equal weights to all points. weights
= 'distance' assigns weights proportional to the inverse of the distance from the query point. Alternatively, a
user-defined function of the distance can be supplied, which will be used to compute the weights.
The use of multi-output nearest neighbors for regression is demonstrated in Face completion with a multi-output
estimators. In this example, the inputs X are the pixels of the upper half of faces and the outputs Y are the pixels of
the lower half of those faces.
Examples:
• Nearest Neighbors regression: an example of regression using nearest neighbors.


• Face completion with a multi-output estimators: an example of multi-output regression using nearest neigh-
bors.
Nearest Neighbor Algorithms
Brute Force
Fast computation of nearest neighbors is an active area of research in machine learning. The most naive neighbor
search implementation involves the brute-force computation of distances between all pairs of points in the dataset: for
𝑁 samples in 𝐷 dimensions, this approach scales as 𝑂[𝐷𝑁 2 ]. Efficient brute-force neighbors searches can be very
competitive for small data samples. However, as the number of samples 𝑁 grows, the brute-force approach quickly
becomes infeasible. In the classes within sklearn.neighbors, brute-force neighbors searches are specified using
the keyword algorithm = 'brute', and are computed using the routines available in sklearn.metrics.
pairwise.
K-D Tree
To address the computational inefficiencies of the brute-force approach, a variety of tree-based data structures have
been invented. In general, these structures attempt to reduce the required number of distance calculations by efficiently
encoding aggregate distance information for the sample. The basic idea is that if point 𝐴 is very distant from point
𝐵, and point 𝐵 is very close to point 𝐶, then we know that points 𝐴 and 𝐶 are very distant, without having to
explicitly calculate their distance. In this way, the computational cost of a nearest neighbors search can be reduced to
𝑂[𝐷𝑁 log(𝑁 )] or better. This is a significant improvement over brute-force for large 𝑁 .
An early approach to taking advantage of this aggregate information was the KD tree data structure (short for K-
dimensional tree), which generalizes two-dimensional Quad-trees and 3-dimensional Oct-trees to an arbitrary number
of dimensions. The KD tree is a binary tree structure which recursively partitions the parameter space along the data
axes, dividing it into nested orthotropic regions into which data points are filed. The construction of a KD tree is very
fast: because partitioning is performed only along the data axes, no 𝐷-dimensional distances need to be computed.
Once constructed, the nearest neighbor of a query point can be determined with only 𝑂[log(𝑁 )] distance computations.
Though the KD tree approach is very fast for low-dimensional (𝐷 < 20) neighbors searches, it becomes inefficient
as 𝐷 grows very large: this is one manifestation of the so-called “curse of dimensionality”. In scikit-learn, KD tree
neighbors searches are specified using the keyword algorithm = 'kd_tree', and are computed using the class
KDTree.
References:
• “Multidimensional binary search trees used for associative searching”, Bentley, J.L., Communications of the
ACM (1975)
Ball Tree
To address the inefficiencies of KD Trees in higher dimensions, the ball tree data structure was developed. Where KD
trees partition data along Cartesian axes, ball trees partition data in a series of nesting hyper-spheres. This makes tree
construction more costly than that of the KD tree, but results in a data structure which can be very efficient on highly
structured data, even in very high dimensions.
A ball tree recursively divides the data into nodes defined by a centroid 𝐶 and radius 𝑟, such that each point in the
node lies within the hyper-sphere defined by 𝑟 and 𝐶. The number of candidate points for a neighbor search is reduced

through use of the triangle inequality:
|𝑥 + 𝑦| ≤ |𝑥| + |𝑦|
With this setup, a single distance calculation between a test point and the centroid is sufficient to determine a lower
and upper bound on the distance to all points within the node. Because of the spherical geometry of the ball tree nodes,
it can out-perform a KD-tree in high dimensions, though the actual performance is highly dependent on the structure
of the training data. In scikit-learn, ball-tree-based neighbors searches are specified using the keyword algorithm
= 'ball_tree', and are computed using the class sklearn.neighbors.BallTree. Alternatively, the user
can work with the BallTree class directly.
References:
• “Five balltree construction algorithms”, Omohundro, S.M., International Computer Science Institute Techni-
cal Report (1989)
Choice of Nearest Neighbors Algorithm
The optimal algorithm for a given dataset is a complicated choice, and depends on a number of factors:
• number of samples 𝑁 (i.e. n_samples) and dimensionality 𝐷 (i.e. n_features).
– Brute force query time grows as 𝑂[𝐷𝑁 ]
– Ball tree query time grows as approximately 𝑂[𝐷 log(𝑁 )]
– KD tree query time changes with 𝐷 in a way that is difficult to precisely characterise. For small 𝐷 (less
than 20 or so) the cost is approximately 𝑂[𝐷 log(𝑁 )], and the KD tree query can be very efficient. For
larger 𝐷, the cost increases to nearly 𝑂[𝐷𝑁 ], and the overhead due to the tree structure can lead to queries
which are slower than brute force.
For small data sets (𝑁 less than 30 or so), log(𝑁 ) is comparable to 𝑁 , and brute force algorithms can be more
efficient than a tree-based approach. Both KDTree and BallTree address this through providing a leaf size
parameter: this controls the number of samples at which a query switches to brute-force. This allows both
algorithms to approach the efficiency of a brute-force computation for small 𝑁 .
• data structure: intrinsic dimensionality of the data and/or sparsity of the data. Intrinsic dimensionality refers
to the dimension 𝑑 ≤ 𝐷 of a manifold on which the data lies, which can be linearly or non-linearly embedded
in the parameter space. Sparsity refers to the degree to which the data fills the parameter space (this is to be
distinguished from the concept as used in “sparse” matrices. The data matrix may have no zero entries, but the
structure can still be “sparse” in this sense).
– Brute force query time is unchanged by data structure.
– Ball tree and KD tree query times can be greatly influenced by data structure. In general, sparser data with a
smaller intrinsic dimensionality leads to faster query times. Because the KD tree internal representation is
aligned with the parameter axes, it will not generally show as much improvement as ball tree for arbitrarily
structured data.
Datasets used in machine learning tend to be very structured, and are very well-suited for tree-based queries.
• number of neighbors 𝑘 requested for a query point.
– Brute force query time is largely unaffected by the value of 𝑘
– Ball tree and KD tree query time will become slower as 𝑘 increases. This is due to two effects: first, a
larger 𝑘 leads to the necessity to search a larger portion of the parameter space. Second, using 𝑘 > 1
requires internal queueing of results as the tree is traversed.

As 𝑘 becomes large compared to 𝑁 , the ability to prune branches in a tree-based query is reduced. In this
situation, Brute force queries can be more efficient.
• number of query points. Both the ball tree and the KD Tree require a construction phase. The cost of this
construction becomes negligible when amortized over many queries. If only a small number of queries will
be performed, however, the construction can make up a significant fraction of the total cost. If very few query
points will be required, brute force is better than a tree-based method.
Currently, algorithm = 'auto' selects 'kd_tree' if 𝑘 < 𝑁/2 and the 'effective_metric_'
is in the 'VALID_METRICS' list of 'kd_tree'. It selects 'ball_tree' if 𝑘 < 𝑁/2 and the
'effective_metric_' is in the 'VALID_METRICS' list of 'ball_tree'. It selects 'brute' if 𝑘 < 𝑁/2
and the 'effective_metric_' is not in the 'VALID_METRICS' list of 'kd_tree' or 'ball_tree'. It
selects 'brute' if 𝑘 >= 𝑁/2. This choice is based on the assumption that the number of query points is at least the
same order as the number of training points, and that leaf_size is close to its default value of 30.
Effect of leaf_size
As noted above, for small sample sizes a brute force search can be more efficient than a tree-based query. This fact is
accounted for in the ball tree and KD tree by internally switching to brute force searches within leaf nodes. The level
of this switch can be specified with the parameter leaf_size. This parameter choice has many effects:
construction time A larger leaf_size leads to a faster tree construction time, because fewer nodes need to be
created
query time Both a large or small leaf_size can lead to suboptimal query cost. For leaf_size approaching
1, the overhead involved in traversing nodes can significantly slow query times. For leaf_size approach-
ing the size of the training set, queries become essentially brute force. A good compromise between these is
leaf_size = 30, the default value of the parameter.
memory As leaf_size increases, the memory required to store a tree structure decreases. This is especially
important in the case of ball tree, which stores a 𝐷-dimensional centroid for each node. The required storage
space for BallTree is approximately 1 / leaf_size times the size of the training set.
leaf_size is not referenced for brute force queries.
Nearest Centroid Classifier
The NearestCentroid classifier is a simple algorithm that represents each class by the centroid of its members.
In effect, this makes it similar to the label updating phase of the sklearn.KMeans algorithm. It also has no param-
eters to choose, making it a good baseline classifier. It does, however, suffer on non-convex classes, as well as when
classes have drastically different variances, as equal variance in all dimensions is assumed. See Linear Discrim-
inant Analysis (sklearn.discriminant_analysis.LinearDiscriminantAnalysis) and Quadratic
Discriminant Analysis (sklearn.discriminant_analysis.QuadraticDiscriminantAnalysis) for
more complex methods that do not make this assumption. Usage of the default NearestCentroid is simple:
>>> from sklearn.neighbors.nearest_centroid import NearestCentroid

>>> X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]])
>>> y = np.array([1, 1, 1, 2, 2, 2])
>>> clf = NearestCentroid()
>>> clf.fit(X, y)
NearestCentroid(metric='euclidean', shrink_threshold=None)
>>> print(clf.predict([[-0.8, -1]]))
[1]

Nearest Shrunken Centroid
The NearestCentroid classifier has a shrink_threshold parameter, which implements the nearest shrunken
centroid classifier. In effect, the value of each feature for each centroid is divided by the within-class variance of that
feature. The feature values are then reduced by shrink_threshold. Most notably, if a particular feature value
crosses zero, it is set to zero. In effect, this removes the feature from affecting the classification. This is useful, for
example, for removing noisy features.
In the example below, using a small shrink threshold increases the accuracy of the model from 0.81 to 0.82.
Examples:
• Nearest Centroid Classification: an example of classification using nearest centroid with different shrink
thresholds.
3.1.7 Gaussian Processes
Gaussian Processes (GP) are a generic supervised learning method designed to solve regression and probabilistic
classification problems.
The advantages of Gaussian processes are:
• The prediction interpolates the observations (at least for regular kernels).
• The prediction is probabilistic (Gaussian) so that one can compute empirical confidence intervals and decide
based on those if one should refit (online fitting, adaptive fitting) the prediction in some region of interest.
• Versatile: different kernels can be specified. Common kernels are provided, but it is also possible to specify
custom kernels.
The disadvantages of Gaussian processes include:
• They are not sparse, i.e., they use the whole samples/features information to perform the prediction.
• They lose efficiency in high dimensional spaces – namely when the number of features exceeds a few dozens.
Gaussian Process Regression (GPR)
The GaussianProcessRegressor implements Gaussian processes (GP) for regression purposes. For this, the
prior of the GP needs to be specified. The prior mean is assumed to be constant and zero (for normalize_y=False)

or the training data’s mean (for normalize_y=True). The prior’s covariance is specified by passing a kernel object.
The hyperparameters of the kernel are optimized during fitting of GaussianProcessRegressor by maximizing the log-
marginal-likelihood (LML) based on the passed optimizer. As the LML may have multiple local optima, the
optimizer can be started repeatedly by specifying n_restarts_optimizer. The first run is always conducted
starting from the initial hyperparameter values of the kernel; subsequent runs are conducted from hyperparameter
values that have been chosen randomly from the range of allowed values. If the initial hyperparameters should be kept
fixed, None can be passed as optimizer.
The noise level in the targets can be specified by passing it via the parameter alpha, either globally as a scalar or
per datapoint. Note that a moderate noise level can also be helpful for dealing with numeric issues during fitting as
it is effectively implemented as Tikhonov regularization, i.e., by adding it to the diagonal of the kernel matrix. An
alternative to specifying the noise level explicitly is to include a WhiteKernel component into the kernel, which can
estimate the global noise level from the data (see example below).
The implementation is based on Algorithm 2.1 of [RW2006]. In addition to the API of standard scikit-learn estimators,
GaussianProcessRegressor:
• allows prediction without prior fitting (based on the GP prior)
• provides an additional method sample_y(X), which evaluates samples drawn from the GPR (prior or poste-
rior) at given inputs
• exposes a method log_marginal_likelihood(theta), which can be used externally for other ways of
selecting hyperparameters, e.g., via Markov chain Monte Carlo.
GPR examples
GPR with noise-level estimation
This example illustrates that GPR with a sum-kernel including a WhiteKernel can estimate the noise level of data. An
illustration of the log-marginal-likelihood (LML) landscape shows that there exist two local maxima of LML.
The first corresponds to a model with a high noise level and a large length scale, which explains all variations in the
data by noise.
The second one has a smaller noise level and shorter length scale, which explains most of the variation by the noise-
free functional relationship. The second model has a higher likelihood; however, depending on the initial value for the
hyperparameters, the gradient-based optimization might also converge to the high-noise solution. It is thus important
to repeat the optimization several times for different initializations.
Comparison of GPR and Kernel Ridge Regression
Both kernel ridge regression (KRR) and GPR learn a target function by employing internally the “kernel trick”. KRR
learns a linear function in the space induced by the respective kernel which corresponds to a non-linear function in
the original space. The linear function in the kernel space is chosen based on the mean-squared error loss with ridge
regularization. GPR uses the kernel to define the covariance of a prior distribution over the target functions and uses
the observed training data to define a likelihood function. Based on Bayes theorem, a (Gaussian) posterior distribution
over target functions is defined, whose mean is used for prediction.
A major difference is that GPR can choose the kernel’s hyperparameters based on gradient-ascent on the marginal
likelihood function while KRR needs to perform a grid search on a cross-validated loss function (mean-squared error
loss). A further difference is that GPR learns a generative, probabilistic model of the target function and can thus
provide meaningful confidence intervals and posterior samples along with the predictions while KRR only provides
predictions.




The following figure illustrates both methods on an artificial dataset, which consists of a sinusoidal target function
and strong noise. The figure compares the learned model of KRR and GPR based on a ExpSineSquared kernel,
which is suited for learning periodic functions. The kernel’s hyperparameters control the smoothness (length_scale)
and periodicity of the kernel (periodicity). Moreover, the noise level of the data is learned explicitly by GPR by an
additional WhiteKernel component in the kernel and by the regularization parameter alpha of KRR.
The figure shows that both methods learn reasonable models of the target function. GPR correctly identifies the peri-
odicity of the function to be roughly 2 * 𝜋 (6.28), while KRR chooses the doubled periodicity 4 * 𝜋 . Besides that, GPR
provides reasonable confidence bounds on the prediction which are not available for KRR. A major difference between
the two methods is the time required for fitting and predicting: while fitting KRR is fast in principle, the grid-search
for hyperparameter optimization scales exponentially with the number of hyperparameters (“curse of dimensional-
ity”). The gradient-based optimization of the parameters in GPR does not suffer from this exponential scaling and is
thus considerable faster on this example with 3-dimensional hyperparameter space. The time for predicting is similar;
however, generating the variance of the predictive distribution of GPR takes considerable longer than just predicting
the mean.
GPR on Mauna Loa CO2 data
This example is based on Section 5.4.3 of [RW2006]. It illustrates an example of complex kernel engineering and
hyperparameter optimization using gradient ascent on the log-marginal-likelihood. The data consists of the monthly
average atmospheric CO2 concentrations (in parts per million by volume (ppmv)) collected at the Mauna Loa Obser-
vatory in Hawaii, between 1958 and 1997. The objective is to model the CO2 concentration as a function of the time
t.
The kernel is composed of several terms that are responsible for explaining different properties of the signal:
• a long term, smooth rising trend is to be explained by an RBF kernel. The RBF kernel with a large length-scale
enforces this component to be smooth; it is not enforced that the trend is rising which leaves this choice to the
GP. The specific length-scale and the amplitude are free hyperparameters.
• a seasonal component, which is to be explained by the periodic ExpSineSquared kernel with a fixed periodicity
of 1 year. The length-scale of this periodic component, controlling its smoothness, is a free parameter. In order
to allow decaying away from exact periodicity, the product with an RBF kernel is taken. The length-scale of this
RBF component controls the decay time and is a further free parameter.

• smaller, medium term irregularities are to be explained by a RationalQuadratic kernel component, whose length-
scale and alpha parameter, which determines the diffuseness of the length-scales, are to be determined. Ac-
cording to [RW2006], these irregularities can better be explained by a RationalQuadratic than an RBF kernel
component, probably because it can accommodate several length-scales.
• a “noise” term, consisting of an RBF kernel contribution, which shall explain the correlated noise components
such as local weather phenomena, and a WhiteKernel contribution for the white noise. The relative amplitudes
and the RBF’s length scale are further free parameters.
Maximizing the log-marginal-likelihood after subtracting the target’s mean yields the following kernel with an LML
of -83.214:
34.4**2 * RBF(length_scale=41.8)
+ 3.27**2 * RBF(length_scale=180) * ExpSineSquared(length_scale=1.44,
periodicity=1)
+ 0.446**2 * RationalQuadratic(alpha=17.7, length_scale=0.957)
+ 0.197**2 * RBF(length_scale=0.138) + WhiteKernel(noise_level=0.0336)
Thus, most of the target signal (34.4ppm) is explained by a long-term rising trend (length-scale 41.8 years). The
periodic component has an amplitude of 3.27ppm, a decay time of 180 years and a length-scale of 1.44. The long
decay time indicates that we have a locally very close to periodic seasonal component. The correlated noise has an
amplitude of 0.197ppm with a length scale of 0.138 years and a white-noise contribution of 0.197ppm. Thus, the
overall noise level is very small, indicating that the data can be very well explained by the model. The figure shows
also that the model makes very confident predictions until around 2015

Gaussian Process Classification (GPC)
The GaussianProcessClassifier implements Gaussian processes (GP) for classification purposes, more
specifically for probabilistic classification, where test predictions take the form of class probabilities. GaussianPro-
cessClassifier places a GP prior on a latent function 𝑓 , which is then squashed through a link function to obtain the
probabilistic classification. The latent function 𝑓 is a so-called nuisance function, whose values are not observed and
are not relevant by themselves. Its purpose is to allow a convenient formulation of the model, and 𝑓 is removed (inte-
grated out) during prediction. GaussianProcessClassifier implements the logistic link function, for which the integral
cannot be computed analytically but is easily approximated in the binary case.
In contrast to the regression setting, the posterior of the latent function 𝑓 is not Gaussian even for a GP prior since
a Gaussian likelihood is inappropriate for discrete class labels. Rather, a non-Gaussian likelihood corresponding to
the logistic link function (logit) is used. GaussianProcessClassifier approximates the non-Gaussian posterior with a
Gaussian based on the Laplace approximation. More details can be found in Chapter 3 of [RW2006].
The GP prior mean is assumed to be zero. The prior’s covariance is specified by passing a kernel object. The hyper-
parameters of the kernel are optimized during fitting of GaussianProcessRegressor by maximizing the log-marginal-
likelihood (LML) based on the passed optimizer. As the LML may have multiple local optima, the optimizer can
be started repeatedly by specifying n_restarts_optimizer. The first run is always conducted starting from the
initial hyperparameter values of the kernel; subsequent runs are conducted from hyperparameter values that have been
chosen randomly from the range of allowed values. If the initial hyperparameters should be kept fixed, None can be
passed as optimizer.
GaussianProcessClassifier supports multi-class classification by performing either one-versus-rest or one-
versus-one based training and prediction. In one-versus-rest, one binary Gaussian process classifier is fitted for each
class, which is trained to separate this class from the rest. In “one_vs_one”, one binary Gaussian process classifier is
fitted for each pair of classes, which is trained to separate these two classes. The predictions of these binary predictors
are combined into multi-class predictions. See the section on multi-class classification for more details.
In the case of Gaussian process classification, “one_vs_one” might be computationally cheaper since it has to solve
many problems involving only a subset of the whole training set rather than fewer problems on the whole dataset. Since
Gaussian process classification scales cubically with the size of the dataset, this might be considerably faster. How-
ever, note that “one_vs_one” does not support predicting probability estimates but only plain predictions. Moreover,
note that GaussianProcessClassifier does not (yet) implement a true multi-class Laplace approximation in-
ternally, but as discussed above is based on solving several binary classification tasks internally, which are combined
using one-versus-rest or one-versus-one.
GPC examples
Probabilistic predictions with GPC
This example illustrates the predicted probability of GPC for an RBF kernel with different choices of the hyperparam-
eters. The first figure shows the predicted probability of GPC with arbitrarily chosen hyperparameters and with the
hyperparameters corresponding to the maximum log-marginal-likelihood (LML).
While the hyperparameters chosen by optimizing LML have a considerable larger LML, they perform slightly worse
according to the log-loss on test data. The figure shows that this is because they exhibit a steep change of the class
probabilities at the class boundaries (which is good) but have predicted probabilities close to 0.5 far away from the
class boundaries (which is bad) This undesirable effect is caused by the Laplace approximation used internally by
GPC.
The second figure shows the log-marginal-likelihood for different choices of the kernel’s hyperparameters, highlighting
the two choices of the hyperparameters used in the first figure by black dots.



Illustration of GPC on the XOR dataset
This example illustrates GPC on XOR data. Compared are a stationary, isotropic kernel (RBF) and a non-stationary
kernel (DotProduct). On this particular dataset, the DotProduct kernel obtains considerably better results because
the class-boundaries are linear and coincide with the coordinate axes. In practice, however, stationary kernels such as
RBF often obtain better results.
Gaussian process classification (GPC) on iris dataset
This example illustrates the predicted probability of GPC for an isotropic and anisotropic RBF kernel on a two-
dimensional version for the iris-dataset. This illustrates the applicability of GPC to non-binary classification. The
anisotropic RBF kernel obtains slightly higher log-marginal-likelihood by assigning different length-scales to the two
feature dimensions.
Kernels for Gaussian Processes
Kernels (also called “covariance functions” in the context of GPs) are a crucial ingredient of GPs which determine
the shape of prior and posterior of the GP. They encode the assumptions on the function being learned by defining the
“similarity” of two datapoints combined with the assumption that similar datapoints should have similar target values.
Two categories of kernels can be distinguished: stationary kernels depend only on the distance of two datapoints
and not on their absolute values 𝑘(𝑥𝑖 , 𝑥𝑗 ) = 𝑘(𝑑(𝑥𝑖 , 𝑥𝑗 )) and are thus invariant to translations in the input space,
while non-stationary kernels depend also on the specific values of the datapoints. Stationary kernels can further be
subdivided into isotropic and anisotropic kernels, where isotropic kernels are also invariant to rotations in the input
space. For more details, we refer to Chapter 4 of [RW2006].
Gaussian Process Kernel API
The main usage of a Kernel is to compute the GP’s covariance between datapoints. For this, the method __call__
of the kernel can be called. This method can either be used to compute the “auto-covariance” of all pairs of datapoints

in a 2d array X, or the “cross-covariance” of all combinations of datapoints of a 2d array X with datapoints in a 2d

array Y. The following identity holds true for all kernels k (except for the WhiteKernel): k(X) == K(X, Y=X)
If only the diagonal of the auto-covariance is being used, the method diag() of a kernel can be called, which is more
computationally efficient than the equivalent call to __call__: np.diag(k(X, X)) == k.diag(X)
Kernels are parameterized by a vector 𝜃 of hyperparameters. These hyperparameters can for instance control length-
scales or periodicity of a kernel (see below). All kernels support computing analytic gradients of the kernel’s auto-
covariance with respect to 𝜃 via setting eval_gradient=True in the __call__ method. This gradient is used by
the Gaussian process (both regressor and classifier) in computing the gradient of the log-marginal-likelihood, which
in turn is used to determine the value of 𝜃, which maximizes the log-marginal-likelihood, via gradient ascent. For
each hyperparameter, the initial value and the bounds need to be specified when creating an instance of the kernel.
The current value of 𝜃 can be get and set via the property theta of the kernel object. Moreover, the bounds of the
hyperparameters can be accessed by the property bounds of the kernel. Note that both properties (theta and bounds)
return log-transformed values of the internally used values since those are typically more amenable to gradient-based
optimization. The specification of each hyperparameter is stored in the form of an instance of Hyperparameter
in the respective kernel. Note that a kernel using a hyperparameter with name “x” must have the attributes self.x and
self.x_bounds.
The abstract base class for all kernels is Kernel. Kernel implements a similar interface as Estimator, providing
the methods get_params(), set_params(), and clone(). This allows setting kernel values also via meta-
estimators such as Pipeline or GridSearch. Note that due to the nested structure of kernels (by applying kernel
operators, see below), the names of kernel parameters might become relatively complicated. In general, for a binary
kernel operator, parameters of the left operand are prefixed with k1__ and parameters of the right operand with k2__.
An additional convenience method is clone_with_theta(theta), which returns a cloned version of the kernel
but with the hyperparameters set to theta. An illustrative example:
>>> from sklearn.gaussian_process.kernels import ConstantKernel, RBF
>>> kernel = ConstantKernel(constant_value=1.0, constant_value_bounds=(0.0, 10.0)) *
˓→RBF(length_scale=0.5, length_scale_bounds=(0.0, 10.0)) + RBF(length_scale=2.0,
˓→length_scale_bounds=(0.0, 10.0))
>>> for hyperparameter in kernel.hyperparameters: print(hyperparameter)

Hyperparameter(name='k1__k1__constant_value', value_type='numeric', bounds=array([[ 0.
˓→, 10.]]), n_elements=1, fixed=False)
Hyperparameter(name='k1__k2__length_scale', value_type='numeric', bounds=array([[ 0.,

˓→10.]]), n_elements=1, fixed=False)

Hyperparameter(name='k2__length_scale', value_type='numeric', bounds=array([[ 0., 10.

˓→]]), n_elements=1, fixed=False)
>>> params = kernel.get_params()

>>> for key in sorted(params): print("%s : %s" % (key, params[key]))
k1 : 1**2 * RBF(length_scale=0.5)
k1__k1 : 1**2
k1__k1__constant_value : 1.0
k1__k1__constant_value_bounds : (0.0, 10.0)
k1__k2 : RBF(length_scale=0.5)
k1__k2__length_scale : 0.5
k1__k2__length_scale_bounds : (0.0, 10.0)
k2 : RBF(length_scale=2)
k2__length_scale : 2.0
k2__length_scale_bounds : (0.0, 10.0)
>>> print(kernel.theta) # Note: log-transformed
[ 0. -0.69314718 0.69314718]
>>> print(kernel.bounds) # Note: log-transformed
[[ -inf 2.30258509]
[ -inf 2.30258509]
[ -inf 2.30258509]]
All Gaussian process kernels are interoperable with sklearn.metrics.pairwise and vice versa: instances
of subclasses of Kernel can be passed as metric to pairwise_kernels from sklearn.metrics.
pairwise. Moreover, kernel functions from pairwise can be used as GP kernels by using the wrapper class
PairwiseKernel. The only caveat is that the gradient of the hyperparameters is not analytic but numeric and
all those kernels support only isotropic distances. The parameter gamma is considered to be a hyperparameter and
may be optimized. The other kernel parameters are set directly at initialization and are kept fixed.
Basic kernels
The ConstantKernel kernel can be used as part of a Product kernel where it scales the magnitude of the other
factor (kernel) or as part of a Sum kernel, where it modifies the mean of the Gaussian process. It depends on a
parameter 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡_𝑣𝑎𝑙𝑢𝑒. It is defined as:
𝑘(𝑥𝑖 , 𝑥𝑗 ) = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡_𝑣𝑎𝑙𝑢𝑒 ∀ 𝑥1 , 𝑥2
The main use-case of the WhiteKernel kernel is as part of a sum-kernel where it explains the noise-component of
the signal. Tuning its parameter 𝑛𝑜𝑖𝑠𝑒_𝑙𝑒𝑣𝑒𝑙 corresponds to estimating the noise-level. It is defined as:
𝑘(𝑥𝑖 , 𝑥𝑗 ) = 𝑛𝑜𝑖𝑠𝑒_𝑙𝑒𝑣𝑒𝑙 if 𝑥𝑖 == 𝑥𝑗 else 0
Kernel operators
Kernel operators take one or two base kernels and combine them into a new kernel. The Sum kernel takes two kernels
𝑘1 and 𝑘2 and combines them via 𝑘𝑠𝑢𝑚 (𝑋, 𝑌 ) = 𝑘1(𝑋, 𝑌 ) + 𝑘2(𝑋, 𝑌 ). The Product kernel takes two kernels 𝑘1
and 𝑘2 and combines them via 𝑘𝑝𝑟𝑜𝑑𝑢𝑐𝑡 (𝑋, 𝑌 ) = 𝑘1(𝑋, 𝑌 ) * 𝑘2(𝑋, 𝑌 ). The Exponentiation kernel takes one
base kernel and a scalar parameter 𝑒𝑥𝑝𝑜𝑛𝑒𝑛𝑡 and combines them via 𝑘𝑒𝑥𝑝 (𝑋, 𝑌 ) = 𝑘(𝑋, 𝑌 )exponent .
Radial-basis function (RBF) kernel
The RBF kernel is a stationary kernel. It is also known as the “squared exponential” kernel. It is parameterized by a
length-scale parameter 𝑙 > 0, which can either be a scalar (isotropic variant of the kernel) or a vector with the same

number of dimensions as the inputs 𝑥 (anisotropic variant of the kernel). The kernel is given by:
(︂ )︂
1 2
𝑘(𝑥𝑖 , 𝑥𝑗 ) = exp − 𝑑(𝑥𝑖 /𝑙, 𝑥𝑗 /𝑙)
2
This kernel is infinitely differentiable, which implies that GPs with this kernel as covariance function have mean square
derivatives of all orders, and are thus very smooth. The prior and posterior of a GP resulting from an RBF kernel are
shown in the following figure:
Matérn kernel
The Matern kernel is a stationary kernel and a generalization of the RBF kernel. It has an additional parameter 𝜈
which controls the smoothness of the resulting function. It is parameterized by a length-scale parameter 𝑙 > 0, which

can either be a scalar (isotropic variant of the kernel) or a vector with the same number of dimensions as the inputs 𝑥
(anisotropic variant of the kernel). The kernel is given by:
(︃ )︃𝜈 (︃ )︃
1 √ √
𝑘(𝑥𝑖 , 𝑥𝑗 ) = 𝜎 2 𝛾 2𝜈𝑑(𝑥𝑖 /𝑙, 𝑥𝑗 /𝑙) 𝐾𝜈 𝛾 2𝜈𝑑(𝑥𝑖 /𝑙, 𝑥𝑗 /𝑙) ,
Γ(𝜈)2𝜈−1
As 𝜈 → ∞, the Matérn kernel converges to the RBF kernel. When 𝜈 = 1/2, the Matérn kernel becomes identical to
the absolute exponential kernel, i.e.,
(︃ )︃
𝑘(𝑥𝑖 , 𝑥𝑗 ) = 𝜎 2 exp − 𝛾𝑑(𝑥𝑖 /𝑙, 𝑥𝑗 /𝑙) 𝜈= 1
2
In particular, 𝜈 = 3/2:
(︃ )︃ (︃ )︃
√ √
𝑘(𝑥𝑖 , 𝑥𝑗 ) = 𝜎 2 1 + 𝛾 3𝑑(𝑥𝑖 /𝑙, 𝑥𝑗 /𝑙) exp − 𝛾 3𝑑(𝑥𝑖 /𝑙, 𝑥𝑗 /𝑙) 𝜈= 3
2
and 𝜈 = 5/2:
(︃ )︃ (︃ )︃
2
√ 5 √
𝑘(𝑥𝑖 , 𝑥𝑗 ) = 𝜎 1 + 𝛾 5𝑑(𝑥𝑖 /𝑙, 𝑥𝑗 /𝑙) + 𝛾 2 𝑑(𝑥𝑖 /𝑙, 𝑥𝑗 /𝑙)2 exp − 𝛾 5𝑑(𝑥𝑖 /𝑙, 𝑥𝑗 /𝑙) 𝜈= 5
2
3
are popular choices for learning functions that are not infinitely differentiable (as assumed by the RBF kernel) but at
least once (𝜈 = 3/2) or twice differentiable (𝜈 = 5/2).
The flexibility of controlling the smoothness of the learned function via 𝜈 allows adapting to the properties of the
true underlying functional relation. The prior and posterior of a GP resulting from a Matérn kernel are shown in the
following figure:
See [RW2006], pp84 for further details regarding the different variants of the Matérn kernel.
Rational quadratic kernel
The RationalQuadratic kernel can be seen as a scale mixture (an infinite sum) of RBF kernels with different
characteristic length-scales. It is parameterized by a length-scale parameter 𝑙 > 0 and a scale mixture parameter 𝛼 > 0
Only the isotropic variant where 𝑙 is a scalar is supported at the moment. The kernel is given by:
)︂−𝛼
𝑑(𝑥𝑖 , 𝑥𝑗 )2
(︂
𝑘(𝑥𝑖 , 𝑥𝑗 ) = 1+
2𝛼𝑙2
The prior and posterior of a GP resulting from a RationalQuadratic kernel are shown in the following figure:
Exp-Sine-Squared kernel
The ExpSineSquared kernel allows modeling periodic functions. It is parameterized by a length-scale parameter
𝑙 > 0 and a periodicity parameter 𝑝 > 0. Only the isotropic variant where 𝑙 is a scalar is supported at the moment.
The kernel is given by:
(︁ )︁
2
𝑘(𝑥𝑖 , 𝑥𝑗 ) = exp −2 (sin(𝜋/𝑝 * 𝑑(𝑥𝑖 , 𝑥𝑗 ))/𝑙)
The prior and posterior of a GP resulting from an ExpSineSquared kernel are shown in the following figure:




Dot-Product kernel
The DotProduct kernel is non-stationary and can be obtained from linear regression by putting 𝑁 (0, 1) priors on
the coefficients of 𝑥𝑑 (𝑑 = 1, ..., 𝐷) and a prior of 𝑁 (0, 𝜎02 ) on the bias. The DotProduct kernel is invariant to a
rotation of the coordinates about the origin, but not translations. It is parameterized by a parameter 𝜎02 . For 𝜎02 = 0,
the kernel is called the homogeneous linear kernel, otherwise it is inhomogeneous. The kernel is given by
𝑘(𝑥𝑖 , 𝑥𝑗 ) = 𝜎02 + 𝑥𝑖 · 𝑥𝑗
The DotProduct kernel is commonly combined with exponentiation. An example with exponent 2 is shown in the
following figure:

References
3.1.8 Cross decomposition
The cross decomposition module contains two main families of algorithms: the partial least squares (PLS) and the
canonical correlation analysis (CCA).
These families of algorithms are useful to find linear relations between two multivariate datasets: the X and Y argu-
ments of the fit method are 2D arrays.
Cross decomposition algorithms find the fundamental relations between two matrices (X and Y). They are latent
variable approaches to modeling the covariance structures in these two spaces. They will try to find the multidi-
mensional direction in the X space that explains the maximum multidimensional variance direction in the Y space.
PLS-regression is particularly suited when the matrix of predictors has more variables than observations, and when
there is multicollinearity among X values. By contrast, standard regression will fail in these cases.
Classes included in this module are PLSRegression PLSCanonical, CCA and PLSSVD
Reference:
• JA Wegelin A survey of Partial Least Squares (PLS) methods, with emphasis on the two-block case
Examples:

• Compare cross decomposition methods
3.1.9 Naive Bayes
Naive Bayes methods are a set of supervised learning algorithms based on applying Bayes’ theorem with the “naive”
assumption of conditional independence between every pair of features given the value of the class variable. Bayes’
theorem states the following relationship, given class variable 𝑦 and dependent feature vector 𝑥1 through 𝑥𝑛 , :
𝑃 (𝑦)𝑃 (𝑥1 , . . . 𝑥𝑛 | 𝑦)
𝑃 (𝑦 | 𝑥1 , . . . , 𝑥𝑛 ) =
𝑃 (𝑥1 , . . . , 𝑥𝑛 )
Using the naive conditional independence assumption that
𝑃 (𝑥𝑖 |𝑦, 𝑥1 , . . . , 𝑥𝑖−1 , 𝑥𝑖+1 , . . . , 𝑥𝑛 ) = 𝑃 (𝑥𝑖 |𝑦),
for all 𝑖, this relationship is simplified to

∏︀𝑛
𝑃 (𝑦) 𝑖=1 𝑃 (𝑥𝑖 | 𝑦)
𝑃 (𝑦 | 𝑥1 , . . . , 𝑥𝑛 ) =
𝑃 (𝑥1 , . . . , 𝑥𝑛 )
Since 𝑃 (𝑥1 , . . . , 𝑥𝑛 ) is constant given the input, we can use the following classification rule:
𝑛
∏︁
𝑃 (𝑦 | 𝑥1 , . . . , 𝑥𝑛 ) ∝ 𝑃 (𝑦) 𝑃 (𝑥𝑖 | 𝑦)
𝑖=1
⇓
𝑛
∏︁
𝑦ˆ = arg max 𝑃 (𝑦) 𝑃 (𝑥𝑖 | 𝑦),
𝑦
𝑖=1
and we can use Maximum A Posteriori (MAP) estimation to estimate 𝑃 (𝑦) and 𝑃 (𝑥𝑖 | 𝑦); the former is then the
relative frequency of class 𝑦 in the training set.
The different naive Bayes classifiers differ mainly by the assumptions they make regarding the distribution of 𝑃 (𝑥𝑖 |
𝑦).
In spite of their apparently over-simplified assumptions, naive Bayes classifiers have worked quite well in many real-
world situations, famously document classification and spam filtering. They require a small amount of training data to
estimate the necessary parameters. (For theoretical reasons why naive Bayes works well, and on which types of data
it does, see the references below.)
Naive Bayes learners and classifiers can be extremely fast compared to more sophisticated methods. The decoupling
of the class conditional feature distributions means that each distribution can be independently estimated as a one
dimensional distribution. This in turn helps to alleviate problems stemming from the curse of dimensionality.
On the flip side, although naive Bayes is known as a decent classifier, it is known to be a bad estimator, so the
probability outputs from predict_proba are not to be taken too seriously.
References:
• H. Zhang (2004). The optimality of Naive Bayes. Proc. FLAIRS.

Gaussian Naive Bayes
GaussianNB implements the Gaussian Naive Bayes algorithm for classification. The likelihood of the features is
assumed to be Gaussian:
(𝑥𝑖 − 𝜇𝑦 )2
(︂ )︂
1
𝑃 (𝑥𝑖 | 𝑦) = exp −
2𝜎𝑦2
√︁
2𝜋𝜎𝑦2
The parameters 𝜎𝑦 and 𝜇𝑦 are estimated using maximum likelihood.

>>> from sklearn.naive_bayes import GaussianNB
>>> gnb = GaussianNB()
>>> y_pred = gnb.fit(iris.data, iris.target).predict(iris.data)
>>> print("Number of mislabeled points out of a total %d points : %d"
... % (iris.data.shape[0],(iris.target != y_pred).sum()))
Number of mislabeled points out of a total 150 points : 6
Multinomial Naive Bayes
MultinomialNB implements the naive Bayes algorithm for multinomially distributed data, and is one of the two
classic naive Bayes variants used in text classification (where the data are typically represented as word vector counts,
although tf-idf vectors are also known to work well in practice). The distribution is parametrized by vectors 𝜃𝑦 =
(𝜃𝑦1 , . . . , 𝜃𝑦𝑛 ) for each class 𝑦, where 𝑛 is the number of features (in text classification, the size of the vocabulary)
and 𝜃𝑦𝑖 is the probability 𝑃 (𝑥𝑖 | 𝑦) of feature 𝑖 appearing in a sample belonging to class 𝑦.
The parameters 𝜃𝑦 is estimated by a smoothed version of maximum likelihood, i.e. relative frequency counting:
𝑁𝑦𝑖 + 𝛼
𝜃ˆ𝑦𝑖 =
𝑁𝑦 + 𝛼𝑛
∑︀
𝑁𝑦𝑖 =
where ∑︀ 𝑥∈𝑇 𝑥𝑖 is the number of times feature 𝑖 appears in a sample of class 𝑦 in the training set 𝑇 , and
𝑛
𝑁𝑦 = 𝑖=1 𝑁𝑦𝑖 is the total count of all features for class 𝑦.
The smoothing priors 𝛼 ≥ 0 accounts for features not present in the learning samples and prevents zero probabilities
in further computations. Setting 𝛼 = 1 is called Laplace smoothing, while 𝛼 < 1 is called Lidstone smoothing.
Complement Naive Bayes
ComplementNB implements the complement naive Bayes (CNB) algorithm. CNB is an adaptation of the standard
multinomial naive Bayes (MNB) algorithm that is particularly suited for imbalanced data sets. Specifically, CNB uses
statistics from the complement of each class to compute the model’s weights. The inventors of CNB show empirically
that the parameter estimates for CNB are more stable than those for MNB. Further, CNB regularly outperforms MNB
(often by a considerable margin) on text classification tasks. The procedure for calculating the weights is as follows:
∑︀
𝛼𝑖 + 𝑗:𝑦𝑗 ̸=𝑐 𝑑𝑖𝑗
𝜃ˆ𝑐𝑖 = ∑︀ ∑︀
𝛼 + 𝑗:𝑦𝑗 ̸=𝑐 𝑘 𝑑𝑘𝑗
𝑤𝑐𝑖 = log 𝜃ˆ𝑐𝑖
𝑤𝑐𝑖
𝑤𝑐𝑖 = ∑︀
𝑗 |𝑤𝑐𝑗 |
where the summations are over all documents 𝑗 not in class 𝑐, 𝑑𝑖𝑗 is either the ∑︀
count or tf-idf value of term 𝑖 in
document 𝑗, 𝛼𝑖 is a smoothing hyperparameter like that found in MNB, and 𝛼 = 𝑖 𝛼𝑖 . The second normalization

addresses the tendency for longer documents to dominate parameter estimates in MNB. The classification rule is:
∑︁
𝑐ˆ = arg min 𝑡𝑖 𝑤𝑐𝑖
𝑐
𝑖
i.e., a document is assigned to the class that is the poorest complement match.
References:
• Rennie, J. D., Shih, L., Teevan, J., & Karger, D. R. (2003). Tackling the poor assumptions of naive bayes text
classifiers. In ICML (Vol. 3, pp. 616-623).
Bernoulli Naive Bayes
BernoulliNB implements the naive Bayes training and classification algorithms for data that is distributed ac-
cording to multivariate Bernoulli distributions; i.e., there may be multiple features but each one is assumed to be a
binary-valued (Bernoulli, boolean) variable. Therefore, this class requires samples to be represented as binary-valued
feature vectors; if handed any other kind of data, a BernoulliNB instance may binarize its input (depending on the
binarize parameter).
The decision rule for Bernoulli naive Bayes is based on
𝑃 (𝑥𝑖 | 𝑦) = 𝑃 (𝑖 | 𝑦)𝑥𝑖 + (1 − 𝑃 (𝑖 | 𝑦))(1 − 𝑥𝑖 )
which differs from multinomial NB’s rule in that it explicitly penalizes the non-occurrence of a feature 𝑖 that is an
indicator for class 𝑦, where the multinomial variant would simply ignore a non-occurring feature.
In the case of text classification, word occurrence vectors (rather than word count vectors) may be used to train and
use this classifier. BernoulliNB might perform better on some datasets, especially those with shorter documents.
It is advisable to evaluate both models, if time permits.
References:
• C.D. Manning, P. Raghavan and H. Schütze (2008). Introduction to Information Retrieval. Cambridge Uni-
versity Press, pp. 234-265.
• A. McCallum and K. Nigam (1998). A comparison of event models for Naive Bayes text classification. Proc.
AAAI/ICML-98 Workshop on Learning for Text Categorization, pp. 41-48.
• V. Metsis, I. Androutsopoulos and G. Paliouras (2006). Spam filtering with Naive Bayes – Which Naive
Bayes? 3rd Conf. on Email and Anti-Spam (CEAS).
Out-of-core naive Bayes model fitting
Naive Bayes models can be used to tackle large scale classification problems for which the full training set might not fit
in memory. To handle this case, MultinomialNB, BernoulliNB, and GaussianNB expose a partial_fit
method that can be used incrementally as done with other classifiers as demonstrated in Out-of-core classification of
text documents. All naive Bayes classifiers support sample weighting.
Contrary to the fit method, the first call to partial_fit needs to be passed the list of all the expected class labels.
For an overview of available strategies in scikit-learn, see also the out-of-core learning documentation.

Note: The partial_fit method call of naive Bayes models introduces some computational overhead. It is
recommended to use data chunk sizes that are as large as possible, that is as the available RAM allows.
3.1.10 Decision Trees
Decision Trees (DTs) are a non-parametric supervised learning method used for classification and regression. The
goal is to create a model that predicts the value of a target variable by learning simple decision rules inferred from the
data features.
For instance, in the example below, decision trees learn from data to approximate a sine curve with a set of if-then-else
decision rules. The deeper the tree, the more complex the decision rules and the fitter the model.
Some advantages of decision trees are:

• Simple to understand and to interpret. Trees can be visualised.
• Requires little data preparation. Other techniques often require data normalisation, dummy variables need to be
created and blank values to be removed. Note however that this module does not support missing values.
• The cost of using the tree (i.e., predicting data) is logarithmic in the number of data points used to train the tree.
• Able to handle both numerical and categorical data. Other techniques are usually specialised in analysing
datasets that have only one type of variable. See algorithms for more information.
• Able to handle multi-output problems.
• Uses a white box model. If a given situation is observable in a model, the explanation for the condition is easily
explained by boolean logic. By contrast, in a black box model (e.g., in an artificial neural network), results may
be more difficult to interpret.
• Possible to validate a model using statistical tests. That makes it possible to account for the reliability of the
model.

• Performs well even if its assumptions are somewhat violated by the true model from which the data were
generated.
The disadvantages of decision trees include:
• Decision-tree learners can create over-complex trees that do not generalise the data well. This is called overfit-
ting. Mechanisms such as pruning (not currently supported), setting the minimum number of samples required
at a leaf node or setting the maximum depth of the tree are necessary to avoid this problem.
• Decision trees can be unstable because small variations in the data might result in a completely different tree
being generated. This problem is mitigated by using decision trees within an ensemble.
• The problem of learning an optimal decision tree is known to be NP-complete under several aspects of optimality
and even for simple concepts. Consequently, practical decision-tree learning algorithms are based on heuristic
algorithms such as the greedy algorithm where locally optimal decisions are made at each node. Such algorithms
cannot guarantee to return the globally optimal decision tree. This can be mitigated by training multiple trees in
an ensemble learner, where the features and samples are randomly sampled with replacement.
• There are concepts that are hard to learn because decision trees do not express them easily, such as XOR, parity
or multiplexer problems.
• Decision tree learners create biased trees if some classes dominate. It is therefore recommended to balance the
dataset prior to fitting with the decision tree.
Classification
DecisionTreeClassifier is a class capable of performing multi-class classification on a dataset.

As with other classifiers, DecisionTreeClassifier takes as input two arrays: an array X, sparse or dense,
of size [n_samples, n_features] holding the training samples, and an array Y of integer values, size
[n_samples], holding the class labels for the training samples:
>>> from sklearn import tree

>>> X = [[0, 0], [1, 1]]
>>> Y = [0, 1]
>>> clf = tree.DecisionTreeClassifier()
>>> clf = clf.fit(X, Y)
After being fitted, the model can then be used to predict the class of samples:
>>> clf.predict([[2., 2.]])

array([1])
Alternatively, the probability of each class can be predicted, which is the fraction of training samples of the same class
in a leaf:
>>> clf.predict_proba([[2., 2.]])

array([[0., 1.]])
DecisionTreeClassifier is capable of both binary (where the labels are [-1, 1]) classification and multiclass
(where the labels are [0, . . . , K-1]) classification.
Using the Iris dataset, we can construct a tree as follows:
>>> from sklearn.datasets import load_iris

>>> iris = load_iris()
>>> clf = tree.DecisionTreeClassifier()
>>> clf = clf.fit(iris.data, iris.target)

Once trained, we can export the tree in Graphviz format using the export_graphviz exporter. If you use the
conda package manager, the graphviz binaries and the python package can be installed with
conda install python-graphviz
Alternatively binaries for graphviz can be downloaded from the graphviz project homepage, and the Python wrapper
installed from pypi with pip install graphviz.
Below is an example graphviz export of the above tree trained on the entire iris dataset; the results are saved in an
output file iris.pdf :
>>> import graphviz

>>> dot_data = tree.export_graphviz(clf, out_file=None)
>>> graph = graphviz.Source(dot_data)
>>> graph.render("iris")
The export_graphviz exporter also supports a variety of aesthetic options, including coloring nodes by their class
(or value for regression) and using explicit variable and class names if desired. Jupyter notebooks also render these
plots inline automatically:
>>> dot_data = tree.export_graphviz(clf, out_file=None,

... feature_names=iris.feature_names,
... class_names=iris.target_names,
... filled=True, rounded=True,
... special_characters=True)
>>> graph = graphviz.Source(dot_data)
>>> graph
Examples:
• Plot the decision surface of a decision tree on the iris dataset
Regression
Decision trees can also be applied to regression problems, using the DecisionTreeRegressor class.
As in the classification setting, the fit method will take as argument arrays X and y, only that in this case y is expected
to have floating point values instead of integer values:

>>> X = [[0, 0], [2, 2]]
>>> y = [0.5, 2.5]
>>> clf = tree.DecisionTreeRegressor()
>>> clf = clf.fit(X, y)
array([0.5])
Examples:
• Decision Tree Regression


Multi-output problems
A multi-output problem is a supervised learning problem with several outputs to predict, that is when Y is a 2d array
of size [n_samples, n_outputs].
When there is no correlation between the outputs, a very simple way to solve this kind of problem is to build n
independent models, i.e. one for each output, and then to use those models to independently predict each one of the
n outputs. However, because it is likely that the output values related to the same input are themselves correlated, an
often better way is to build a single model capable of predicting simultaneously all n outputs. First, it requires lower
training time since only a single estimator is built. Second, the generalization accuracy of the resulting estimator may
often be increased.
With regard to decision trees, this strategy can readily be used to support multi-output problems. This requires the
following changes:
• Store n output values in leaves, instead of 1;
• Use splitting criteria that compute the average reduction across all n outputs.
This module offers support for multi-output problems by implementing this strategy in both
DecisionTreeClassifier and DecisionTreeRegressor. If a decision tree is fit on an output
array Y of size [n_samples, n_outputs] then the resulting estimator will:
• Output n_output values upon predict;
• Output a list of n_output arrays of class probabilities upon predict_proba.
The use of multi-output trees for regression is demonstrated in Multi-output Decision Tree Regression. In this example,
the input X is a single real value and the outputs Y are the sine and cosine of X.
The use of multi-output trees for classification is demonstrated in Face completion with a multi-output estimators. In
this example, the inputs X are the pixels of the upper half of faces and the outputs Y are the pixels of the lower half of
those faces.

Examples:
• Multi-output Decision Tree Regression

• Face completion with a multi-output estimators
References:
• M. Dumont et al, Fast multi-class image annotation with random subwindows and multiple output randomized
trees, International Conference on Computer Vision Theory and Applications 2009
Complexity
In general, the run time cost to construct a balanced binary tree is 𝑂(𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠 𝑛𝑓 𝑒𝑎𝑡𝑢𝑟𝑒𝑠 log(𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠 )) and query
time 𝑂(log(𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠 )). Although the tree construction algorithm attempts to generate balanced trees, they will not
always be balanced. Assuming that the subtrees remain approximately balanced, the cost at each node consists of
searching through 𝑂(𝑛𝑓 𝑒𝑎𝑡𝑢𝑟𝑒𝑠 ) to find the feature that offers the largest reduction in entropy. This has a cost of
𝑂(𝑛𝑓 𝑒𝑎𝑡𝑢𝑟𝑒𝑠 𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠 log(𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠 )) at each node, leading to a total cost over the entire trees (by summing the cost at
each node) of 𝑂(𝑛𝑓 𝑒𝑎𝑡𝑢𝑟𝑒𝑠 𝑛2𝑠𝑎𝑚𝑝𝑙𝑒𝑠 log(𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠 )).
Scikit-learn offers a more efficient implementation for the construction of decision trees. A naive implementation
(as above) would recompute the class label histograms (for classification) or the means (for regression) at for each
new split point along a given feature. Presorting the feature over all relevant samples, and retaining a running la-
bel count, will reduce the complexity at each node to 𝑂(𝑛𝑓 𝑒𝑎𝑡𝑢𝑟𝑒𝑠 log(𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠 )), which results in a total cost of
𝑂(𝑛𝑓 𝑒𝑎𝑡𝑢𝑟𝑒𝑠 𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠 log(𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠 )). This is an option for all tree based algorithms. By default it is turned on for
gradient boosting, where in general it makes training faster, but turned off for all other algorithms as it tends to slow
down training when training deep trees.


Tips on practical use
• Decision trees tend to overfit on data with a large number of features. Getting the right ratio of samples to
number of features is important, since a tree with few samples in high dimensional space is very likely to
overfit.
• Consider performing dimensionality reduction (PCA, ICA, or Feature selection) beforehand to give your tree a
better chance of finding features that are discriminative.
• Visualise your tree as you are training by using the export function. Use max_depth=3 as an initial tree
depth to get a feel for how the tree is fitting to your data, and then increase the depth.
• Remember that the number of samples required to populate the tree doubles for each additional level the tree
grows to. Use max_depth to control the size of the tree to prevent overfitting.
• Use min_samples_split or min_samples_leaf to ensure that multiple samples inform every decision
in the tree, by controlling which splits will be considered. A very small number will usually mean the tree will
overfit, whereas a large number will prevent the tree from learning the data. Try min_samples_leaf=5 as an
initial value. If the sample size varies greatly, a float number can be used as percentage in these two parameters.
While min_samples_split can create arbitrarily small leaves, min_samples_leaf guarantees that each
leaf has a minimum size, avoiding low-variance, over-fit leaf nodes in regression problems. For classification
with few classes, min_samples_leaf=1 is often the best choice.
• Balance your dataset before training to prevent the tree from being biased toward the classes that are dominant.
Class balancing can be done by sampling an equal number of samples from each class, or preferably by nor-
malizing the sum of the sample weights (sample_weight) for each class to the same value. Also note that
weight-based pre-pruning criteria, such as min_weight_fraction_leaf, will then be less biased toward
dominant classes than criteria that are not aware of the sample weights, like min_samples_leaf.
• If the samples are weighted, it will be easier to optimize the tree structure using weight-based pre-pruning
criterion such as min_weight_fraction_leaf, which ensure that leaf nodes contain at least a fraction of
the overall sum of the sample weights.
• All decision trees use np.float32 arrays internally. If training data is not in this format, a copy of the dataset
will be made.
• If the input matrix X is very sparse, it is recommended to convert to sparse csc_matrix before calling fit and
sparse csr_matrix before calling predict. Training time can be orders of magnitude faster for a sparse matrix
input compared to a dense matrix when features have zero values in most of the samples.
Tree algorithms: ID3, C4.5, C5.0 and CART
What are all the various decision tree algorithms and how do they differ from each other? Which one is implemented
in scikit-learn?
ID3 (Iterative Dichotomiser 3) was developed in 1986 by Ross Quinlan. The algorithm creates a multiway tree, finding
for each node (i.e. in a greedy manner) the categorical feature that will yield the largest information gain for categorical
targets. Trees are grown to their maximum size and then a pruning step is usually applied to improve the ability of the
tree to generalise to unseen data.
C4.5 is the successor to ID3 and removed the restriction that features must be categorical by dynamically defining
a discrete attribute (based on numerical variables) that partitions the continuous attribute value into a discrete set of
intervals. C4.5 converts the trained trees (i.e. the output of the ID3 algorithm) into sets of if-then rules. These accuracy
of each rule is then evaluated to determine the order in which they should be applied. Pruning is done by removing a
rule’s precondition if the accuracy of the rule improves without it.
C5.0 is Quinlan’s latest version release under a proprietary license. It uses less memory and builds smaller rulesets
than C4.5 while being more accurate.

CART (Classification and Regression Trees) is very similar to C4.5, but it differs in that it supports numerical target
variables (regression) and does not compute rule sets. CART constructs binary trees using the feature and threshold
that yield the largest information gain at each node.
scikit-learn uses an optimised version of the CART algorithm; however, scikit-learn implementation does not support
categorical variables for now.
Given training vectors 𝑥𝑖 ∈ 𝑅𝑛 , i=1,. . . , l and a label vector 𝑦 ∈ 𝑅𝑙 , a decision tree recursively partitions the space
such that the samples with the same labels are grouped together.
Let the data at node 𝑚 be represented by 𝑄. For each candidate split 𝜃 = (𝑗, 𝑡𝑚 ) consisting of a feature 𝑗 and threshold
𝑡𝑚 , partition the data into 𝑄𝑙𝑒𝑓 𝑡 (𝜃) and 𝑄𝑟𝑖𝑔ℎ𝑡 (𝜃) subsets
𝑄𝑙𝑒𝑓 𝑡 (𝜃) = (𝑥, 𝑦)|𝑥𝑗 <= 𝑡𝑚

𝑄𝑟𝑖𝑔ℎ𝑡 (𝜃) = 𝑄 ∖ 𝑄𝑙𝑒𝑓 𝑡 (𝜃)
The impurity at 𝑚 is computed using an impurity function 𝐻(), the choice of which depends on the task being solved
(classification or regression)
𝑛𝑙𝑒𝑓 𝑡 𝑛𝑟𝑖𝑔ℎ𝑡
𝐺(𝑄, 𝜃) = 𝐻(𝑄𝑙𝑒𝑓 𝑡 (𝜃)) + 𝐻(𝑄𝑟𝑖𝑔ℎ𝑡 (𝜃))
𝑁𝑚 𝑁𝑚
Select the parameters that minimises the impurity
𝜃* = argmin𝜃 𝐺(𝑄, 𝜃)
Recurse for subsets 𝑄𝑙𝑒𝑓 𝑡 (𝜃* ) and 𝑄𝑟𝑖𝑔ℎ𝑡 (𝜃* ) until the maximum allowable depth is reached, 𝑁𝑚 < min𝑠𝑎𝑚𝑝𝑙𝑒𝑠 or
𝑁𝑚 = 1.
Classification criteria
If a target is a classification outcome taking on values 0,1,. . . ,K-1, for node 𝑚, representing a region 𝑅𝑚 with 𝑁𝑚
observations, let
∑︁
𝑝𝑚𝑘 = 1/𝑁𝑚 𝐼(𝑦𝑖 = 𝑘)
𝑥𝑖 ∈𝑅𝑚
be the proportion of class k observations in node 𝑚

Common measures of impurity are Gini
∑︁
𝐻(𝑋𝑚 ) = 𝑝𝑚𝑘 (1 − 𝑝𝑚𝑘 )
𝑘
Entropy
∑︁
𝐻(𝑋𝑚 ) = − 𝑝𝑚𝑘 log(𝑝𝑚𝑘 )
𝑘
and Misclassification
𝐻(𝑋𝑚 ) = 1 − max(𝑝𝑚𝑘 )
where 𝑋𝑚 is the training data in node 𝑚

Regression criteria
If the target is a continuous value, then for node 𝑚, representing a region 𝑅𝑚 with 𝑁𝑚 observations, common criteria
to minimise as for determining locations for future splits are Mean Squared Error, which minimizes the L2 error
using mean values at terminal nodes, and Mean Absolute Error, which minimizes the L1 error using median values at
terminal nodes.
Mean Squared Error:
1 ∑︁
𝑦¯𝑚 = 𝑦𝑖
𝑁𝑚
𝑖∈𝑁𝑚
1 ∑︁
𝐻(𝑋𝑚 ) = (𝑦𝑖 − 𝑦¯𝑚 )2
𝑁𝑚
𝑖∈𝑁𝑚
Mean Absolute Error:

1 ∑︁
𝑦¯𝑚 = 𝑦𝑖
𝑁𝑚
𝑖∈𝑁𝑚
1 ∑︁
𝐻(𝑋𝑚 ) = |𝑦𝑖 − 𝑦¯𝑚 |
𝑁𝑚
𝑖∈𝑁𝑚
where 𝑋𝑚 is the training data in node 𝑚
References:
• https://en.wikipedia.org/wiki/Decision_tree_learning
• https://en.wikipedia.org/wiki/Predictive_analytics
• L. Breiman, J. Friedman, R. Olshen, and C. Stone. Classification and Regression Trees. Wadsworth, Belmont,
CA, 1984.
• J.R. Quinlan. C4. 5: programs for machine learning. Morgan Kaufmann, 1993.
• T. Hastie, R. Tibshirani and J. Friedman. Elements of Statistical Learning, Springer, 2009.
3.1.11 Ensemble methods
The goal of ensemble methods is to combine the predictions of several base estimators built with a given learning
algorithm in order to improve generalizability / robustness over a single estimator.
Two families of ensemble methods are usually distinguished:
• In averaging methods, the driving principle is to build several estimators independently and then to average
their predictions. On average, the combined estimator is usually better than any of the single base estimator
because its variance is reduced.
Examples: Bagging methods, Forests of randomized trees, . . .
• By contrast, in boosting methods, base estimators are built sequentially and one tries to reduce the bias of the
combined estimator. The motivation is to combine several weak models to produce a powerful ensemble.
Examples: AdaBoost, Gradient Tree Boosting, . . .

Bagging meta-estimator
In ensemble algorithms, bagging methods form a class of algorithms which build several instances of a black-box
estimator on random subsets of the original training set and then aggregate their individual predictions to form a final
prediction. These methods are used as a way to reduce the variance of a base estimator (e.g., a decision tree), by
introducing randomization into its construction procedure and then making an ensemble out of it. In many cases,
bagging methods constitute a very simple way to improve with respect to a single model, without making it necessary
to adapt the underlying base algorithm. As they provide a way to reduce overfitting, bagging methods work best with
strong and complex models (e.g., fully developed decision trees), in contrast with boosting methods which usually
work best with weak models (e.g., shallow decision trees).
Bagging methods come in many flavours but mostly differ from each other by the way they draw random subsets of
the training set:
• When random subsets of the dataset are drawn as random subsets of the samples, then this algorithm is known
as Pasting [B1999].
• When samples are drawn with replacement, then the method is known as Bagging [B1996].
• When random subsets of the dataset are drawn as random subsets of the features, then the method is known as
Random Subspaces [H1998].
• Finally, when base estimators are built on subsets of both samples and features, then the method is known as
Random Patches [LG2012].
In scikit-learn, bagging methods are offered as a unified BaggingClassifier meta-estimator (resp.
BaggingRegressor), taking as input a user-specified base estimator along with parameters specifying the strategy
to draw random subsets. In particular, max_samples and max_features control the size of the subsets (in terms
of samples and features), while bootstrap and bootstrap_features control whether samples and features
are drawn with or without replacement. When using a subset of the available samples the generalization accuracy can
be estimated with the out-of-bag samples by setting oob_score=True. As an example, the snippet below illustrates
how to instantiate a bagging ensemble of KNeighborsClassifier base estimators, each built on random subsets
of 50% of the samples and 50% of the features.
>>> from sklearn.ensemble import BaggingClassifier

>>> bagging = BaggingClassifier(KNeighborsClassifier(),
... max_samples=0.5, max_features=0.5)
Examples:
• Single estimator versus bagging: bias-variance decomposition
References
Forests of randomized trees
The sklearn.ensemble module includes two averaging algorithms based on randomized decision trees: the Ran-
domForest algorithm and the Extra-Trees method. Both algorithms are perturb-and-combine techniques [B1998]
specifically designed for trees. This means a diverse set of classifiers is created by introducing randomness in the
classifier construction. The prediction of the ensemble is given as the averaged prediction of the individual classifiers.

As other classifiers, forest classifiers have to be fitted with two arrays: a sparse or dense array X of size [n_samples,
n_features] holding the training samples, and an array Y of size [n_samples] holding the target values (class
labels) for the training samples:
>>> from sklearn.ensemble import RandomForestClassifier
>>> X = [[0, 0], [1, 1]]
>>> Y = [0, 1]
>>> clf = RandomForestClassifier(n_estimators=10)
>>> clf = clf.fit(X, Y)
Like decision trees, forests of trees also extend to multi-output problems (if Y is an array of size [n_samples,
n_outputs]).
Random Forests
In random forests (see RandomForestClassifier and RandomForestRegressor classes), each tree in the
ensemble is built from a sample drawn with replacement (i.e., a bootstrap sample) from the training set. In addition,
when splitting a node during the construction of the tree, the split that is chosen is no longer the best split among all
features. Instead, the split that is picked is the best split among a random subset of the features. As a result of this
randomness, the bias of the forest usually slightly increases (with respect to the bias of a single non-random tree) but,
due to averaging, its variance also decreases, usually more than compensating for the increase in bias, hence yielding
an overall better model.
In contrast to the original publication [B2001], the scikit-learn implementation combines classifiers by averaging their
probabilistic prediction, instead of letting each classifier vote for a single class.
Extremely Randomized Trees
In extremely randomized trees (see ExtraTreesClassifier and ExtraTreesRegressor classes), random-

ness goes one step further in the way splits are computed. As in random forests, a random subset of candidate features
is used, but instead of looking for the most discriminative thresholds, thresholds are drawn at random for each candi-
date feature and the best of these randomly-generated thresholds is picked as the splitting rule. This usually allows to
reduce the variance of the model a bit more, at the expense of a slightly greater increase in bias:
>>> from sklearn.model_selection import cross_val_score
>>> from sklearn.datasets import make_blobs
>>> from sklearn.ensemble import ExtraTreesClassifier
>>> from sklearn.tree import DecisionTreeClassifier
>>> X, y = make_blobs(n_samples=10000, n_features=10, centers=100,

... random_state=0)
>>> clf = DecisionTreeClassifier(max_depth=None, min_samples_split=2,

... random_state=0)
>>> scores = cross_val_score(clf, X, y, cv=5)
>>> scores.mean()
0.98...
>>> clf = RandomForestClassifier(n_estimators=10, max_depth=None,

... min_samples_split=2, random_state=0)
>>> scores.mean()
0.999...

>>> clf = ExtraTreesClassifier(n_estimators=10, max_depth=None,

... min_samples_split=2, random_state=0)
>>> scores.mean() > 0.999
True
Parameters
The main parameters to adjust when using these methods is n_estimators and max_features. The former
is the number of trees in the forest. The larger the better, but also the longer it will take to compute. In addition,
note that results will stop getting significantly better beyond a critical number of trees. The latter is the size of
the random subsets of features to consider when splitting a node. The lower the greater the reduction of variance,
but also the greater the increase in bias. Empirical good default values are max_features=n_features for
regression problems, and max_features=sqrt(n_features) for classification tasks (where n_features is
the number of features in the data). Good results are often achieved when setting max_depth=None in combination
with min_samples_split=2 (i.e., when fully developing the trees). Bear in mind though that these values are
usually not optimal, and might result in models that consume a lot of RAM. The best parameter values should always be
cross-validated. In addition, note that in random forests, bootstrap samples are used by default (bootstrap=True)
while the default strategy for extra-trees is to use the whole dataset (bootstrap=False). When using bootstrap
sampling the generalization accuracy can be estimated on the left out or out-of-bag samples. This can be enabled by
setting oob_score=True.
Note: The size of the model with the default parameters is 𝑂(𝑀 * 𝑁 * 𝑙𝑜𝑔(𝑁 )), where 𝑀 is the number of
trees and 𝑁 is the number of samples. In order to reduce the size of the model, you can change these parameters:
min_samples_split, max_leaf_nodes, max_depth and min_samples_leaf.

Parallelization
Finally, this module also features the parallel construction of the trees and the parallel computation of the predictions
through the n_jobs parameter. If n_jobs=k then computations are partitioned into k jobs, and run on k cores of
the machine. If n_jobs=-1 then all cores available on the machine are used. Note that because of inter-process
communication overhead, the speedup might not be linear (i.e., using k jobs will unfortunately not be k times as fast).
Significant speedup can still be achieved though when building a large number of trees, or when building a single tree
requires a fair amount of time (e.g., on large datasets).
Examples:
• Plot the decision surfaces of ensembles of trees on the iris dataset

• Pixel importances with a parallel forest of trees
References
• P. Geurts, D. Ernst., and L. Wehenkel, “Extremely randomized trees”, Machine Learning, 63(1), 3-42, 2006.
Feature importance evaluation
The relative rank (i.e. depth) of a feature used as a decision node in a tree can be used to assess the relative importance
of that feature with respect to the predictability of the target variable. Features used at the top of the tree contribute
to the final prediction decision of a larger fraction of the input samples. The expected fraction of the samples they
contribute to can thus be used as an estimate of the relative importance of the features. In scikit-learn, the fraction of
samples a feature contributes to is combined with the decrease in impurity from splitting them to create a normalized
estimate of the predictive power of that feature.
By averaging the estimates of predictive ability over several randomized trees one can reduce the variance of such
an estimate and use it for feature selection. This is known as the mean decrease in impurity, or MDI. Refer to [L2014]
for more information on MDI and feature importance evaluation with Random Forests.
The following example shows a color-coded representation of the relative importances of each individual pixel for a
face recognition task using a ExtraTreesClassifier model.
In practice those estimates are stored as an attribute named feature_importances_ on the fitted model. This
is an array with shape (n_features,) whose values are positive and sum to 1.0. The higher the value, the more
important is the contribution of the matching feature to the prediction function.
Examples:

• Feature importances with forests of trees
References

Totally Random Trees Embedding
RandomTreesEmbedding implements an unsupervised transformation of the data. Using a forest of completely

random trees, RandomTreesEmbedding encodes the data by the indices of the leaves a data point ends up in. This
index is then encoded in a one-of-K manner, leading to a high dimensional, sparse binary coding. This coding can be
computed very efficiently and can then be used as a basis for other learning tasks. The size and sparsity of the code
can be influenced by choosing the number of trees and the maximum depth per tree. For each tree in the ensemble, the
coding contains one entry of one. The size of the coding is at most n_estimators * 2 ** max_depth, the
maximum number of leaves in the forest.
As neighboring data points are more likely to lie within the same leaf of a tree, the transformation performs an implicit,
non-parametric density estimation.
Examples:
• Hashing feature transformation using Totally Random Trees

• Manifold learning on handwritten digits: Locally Linear Embedding, Isomap. . . compares non-linear dimen-
sionality reduction techniques on handwritten digits.
• Feature transformations with ensembles of trees compares supervised and unsupervised tree based feature
transformations.
See also:
Manifold learning techniques can also be useful to derive non-linear representations of feature space, also these ap-
proaches focus also on dimensionality reduction.

AdaBoost
The module sklearn.ensemble includes the popular boosting algorithm AdaBoost, introduced in 1995 by Freund
and Schapire [FS1995].
The core principle of AdaBoost is to fit a sequence of weak learners (i.e., models that are only slightly better than
random guessing, such as small decision trees) on repeatedly modified versions of the data. The predictions from
all of them are then combined through a weighted majority vote (or sum) to produce the final prediction. The data
modifications at each so-called boosting iteration consist of applying weights 𝑤1 , 𝑤2 , . . . , 𝑤𝑁 to each of the training
samples. Initially, those weights are all set to 𝑤𝑖 = 1/𝑁 , so that the first step simply trains a weak learner on the
original data. For each successive iteration, the sample weights are individually modified and the learning algorithm is
reapplied to the reweighted data. At a given step, those training examples that were incorrectly predicted by the boosted
model induced at the previous step have their weights increased, whereas the weights are decreased for those that were
predicted correctly. As iterations proceed, examples that are difficult to predict receive ever-increasing influence. Each
subsequent weak learner is thereby forced to concentrate on the examples that are missed by the previous ones in the
sequence [HTF].
AdaBoost can be used both for classification and regression problems:

• For multi-class classification, AdaBoostClassifier implements AdaBoost-SAMME and AdaBoost-
SAMME.R [ZZRH2009].
• For regression, AdaBoostRegressor implements AdaBoost.R2 [D1997].
Usage
The following example shows how to fit an AdaBoost classifier with 100 weak learners:
>>> from sklearn.ensemble import AdaBoostClassifier


>>> clf = AdaBoostClassifier(n_estimators=100)
>>> scores = cross_val_score(clf, iris.data, iris.target, cv=5)
>>> scores.mean()
0.9...
The number of weak learners is controlled by the parameter n_estimators. The learning_rate parameter
controls the contribution of the weak learners in the final combination. By default, weak learners are decision stumps.
Different weak learners can be specified through the base_estimator parameter. The main parameters to tune to
obtain good results are n_estimators and the complexity of the base estimators (e.g., its depth max_depth or
minimum required number of samples to consider a split min_samples_split).
Examples:
• Discrete versus Real AdaBoost compares the classification error of a decision stump, decision tree, and a
boosted decision stump using AdaBoost-SAMME and AdaBoost-SAMME.R.
• Multi-class AdaBoosted Decision Trees shows the performance of AdaBoost-SAMME and AdaBoost-
SAMME.R on a multi-class problem.
• Two-class AdaBoost shows the decision boundary and decision function values for a non-linearly separable
two-class problem using AdaBoost-SAMME.
• Decision Tree Regression with AdaBoost demonstrates regression with the AdaBoost.R2 algorithm.
References
Gradient Tree Boosting
Gradient Tree Boosting or Gradient Boosted Regression Trees (GBRT) is a generalization of boosting to arbitrary
differentiable loss functions. GBRT is an accurate and effective off-the-shelf procedure that can be used for both
regression and classification problems. Gradient Tree Boosting models are used in a variety of areas including Web
search ranking and ecology.
The advantages of GBRT are:
• Natural handling of data of mixed type (= heterogeneous features)
• Predictive power
• Robustness to outliers in output space (via robust loss functions)
The disadvantages of GBRT are:
• Scalability, due to the sequential nature of boosting it can hardly be parallelized.
The module sklearn.ensemble provides methods for both classification and regression via gradient boosted
regression trees.
Classification
GradientBoostingClassifier supports both binary and multi-class classification. The following example
shows how to fit a gradient boosting classifier with 100 decision stumps as weak learners:

>>> from sklearn.datasets import make_hastie_10_2

>>> from sklearn.ensemble import GradientBoostingClassifier
>>> X, y = make_hastie_10_2(random_state=0)
>>> X_train, X_test = X[:2000], X[2000:]
>>> y_train, y_test = y[:2000], y[2000:]
>>> clf = GradientBoostingClassifier(n_estimators=100, learning_rate=1.0,

... max_depth=1, random_state=0).fit(X_train, y_train)
>>> clf.score(X_test, y_test)
0.913...
The number of weak learners (i.e. regression trees) is controlled by the parameter n_estimators; The size of each
tree can be controlled either by setting the tree depth via max_depth or by setting the number of leaf nodes via
max_leaf_nodes. The learning_rate is a hyper-parameter in the range (0.0, 1.0] that controls overfitting via
shrinkage .
Note: Classification with more than 2 classes requires the induction of n_classes regression trees at each
iteration, thus, the total number of induced trees equals n_classes * n_estimators. For datasets with
a large number of classes we strongly recommend to use RandomForestClassifier as an alternative to
GradientBoostingClassifier .
Regression
GradientBoostingRegressor supports a number of different loss functions for regression which can be speci-
fied via the argument loss; the default loss function for regression is least squares ('ls').

>>> from sklearn.metrics import mean_squared_error
>>> from sklearn.datasets import make_friedman1
>>> from sklearn.ensemble import GradientBoostingRegressor
>>> X, y = make_friedman1(n_samples=1200, random_state=0, noise=1.0)

>>> X_train, X_test = X[:200], X[200:]
>>> y_train, y_test = y[:200], y[200:]
>>> est = GradientBoostingRegressor(n_estimators=100, learning_rate=0.1,
... max_depth=1, random_state=0, loss='ls').fit(X_train, y_train)
>>> mean_squared_error(y_test, est.predict(X_test))
5.00...
The figure below shows the results of applying GradientBoostingRegressor with least squares loss and 500
base learners to the Boston house price dataset (sklearn.datasets.load_boston). The plot on the left shows
the train and test error at each iteration. The train error at each iteration is stored in the train_score_ attribute
of the gradient boosting model. The test error at each iterations can be obtained via the staged_predict method
which returns a generator that yields the predictions at each stage. Plots like these can be used to determine the optimal
number of trees (i.e. n_estimators) by early stopping. The plot on the right shows the feature importances which
can be obtained via the feature_importances_ property.
Examples:
• Gradient Boosting regression

• Gradient Boosting Out-of-Bag estimates
Fitting additional weak-learners
Both GradientBoostingRegressor and GradientBoostingClassifier support

warm_start=True which allows you to add more estimators to an already fitted model.
>>> _ = est.set_params(n_estimators=200, warm_start=True) # set warm_start and new

˓→nr of trees
>>> _ = est.fit(X_train, y_train) # fit additional 100 trees to est

>>> mean_squared_error(y_test, est.predict(X_test))
3.84...
Controlling the tree size
The size of the regression tree base learners defines the level of variable interactions that can be captured by the
gradient boosting model. In general, a tree of depth h can capture interactions of order h . There are two ways in
which the size of the individual regression trees can be controlled.
If you specify max_depth=h then complete binary trees of depth h will be grown. Such trees will have (at most)
2**h leaf nodes and 2**h - 1 split nodes.
Alternatively, you can control the tree size by specifying the number of leaf nodes via the parameter
max_leaf_nodes. In this case, trees will be grown using best-first search where nodes with the highest improve-
ment in impurity will be expanded first. A tree with max_leaf_nodes=k has k - 1 split nodes and thus can
model interactions of up to order max_leaf_nodes - 1 .
We found that max_leaf_nodes=k gives comparable results to max_depth=k-1 but is significantly faster to
train at the expense of a slightly higher training error. The parameter max_leaf_nodes corresponds to the variable
J in the chapter on gradient boosting in [F2001] and is related to the parameter interaction.depth in R’s gbm
package where max_leaf_nodes == interaction.depth + 1 .

GBRT considers additive models of the following form:
𝑀
∑︁
𝐹 (𝑥) = 𝛾𝑚 ℎ𝑚 (𝑥)
𝑚=1
where ℎ𝑚 (𝑥) are the basis functions which are usually called weak learners in the context of boosting. Gradient Tree
Boosting uses decision trees of fixed size as weak learners. Decision trees have a number of abilities that make them
valuable for boosting, namely the ability to handle data of mixed type and the ability to model complex functions.
Similar to other boosting algorithms, GBRT builds the additive model in a greedy fashion:
𝐹𝑚 (𝑥) = 𝐹𝑚−1 (𝑥) + 𝛾𝑚 ℎ𝑚 (𝑥),
where the newly added tree ℎ𝑚 tries to minimize the loss 𝐿, given the previous ensemble 𝐹𝑚−1 :
𝑛
∑︁
ℎ𝑚 = arg min 𝐿(𝑦𝑖 , 𝐹𝑚−1 (𝑥𝑖 ) + ℎ(𝑥𝑖 )).
ℎ
𝑖=1
The initial model 𝐹0 is problem specific, for least-squares regression one usually chooses the mean of the target values.
Note: The initial model can also be specified via the init argument. The passed object has to implement fit and
predict.
Gradient Boosting attempts to solve this minimization problem numerically via steepest descent: The steepest descent
direction is the negative gradient of the loss function evaluated at the current model 𝐹𝑚−1 which can be calculated for
any differentiable loss function:
𝑛
∑︁
𝐹𝑚 (𝑥) = 𝐹𝑚−1 (𝑥) − 𝛾𝑚 ∇𝐹 𝐿(𝑦𝑖 , 𝐹𝑚−1 (𝑥𝑖 ))
𝑖=1
Where the step length 𝛾𝑚 is chosen using line search:
𝑛
∑︁ 𝜕𝐿(𝑦𝑖 , 𝐹𝑚−1 (𝑥𝑖 ))
𝛾𝑚 = arg min 𝐿(𝑦𝑖 , 𝐹𝑚−1 (𝑥𝑖 ) − 𝛾 )
𝛾
𝑖=1
𝜕𝐹𝑚−1 (𝑥𝑖 )
The algorithms for regression and classification only differ in the concrete loss function used.
Loss Functions
The following loss functions are supported and can be specified using the parameter loss:
• Regression

– Least squares ('ls'): The natural choice for regression due to its superior computational properties. The
initial model is given by the mean of the target values.
– Least absolute deviation ('lad'): A robust loss function for regression. The initial model is given by the
median of the target values.
– Huber ('huber'): Another robust loss function that combines least squares and least absolute deviation;
use alpha to control the sensitivity with regards to outliers (see [F2001] for more details).
– Quantile ('quantile'): A loss function for quantile regression. Use 0 < alpha < 1 to specify the
quantile. This loss function can be used to create prediction intervals (see Prediction Intervals for Gradient
Boosting Regression).
• Classification
– Binomial deviance ('deviance'): The negative binomial log-likelihood loss function for binary classi-
fication (provides probability estimates). The initial model is given by the log odds-ratio.
– Multinomial deviance ('deviance'): The negative multinomial log-likelihood loss function for multi-
class classification with n_classes mutually exclusive classes. It provides probability estimates. The
initial model is given by the prior probability of each class. At each iteration n_classes regression trees
have to be constructed which makes GBRT rather inefficient for data sets with a large number of classes.
– Exponential loss ('exponential'): The same loss function as AdaBoostClassifier. Less robust
to mislabeled examples than 'deviance'; can only be used for binary classification.
Regularization
Shrinkage
[F2001] proposed a simple regularization strategy that scales the contribution of each weak learner by a factor 𝜈:
𝐹𝑚 (𝑥) = 𝐹𝑚−1 (𝑥) + 𝜈𝛾𝑚 ℎ𝑚 (𝑥)
The parameter 𝜈 is also called the learning rate because it scales the step length the gradient descent procedure; it can
be set via the learning_rate parameter.
The parameter learning_rate strongly interacts with the parameter n_estimators, the number of weak learn-
ers to fit. Smaller values of learning_rate require larger numbers of weak learners to maintain a constant training
error. Empirical evidence suggests that small values of learning_rate favor better test error. [HTF2009] recom-
mend to set the learning rate to a small constant (e.g. learning_rate <= 0.1) and choose n_estimators by
early stopping. For a more detailed discussion of the interaction between learning_rate and n_estimators
see [R2007].
Subsampling
[F1999] proposed stochastic gradient boosting, which combines gradient boosting with bootstrap averaging (bagging).
At each iteration the base classifier is trained on a fraction subsample of the available training data. The subsample
is drawn without replacement. A typical value of subsample is 0.5.
The figure below illustrates the effect of shrinkage and subsampling on the goodness-of-fit of the model. We can
clearly see that shrinkage outperforms no-shrinkage. Subsampling with shrinkage can further increase the accuracy of
the model. Subsampling without shrinkage, on the other hand, does poorly.
Another strategy to reduce the variance is by subsampling the features analogous to the random splits in
RandomForestClassifier . The number of subsampled features can be controlled via the max_features
parameter.

Note: Using a small max_features value can significantly decrease the runtime.
Stochastic gradient boosting allows to compute out-of-bag estimates of the test deviance by computing the improve-
ment in deviance on the examples that are not included in the bootstrap sample (i.e. the out-of-bag examples). The
improvements are stored in the attribute oob_improvement_. oob_improvement_[i] holds the improvement
in terms of the loss on the OOB samples if you add the i-th stage to the current predictions. Out-of-bag estimates can
be used for model selection, for example to determine the optimal number of iterations. OOB estimates are usually
very pessimistic thus we recommend to use cross-validation instead and only use OOB if cross-validation is too time
consuming.
Examples:
• Gradient Boosting regularization

• OOB Errors for Random Forests
Interpretation
Individual decision trees can be interpreted easily by simply visualizing the tree structure. Gradient boosting models,
however, comprise hundreds of regression trees thus they cannot be easily interpreted by visual inspection of the
individual trees. Fortunately, a number of techniques have been proposed to summarize and interpret gradient boosting
models.

Feature importance
Often features do not contribute equally to predict the target response; in many situations the majority of the features
are in fact irrelevant. When interpreting a model, the first question usually is: what are those important features and
how do they contributing in predicting the target response?
Individual decision trees intrinsically perform feature selection by selecting appropriate split points. This information
can be used to measure the importance of each feature; the basic idea is: the more often a feature is used in the split
points of a tree the more important that feature is. This notion of importance can be extended to decision tree ensembles
by simply averaging the feature importance of each tree (see Feature importance evaluation for more details).
The feature importance scores of a fit gradient boosting model can be accessed via the feature_importances_
property:

... max_depth=1, random_state=0).fit(X, y)
>>> clf.feature_importances_
array([0.10..., 0.10..., 0.11..., ...
Examples:
Partial dependence
Partial dependence plots (PDP) show the dependence between the target response and a set of ‘target’ features,
marginalizing over the values of all other features (the ‘complement’ features). Intuitively, we can interpret the partial
dependence as the expected target response1 as a function of the ‘target’ features2 .
Due to the limits of human perception the size of the target feature set must be small (usually, one or two) thus the
target features are usually chosen among the most important features.
The Figure below shows four one-way and one two-way partial dependence plots for the California housing dataset:
One-way PDPs tell us about the interaction between the target response and the target feature (e.g. linear, non-linear).
The upper left plot in the above Figure shows the effect of the median income in a district on the median house price;
we can clearly see a linear relationship among them.
PDPs with two target features show the interactions among the two features. For example, the two-variable PDP in
the above Figure shows the dependence of median house price on joint values of house age and avg. occupants per
household. We can clearly see an interaction between the two features: For an avg. occupancy greater than two, the
house price is nearly independent of the house age, whereas for values less than two there is a strong dependence on
age.
The module partial_dependence provides a convenience function plot_partial_dependence to cre-
ate one-way and two-way partial dependence plots. In the below example we show how to create a grid of partial
dependence plots: two one-way PDPs for the features 0 and 1 and a two-way PDP between the two features:
1
For classification with loss='deviance' the target response is logit(p).
2
More precisely its the expectation of the target response after accounting for the initial model; partial dependence plots do not include the
init model.


>>> from sklearn.ensemble.partial_dependence import plot_partial_dependence
... max_depth=1, random_state=0).fit(X, y)
>>> features = [0, 1, (0, 1)]
>>> fig, axs = plot_partial_dependence(clf, X, features)
For multi-class models, you need to set the class label for which the PDPs should be created via the label argument:

>>> mc_clf = GradientBoostingClassifier(n_estimators=10,
... max_depth=1).fit(iris.data, iris.target)
>>> features = [3, 2, (3, 2)]
>>> fig, axs = plot_partial_dependence(mc_clf, X, features, label=0)
If you need the raw values of the partial dependence function rather than the plots you can use the
partial_dependence function:
>>> from sklearn.ensemble.partial_dependence import partial_dependence
>>> pdp, axes = partial_dependence(clf, [0], X=X)

>>> pdp
array([[ 2.46643157, 2.46643157, ...
>>> axes
[array([-1.62497054, -1.59201391, ...
The function requires either the argument grid which specifies the values of the target features on which the partial
dependence function should be evaluated or the argument X which is a convenience mode for automatically creating
grid from the training data. If X is given, the axes value returned by the function gives the axis for each target
feature.

For each value of the ‘target’ features in the grid the partial dependence function need to marginalize the predictions
of a tree over all possible values of the ‘complement’ features. In decision trees this function can be evaluated effi-
ciently without reference to the training data. For each grid point a weighted tree traversal is performed: if a split node
involves a ‘target’ feature, the corresponding left or right branch is followed, otherwise both branches are followed,
each branch is weighted by the fraction of training samples that entered that branch. Finally, the partial dependence
is given by a weighted average of all visited leaves. For tree ensembles the results of each individual tree are again
averaged.
Examples:
• Partial Dependence Plots
References
Voting Classifier
The idea behind the VotingClassifier is to combine conceptually different machine learning classifiers and use
a majority vote or the average predicted probabilities (soft vote) to predict the class labels. Such a classifier can be
useful for a set of equally well performing model in order to balance out their individual weaknesses.
Majority Class Labels (Majority/Hard Voting)
In majority voting, the predicted class label for a particular sample is the class label that represents the majority (mode)
of the class labels predicted by each individual classifier.
E.g., if the prediction for a given sample is
• classifier 1 -> class 1
the VotingClassifier (with voting='hard') would classify the sample as “class 1” based on the majority class label.
In the cases of a tie, the VotingClassifier will select the class based on the ascending sort order. E.g., in the following
scenario
the class label 1 will be assigned to the sample.
Usage
The following example shows how to fit the majority rule classifier:

>>> from sklearn.linear_model import LogisticRegression

>>> from sklearn.ensemble import VotingClassifier

>>> X, y = iris.data[:, 1:3], iris.target
>>> clf1 = LogisticRegression(solver='lbfgs', multi_class='multinomial',

... random_state=1)
>>> clf2 = RandomForestClassifier(n_estimators=50, random_state=1)
>>> clf3 = GaussianNB()
>>> eclf = VotingClassifier(estimators=[('lr', clf1), ('rf', clf2), ('gnb', clf3)],

˓→voting='hard')
>>> for clf, label in zip([clf1, clf2, clf3, eclf], ['Logistic Regression', 'Random
˓→Forest', 'naive Bayes', 'Ensemble']):
... scores = cross_val_score(clf, X, y, cv=5, scoring='accuracy')

... print("Accuracy: %0.2f (+/- %0.2f) [%s]" % (scores.mean(), scores.std(),
˓→label))
Accuracy: 0.95 (+/- 0.04) [Logistic Regression]

Accuracy: 0.94 (+/- 0.04) [Random Forest]
Accuracy: 0.91 (+/- 0.04) [naive Bayes]
Accuracy: 0.95 (+/- 0.04) [Ensemble]
Weighted Average Probabilities (Soft Voting)
In contrast to majority voting (hard voting), soft voting returns the class label as argmax of the sum of predicted
probabilities.
Specific weights can be assigned to each classifier via the weights parameter. When weights are provided, the
predicted class probabilities for each classifier are collected, multiplied by the classifier weight, and averaged. The
final class label is then derived from the class label with the highest average probability.
To illustrate this with a simple example, let’s assume we have 3 classifiers and a 3-class classification problems where
we assign equal weights to all classifiers: w1=1, w2=1, w3=1.
The weighted average probabilities for a sample would then be calculated as follows:
classifier class 1 class 2 class 3

classifier 1 w1 * 0.2 w1 * 0.5 w1 * 0.3
classifier 2 w2 * 0.6 w2 * 0.3 w2 * 0.1
classifier 3 w3 * 0.3 w3 * 0.4 w3 * 0.3
weighted average 0.37 0.4 0.23
Here, the predicted class label is 2, since it has the highest average probability.
The following example illustrates how the decision regions may change when a soft VotingClassifier is used based on
an linear Support Vector Machine, a Decision Tree, and a K-nearest neighbor classifier:
>>> from itertools import product
>>> from sklearn.ensemble import VotingClassifier
>>> # Loading some example data


>>> X = iris.data[:, [0, 2]]
>>> y = iris.target
>>> # Training classifiers

>>> clf1 = DecisionTreeClassifier(max_depth=4)
>>> clf2 = KNeighborsClassifier(n_neighbors=7)
>>> clf3 = SVC(gamma='scale', kernel='rbf', probability=True)
>>> eclf = VotingClassifier(estimators=[('dt', clf1), ('knn', clf2), ('svc', clf3)],
... voting='soft', weights=[2, 1, 2])
>>> clf1 = clf1.fit(X, y)

>>> eclf = eclf.fit(X, y)
Using the VotingClassifier with GridSearch
The VotingClassifier can also be used together with GridSearch in order to tune the hyperparameters of the individual
estimators:


... random_state=1)
>>> clf2 = RandomForestClassifier(random_state=1)
˓→voting='soft')
>>> params = {'lr__C': [1.0, 100.0], 'rf__n_estimators': [20, 200]}
>>> grid = GridSearchCV(estimator=eclf, param_grid=params, cv=5)

>>> grid = grid.fit(iris.data, iris.target)
Usage
In order to predict the class labels based on the predicted class-probabilities (scikit-learn estimators in the VotingClas-
sifier must support predict_proba method):

˓→voting='soft')
Optionally, weights can be provided for the individual classifiers:

... voting='soft', weights=[2, 5, 1])
3.1.12 Multiclass and multilabel algorithms
Warning: All classifiers in scikit-learn do multiclass classification out-of-the-box. You don’t need to use the
sklearn.multiclass module unless you want to experiment with different multiclass strategies.
The sklearn.multiclass module implements meta-estimators to solve multiclass and multilabel clas-
sification problems by decomposing such problems into binary classification problems. Multitarget regression is also
supported.
• Multiclass classification means a classification task with more than two classes; e.g., classify a set of images of
fruits which may be oranges, apples, or pears. Multiclass classification makes the assumption that each sample
is assigned to one and only one label: a fruit can be either an apple or a pear but not both at the same time.
• Multilabel classification assigns to each sample a set of target labels. This can be thought as predicting proper-
ties of a data-point that are not mutually exclusive, such as topics that are relevant for a document. A text might
be about any of religion, politics, finance or education at the same time or none of these.
• Multioutput regression assigns each sample a set of target values. This can be thought of as predicting several
properties for each data-point, such as wind direction and magnitude at a certain location.
• Multioutput-multiclass classification and multi-task classification means that a single estimator has to handle
several joint classification tasks. This is both a generalization of the multi-label classification task, which only
considers binary classification, as well as a generalization of the multi-class classification task. The output
format is a 2d numpy array or sparse matrix.
The set of labels can be different for each output variable. For instance, a sample could be assigned “pear” for
an output variable that takes possible values in a finite set of species such as “pear”, “apple”; and “blue” or

“green” for a second output variable that takes possible values in a finite set of colors such as “green”, “red”,
“blue”, “yellow”. . .
This means that any classifiers handling multi-output multiclass or multi-task classification tasks, support the
multi-label classification task as a special case. Multi-task classification is similar to the multi-output classifica-
tion task with different model formulations. For more information, see the relevant estimator documentation.
All scikit-learn classifiers are capable of multiclass classification, but the meta-estimators offered by sklearn.
multiclass permit changing the way they handle more than two classes because this may have an effect on classifier
performance (either in terms of generalization error or required computational resources).
Below is a summary of the classifiers supported by scikit-learn grouped by strategy; you don’t need the meta-estimators
in this class if you’re using one of these, unless you want custom multiclass behavior:
• Inherently multiclass:
– sklearn.naive_bayes.BernoulliNB
– sklearn.tree.DecisionTreeClassifier
– sklearn.tree.ExtraTreeClassifier
– sklearn.ensemble.ExtraTreesClassifier
– sklearn.naive_bayes.GaussianNB
– sklearn.neighbors.KNeighborsClassifier
– sklearn.semi_supervised.LabelPropagation
– sklearn.semi_supervised.LabelSpreading
– sklearn.discriminant_analysis.LinearDiscriminantAnalysis
– sklearn.svm.LinearSVC (setting multi_class=”crammer_singer”)
– sklearn.linear_model.LogisticRegression (setting multi_class=”multinomial”)
– sklearn.linear_model.LogisticRegressionCV (setting multi_class=”multinomial”)
– sklearn.neural_network.MLPClassifier
– sklearn.neighbors.NearestCentroid
– sklearn.discriminant_analysis.QuadraticDiscriminantAnalysis
– sklearn.neighbors.RadiusNeighborsClassifier
– sklearn.ensemble.RandomForestClassifier
– sklearn.linear_model.RidgeClassifier
– sklearn.linear_model.RidgeClassifierCV
• Multiclass as One-Vs-One:
– sklearn.svm.NuSVC
– sklearn.svm.SVC.
– sklearn.gaussian_process.GaussianProcessClassifier (setting multi_class =
“one_vs_one”)
• Multiclass as One-Vs-All:
– sklearn.ensemble.GradientBoostingClassifier
– sklearn.gaussian_process.GaussianProcessClassifier (setting multi_class =
“one_vs_rest”)

– sklearn.svm.LinearSVC (setting multi_class=”ovr”)

– sklearn.linear_model.LogisticRegression (setting multi_class=”ovr”)
– sklearn.linear_model.LogisticRegressionCV (setting multi_class=”ovr”)
– sklearn.linear_model.SGDClassifier
– sklearn.linear_model.Perceptron
– sklearn.linear_model.PassiveAggressiveClassifier
• Support multilabel:
– sklearn.linear_model.RidgeClassifierCV
• Support multiclass-multioutput:
Warning: At present, no metric in sklearn.metrics supports the multioutput-multiclass classification task.
Multilabel classification format
In multilabel learning, the joint set of binary classification tasks is expressed with label binary indicator array: each
sample is one row of a 2d array of shape (n_samples, n_classes) with binary values: the one, i.e. the non zero elements,
corresponds to the subset of labels. An array such as np.array([[1, 0, 0], [0, 1, 1], [0, 0, 0]])
represents label 0 in the first sample, labels 1 and 2 in the second sample, and no labels in the third sample.
Producing multilabel data as a list of sets of labels may be more intuitive. The MultiLabelBinarizer transformer
can be used to convert between a collection of collections of labels and the indicator format.

>>> y = [[2, 3, 4], [2], [0, 1, 3], [0, 1, 2, 3, 4], [0, 1, 2]]
>>> MultiLabelBinarizer().fit_transform(y)
array([[0, 0, 1, 1, 1],
[0, 0, 1, 0, 0],
[1, 1, 0, 1, 0],

[1, 1, 1, 1, 1],
[1, 1, 1, 0, 0]])
One-Vs-The-Rest
This strategy, also known as one-vs-all, is implemented in OneVsRestClassifier. The strategy consists in
fitting one classifier per class. For each classifier, the class is fitted against all the other classes. In addition to its
computational efficiency (only n_classes classifiers are needed), one advantage of this approach is its interpretability.
Since each class is represented by one and only one classifier, it is possible to gain knowledge about the class by
inspecting its corresponding classifier. This is the most commonly used strategy and is a fair default choice.
Multiclass learning
Below is an example of multiclass learning using OvR:

>>> from sklearn.multiclass import OneVsRestClassifier
>>> from sklearn.svm import LinearSVC
>>> OneVsRestClassifier(LinearSVC(random_state=0)).fit(X, y).predict(X)
array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 2, 2, 2, 1, 2, 2, 2, 2,
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2])
Multilabel learning
OneVsRestClassifier also supports multilabel classification. To use this feature, feed the classifier an indicator
matrix, in which cell [i, j] indicates the presence of label j in sample i.
Examples:
• Multilabel classification
One-Vs-One
OneVsOneClassifier constructs one classifier per pair of classes. At prediction time, the class which received
the most votes is selected. In the event of a tie (among two classes with an equal number of votes), it selects the class
with the highest aggregate classification confidence by summing over the pair-wise classification confidence levels
computed by the underlying binary classifiers.
Since it requires to fit n_classes * (n_classes - 1) / 2 classifiers, this method is usually slower than
one-vs-the-rest, due to its O(n_classes^2) complexity. However, this method may be advantageous for algorithms
such as kernel algorithms which don’t scale well with n_samples. This is because each individual learning problem
only involves a small subset of the data whereas, with one-vs-the-rest, the complete dataset is used n_classes times.

Multiclass learning
Below is an example of multiclass learning using OvO:

>>> from sklearn.multiclass import OneVsOneClassifier
>>> OneVsOneClassifier(LinearSVC(random_state=0)).fit(X, y).predict(X)
array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 2, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2])
References:
• “Pattern Recognition and Machine Learning. Springer”, Christopher M. Bishop, page 183, (First Edition)

Error-Correcting Output-Codes
Output-code based strategies are fairly different from one-vs-the-rest and one-vs-one. With these strategies, each class
is represented in a Euclidean space, where each dimension can only be 0 or 1. Another way to put it is that each class
is represented by a binary code (an array of 0 and 1). The matrix which keeps track of the location/code of each class
is called the code book. The code size is the dimensionality of the aforementioned space. Intuitively, each class should
be represented by a code as unique as possible and a good code book should be designed to optimize classification
accuracy. In this implementation, we simply use a randomly-generated code book as advocated in3 although more
elaborate methods may be added in the future.
At fitting time, one binary classifier per bit in the code book is fitted. At prediction time, the classifiers are used to
project new points in the class space and the class closest to the points is chosen.
In OutputCodeClassifier, the code_size attribute allows the user to control the number of classifiers which
will be used. It is a percentage of the total number of classes.
A number between 0 and 1 will require fewer classifiers than one-vs-the-rest. In theory, log2(n_classes) /
n_classes is sufficient to represent each class unambiguously. However, in practice, it may not lead to good
accuracy since log2(n_classes) is much smaller than n_classes.
A number greater than 1 will require more classifiers than one-vs-the-rest. In this case, some classifiers will in theory
correct for the mistakes made by other classifiers, hence the name “error-correcting”. In practice, however, this may
not happen as classifier mistakes will typically be correlated. The error-correcting output codes have a similar effect
to bagging.
Multiclass learning
Below is an example of multiclass learning using Output-Codes:

>>> from sklearn.multiclass import OutputCodeClassifier
>>> clf = OutputCodeClassifier(LinearSVC(random_state=0),
... code_size=2, random_state=0)
>>> clf.fit(X, y).predict(X)
array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1,
1, 2, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 2, 2, 2, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
2, 2, 2, 2, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 2, 2, 2, 1, 1, 2, 2, 2,
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2])
References:
• “Solving multiclass learning problems via error-correcting output codes”, Dietterich T., Bakiri G., Journal of
Artificial Intelligence Research 2, 1995.
• “The Elements of Statistical Learning”, Hastie T., Tibshirani R., Friedman J., page 606 (second-edition) 2008.
3 “The error coding method and PICTs”, James G., Hastie T., Journal of Computational and Graphical statistics 7, 1998.

Multioutput regression
Multioutput regression support can be added to any regressor with MultiOutputRegressor. This strategy con-
sists of fitting one regressor per target. Since each target is represented by exactly one regressor it is possible to
gain knowledge about the target by inspecting its corresponding regressor. As MultiOutputRegressor fits one
regressor per target it can not take advantage of correlations between targets.
Below is an example of multioutput regression:
>>> from sklearn.datasets import make_regression

>>> from sklearn.multioutput import MultiOutputRegressor
>>> X, y = make_regression(n_samples=10, n_targets=3, random_state=1)
>>> MultiOutputRegressor(GradientBoostingRegressor(random_state=0)).fit(X, y).
˓→predict(X)
array([[-154.75474165, -147.03498585, -50.03812219],

[ 7.12165031, 5.12914884, -81.46081961],
[-187.8948621 , -100.44373091, 13.88978285],
[-141.62745778, 95.02891072, -191.48204257],
[ 97.03260883, 165.34867495, 139.52003279],
[ 123.92529176, 21.25719016, -7.84253 ],
[-122.25193977, -85.16443186, -107.12274212],
[ -30.170388 , -94.80956739, 12.16979946],
[ 140.72667194, 176.50941682, -17.50447799],
[ 149.37967282, -81.15699552, -5.72850319]])
Multioutput classification
Multioutput classification support can be added to any classifier with MultiOutputClassifier. This strategy
consists of fitting one classifier per target. This allows multiple target variable classifications. The purpose of this class
is to extend estimators to be able to estimate a series of target functions (f1,f2,f3. . . ,fn) that are trained on a single X
predictor matrix to predict a series of responses (y1,y2,y3. . . ,yn).
Below is an example of multioutput classification:
>>> from sklearn.datasets import make_classification

>>> from sklearn.multioutput import MultiOutputClassifier
>>> from sklearn.utils import shuffle
>>> X, y1 = make_classification(n_samples=10, n_features=100, n_informative=30, n_
˓→classes=3, random_state=1)
>>> y2 = shuffle(y1, random_state=1)

>>> y3 = shuffle(y1, random_state=2)
>>> Y = np.vstack((y1, y2, y3)).T
>>> n_samples, n_features = X.shape # 10,100
>>> n_outputs = Y.shape[1] # 3
>>> n_classes = 3
>>> forest = RandomForestClassifier(n_estimators=100, random_state=1)
>>> multi_target_forest = MultiOutputClassifier(forest, n_jobs=-1)
>>> multi_target_forest.fit(X, Y).predict(X)
array([[2, 2, 0],
[1, 2, 1],
[2, 1, 0],
[0, 0, 2],
[0, 2, 1],
[0, 0, 2],

[1, 1, 0],
[1, 1, 1],
[0, 0, 2],
[2, 0, 0]])
Classifier Chain
Classifier chains (see ClassifierChain) are a way of combining a number of binary classifiers into a single
multi-label model that is capable of exploiting correlations among targets.
For a multi-label classification problem with N classes, N binary classifiers are assigned an integer between 0 and N-1.
These integers define the order of models in the chain. Each classifier is then fit on the available training data plus the
true labels of the classes whose models were assigned a lower number.
When predicting, the true labels will not be available. Instead the predictions of each model are passed on to the
subsequent models in the chain to be used as features.
Clearly the order of the chain is important. The first model in the chain has no information about the other labels while
the last model in the chain has features indicating the presence of all of the other labels. In general one does not know
the optimal ordering of the models in the chain so typically many randomly ordered chains are fit and their predictions
are averaged together.
References:
Jesse Read, Bernhard Pfahringer, Geoff Holmes, Eibe Frank, “Classifier Chains for Multi-label Classifica-
tion”, 2009.
Regressor Chain
Regressor chains (see RegressorChain) is analogous to ClassifierChain as a way of combining a number of re-
gressions into a single multi-target model that is capable of exploiting correlations among targets.
3.1.13 Feature selection
The classes in the sklearn.feature_selection module can be used for feature selection/dimensionality re-
duction on sample sets, either to improve estimators’ accuracy scores or to boost their performance on very high-
dimensional datasets.
Removing features with low variance
VarianceThreshold is a simple baseline approach to feature selection. It removes all features whose variance
doesn’t meet some threshold. By default, it removes all zero-variance features, i.e. features that have the same value
in all samples.
As an example, suppose that we have a dataset with boolean features, and we want to remove all features that are
either one or zero (on or off) in more than 80% of the samples. Boolean features are Bernoulli random variables, and
the variance of such variables is given by
Var[𝑋] = 𝑝(1 − 𝑝)
so we can select using the threshold .8 * (1 - .8):

>>> from sklearn.feature_selection import VarianceThreshold

>>> X = [[0, 0, 1], [0, 1, 0], [1, 0, 0], [0, 1, 1], [0, 1, 0], [0, 1, 1]]
>>> sel = VarianceThreshold(threshold=(.8 * (1 - .8)))
>>> sel.fit_transform(X)
array([[0, 1],
[1, 0],
[0, 0],
[1, 1],
[1, 0],
[1, 1]])
As expected, VarianceThreshold has removed the first column, which has a probability 𝑝 = 5/6 > .8 of
containing a zero.
Univariate feature selection
Univariate feature selection works by selecting the best features based on univariate statistical tests. It can be seen
as a preprocessing step to an estimator. Scikit-learn exposes feature selection routines as objects that implement the
transform method:
• SelectKBest removes all but the 𝑘 highest scoring features
• SelectPercentile removes all but a user-specified highest scoring percentage of features
• using common univariate statistical tests for each feature: false positive rate SelectFpr, false discovery rate
SelectFdr, or family wise error SelectFwe.
• GenericUnivariateSelect allows to perform univariate feature selection with a configurable strategy.
This allows to select the best univariate selection strategy with hyper-parameter search estimator.
For instance, we can perform a 𝜒2 test to the samples to retrieve only the two best features as follows:

>>> from sklearn.feature_selection import SelectKBest
>>> from sklearn.feature_selection import chi2
>>> X.shape
(150, 4)
>>> X_new = SelectKBest(chi2, k=2).fit_transform(X, y)
>>> X_new.shape
(150, 2)
These objects take as input a scoring function that returns univariate scores and p-values (or only scores for
SelectKBest and SelectPercentile):
• For regression: f_regression, mutual_info_regression
• For classification: chi2, f_classif, mutual_info_classif
The methods based on F-test estimate the degree of linear dependency between two random variables. On the other
hand, mutual information methods can capture any kind of statistical dependency, but being nonparametric, they
require more samples for accurate estimation.
Feature selection with sparse data

If you use sparse data (i.e. data represented as sparse matrices), chi2, mutual_info_regression,
mutual_info_classif will deal with the data without making it dense.
Warning: Beware not to use a regression scoring function with a classification problem, you will get useless
results.
Examples:
• Univariate Feature Selection

• Comparison of F-test and mutual information
Recursive feature elimination
Given an external estimator that assigns weights to features (e.g., the coefficients of a linear model), recursive feature
elimination (RFE) is to select features by recursively considering smaller and smaller sets of features. First, the
estimator is trained on the initial set of features and the importance of each feature is obtained either through a coef_
attribute or through a feature_importances_ attribute. Then, the least important features are pruned from
current set of features.That procedure is recursively repeated on the pruned set until the desired number of features to
select is eventually reached.
RFECV performs RFE in a cross-validation loop to find the optimal number of features.
Examples:
• Recursive feature elimination: A recursive feature elimination example showing the relevance of pixels in a
digit classification task.
• Recursive feature elimination with cross-validation: A recursive feature elimination example with automatic
tuning of the number of features selected with cross-validation.
Feature selection using SelectFromModel
SelectFromModel is a meta-transformer that can be used along with any estimator that has a coef_ or
feature_importances_ attribute after fitting. The features are considered unimportant and removed, if the
corresponding coef_ or feature_importances_ values are below the provided threshold parameter. Apart
from specifying the threshold numerically, there are built-in heuristics for finding a threshold using a string argument.
Available heuristics are “mean”, “median” and float multiples of these like “0.1*mean”.
For examples on how it is to be used refer to the sections below.
Examples
• Feature selection using SelectFromModel and LassoCV: Selecting the two most important features from the
Boston dataset without knowing the threshold beforehand.

L1-based feature selection
Linear models penalized with the L1 norm have sparse solutions: many of their estimated coefficients are zero.
When the goal is to reduce the dimensionality of the data to use with another classifier, they can be used along
with feature_selection.SelectFromModel to select the non-zero coefficients. In particular, sparse
estimators useful for this purpose are the linear_model.Lasso for regression, and of linear_model.
LogisticRegression and svm.LinearSVC for classification:

>>> from sklearn.feature_selection import SelectFromModel
>>> X.shape
(150, 4)
>>> lsvc = LinearSVC(C=0.01, penalty="l1", dual=False).fit(X, y)
>>> model = SelectFromModel(lsvc, prefit=True)
>>> X_new = model.transform(X)
>>> X_new.shape
(150, 3)
With SVMs and logistic-regression, the parameter C controls the sparsity: the smaller C the fewer features selected.
With Lasso, the higher the alpha parameter, the fewer features selected.
Examples:
• Classification of text documents using sparse features: Comparison of different algorithms for document
classification including L1-based feature selection.
L1-recovery and compressive sensing
For a good choice of alpha, the Lasso can fully recover the exact set of non-zero variables using only few obser-
vations, provided certain specific conditions are met. In particular, the number of samples should be “sufficiently
large”, or L1 models will perform at random, where “sufficiently large” depends on the number of non-zero co-
efficients, the logarithm of the number of features, the amount of noise, the smallest absolute value of non-zero
coefficients, and the structure of the design matrix X. In addition, the design matrix must display certain specific
properties, such as not being too correlated.
There is no general rule to select an alpha parameter for recovery of non-zero coefficients. It can by set by cross-
validation (LassoCV or LassoLarsCV), though this may lead to under-penalized models: including a small
number of non-relevant variables is not detrimental to prediction score. BIC (LassoLarsIC) tends, on the oppo-
site, to set high values of alpha.
Reference Richard G. Baraniuk “Compressive Sensing”, IEEE Signal Processing Magazine [120] July 2007 http:
//users.isr.ist.utl.pt/~aguiar/CS_notes.pdf
Tree-based feature selection
Tree-based estimators (see the sklearn.tree module and forest of trees in the sklearn.ensemble module)
can be used to compute feature importances, which in turn can be used to discard irrelevant features (when coupled
with the sklearn.feature_selection.SelectFromModel meta-transformer):

>>> from sklearn.ensemble import ExtraTreesClassifier

>>> from sklearn.feature_selection import SelectFromModel
>>> X.shape
(150, 4)
>>> clf = ExtraTreesClassifier(n_estimators=50)
>>> clf.feature_importances_
array([ 0.04..., 0.05..., 0.4..., 0.4...])
>>> model = SelectFromModel(clf, prefit=True)
>>> X_new = model.transform(X)
>>> X_new.shape
(150, 2)
Examples:
• Feature importances with forests of trees: example on synthetic data showing the recovery of the actually
meaningful features.
• Pixel importances with a parallel forest of trees: example on face recognition data.
Feature selection as part of a pipeline
Feature selection is usually used as a pre-processing step before doing the actual learning. The recommended way to
do this in scikit-learn is to use a sklearn.pipeline.Pipeline:
clf = Pipeline([
('feature_selection', SelectFromModel(LinearSVC(penalty="l1"))),
('classification', RandomForestClassifier())
])
clf.fit(X, y)
In this snippet we make use of a sklearn.svm.LinearSVC coupled with sklearn.feature_selection.

SelectFromModel to evaluate feature importances and select the most relevant features. Then, a sklearn.
ensemble.RandomForestClassifier is trained on the transformed output, i.e. using only relevant features.
You can perform similar operations with the other feature selection methods and also classifiers that provide a way to
evaluate feature importances of course. See the sklearn.pipeline.Pipeline examples for more details.
3.1.14 Semi-Supervised
Semi-supervised learning is a situation in which in your training data some of the samples are not labeled. The semi-
supervised estimators in sklearn.semi_supervised are able to make use of this additional unlabeled data to
better capture the shape of the underlying data distribution and generalize better to new samples. These algorithms
can perform well when we have a very small amount of labeled points and a large amount of unlabeled points.
Unlabeled entries in y
It is important to assign an identifier to unlabeled points along with the labeled data when training the model with
the fit method. The identifier that this implementation uses is the integer value −1.

Label Propagation
Label propagation denotes a few variations of semi-supervised graph inference algorithms.

A few features available in this model:
• Can be used for classification and regression tasks
• Kernel methods to project data into alternate dimensional spaces
scikit-learn provides two label propagation models: LabelPropagation and LabelSpreading. Both work by
constructing a similarity graph over all items in the input dataset.
Fig. 3.1: An illustration of label-propagation: the structure of unlabeled observations is consistent with the class
structure, and thus the class label can be propagated to the unlabeled observations of the training set.
LabelPropagation and LabelSpreading differ in modifications to the similarity matrix that graph and the
clamping effect on the label distributions. Clamping allows the algorithm to change the weight of the true ground
labeled data to some degree. The LabelPropagation algorithm performs hard clamping of input labels, which
means 𝛼 = 0. This clamping factor can be relaxed, to say 𝛼 = 0.2, which means that we will always retain 80 percent
of our original label distribution, but the algorithm gets to change its confidence of the distribution within 20 percent.
LabelPropagation uses the raw similarity matrix constructed from the data with no modifications. In contrast,
LabelSpreading minimizes a loss function that has regularization properties, as such it is often more robust to
noise. The algorithm iterates on a modified version of the original graph and normalizes the edge weights by computing
the normalized graph Laplacian matrix. This procedure is also used in Spectral clustering.
Label propagation models have two built-in kernel methods. Choice of kernel effects both scalability and performance
of the algorithms. The following are available:
• rbf (exp(−𝛾|𝑥 − 𝑦|2 ), 𝛾 > 0). 𝛾 is specified by keyword gamma.
• knn (1[𝑥′ ∈ 𝑘𝑁 𝑁 (𝑥)]). 𝑘 is specified by keyword n_neighbors.
The RBF kernel will produce a fully connected graph which is represented in memory by a dense matrix. This matrix
may be very large and combined with the cost of performing a full matrix multiplication calculation for each iteration
of the algorithm can lead to prohibitively long running times. On the other hand, the KNN kernel will produce a much
more memory-friendly sparse matrix which can drastically reduce running times.
Examples

• Decision boundary of label propagation versus SVM on the Iris dataset

• Label Propagation learning a complex structure
• Label Propagation digits: Demonstrating performance
• Label Propagation digits active learning
References
[1] Yoshua Bengio, Olivier Delalleau, Nicolas Le Roux. In Semi-Supervised Learning (2006), pp. 193-216
[2] Olivier Delalleau, Yoshua Bengio, Nicolas Le Roux. Efficient Non-Parametric Function Induction in Semi-
Supervised Learning. AISTAT 2005 https://research.microsoft.com/en-us/people/nicolasl/efficient_ssl.pdf
3.1.15 Isotonic regression
The class IsotonicRegression fits a non-decreasing function to data. It solves the following problem:
minimize 𝑖 𝑤𝑖 (𝑦𝑖 − 𝑦ˆ𝑖 )2
∑︀
subject to 𝑦ˆ𝑚𝑖𝑛 = 𝑦ˆ1 ≤ 𝑦ˆ2 ... ≤ 𝑦ˆ𝑛 = 𝑦ˆ𝑚𝑎𝑥

where each 𝑤𝑖 is strictly positive and each 𝑦𝑖 is an arbitrary real number. It yields the vector which is composed of
non-decreasing elements the closest in terms of mean squared error. In practice this list of elements forms a function
that is piecewise linear.
3.1.16 Probability calibration
When performing classification you often want not only to predict the class label, but also obtain a probability of the
respective label. This probability gives you some kind of confidence on the prediction. Some models can give you
poor estimates of the class probabilities and some even do not support probability prediction. The calibration module
allows you to better calibrate the probabilities of a given model, or to add support for probability prediction.
Well calibrated classifiers are probabilistic classifiers for which the output of the predict_proba method can be directly
interpreted as a confidence level. For instance, a well calibrated (binary) classifier should classify the samples such
that among the samples to which it gave a predict_proba value close to 0.8, approximately 80% actually belong to the
positive class. The following plot compares how well the probabilistic predictions of different classifiers are calibrated:
LogisticRegression returns well calibrated predictions by default as it directly optimizes log-loss. In contrast,
the other methods return biased probabilities; with different biases per method:
• GaussianNB tends to push probabilities to 0 or 1 (note the counts in the histograms). This is mainly because
it makes the assumption that features are conditionally independent given the class, which is not the case in this
dataset which contains 2 redundant features.
• RandomForestClassifier shows the opposite behavior: the histograms show peaks at approximately
0.2 and 0.9 probability, while probabilities close to 0 or 1 are very rare. An explanation for this is given by
Niculescu-Mizil and Caruana4 : “Methods such as bagging and random forests that average predictions from a
base set of models can have difficulty making predictions near 0 and 1 because variance in the underlying base
models will bias predictions that should be near zero or one away from these values. Because predictions are
restricted to the interval [0,1], errors caused by variance tend to be one-sided near zero and one. For example,
if a model should predict p = 0 for a case, the only way bagging can achieve this is if all bagged trees predict
4 Predicting Good Probabilities with Supervised Learning, A. Niculescu-Mizil & R. Caruana, ICML 2005



zero. If we add noise to the trees that bagging is averaging over, this noise will cause some trees to predict
values larger than 0 for this case, thus moving the average prediction of the bagged ensemble away from 0. We
observe this effect most strongly with random forests because the base-level trees trained with random forests
have relatively high variance due to feature subsetting.” As a result, the calibration curve also referred to as the
reliability diagram (Wilks 19955 ) shows a characteristic sigmoid shape, indicating that the classifier could trust
its “intuition” more and return probabilities closer to 0 or 1 typically.
• Linear Support Vector Classification (LinearSVC) shows an even more sigmoid curve as the RandomForest-
Classifier, which is typical for maximum-margin methods (compare Niculescu-Mizil and Caruana4 ), which
focus on hard samples that are close to the decision boundary (the support vectors).
Two approaches for performing calibration of probabilistic predictions are provided: a parametric approach based on
Platt’s sigmoid model and a non-parametric approach based on isotonic regression (sklearn.isotonic). Proba-
bility calibration should be done on new data not used for model fitting. The class CalibratedClassifierCV
uses a cross-validation generator and estimates for each split the model parameter on the train samples and the cali-
bration of the test samples. The probabilities predicted for the folds are then averaged. Already fitted classifiers can
be calibrated by CalibratedClassifierCV via the parameter cv=”prefit”. In this case, the user has to take care
manually that data for model fitting and calibration are disjoint.
The following images demonstrate the benefit of probability calibration. The first image present a dataset with 2
classes and 3 blobs of data. The blob in the middle contains random samples of each class. The probability for the
samples in this blob should be 0.5.
The following image shows on the data above the estimated probability using a Gaussian naive Bayes classifier without
5 On the combination of forecast probabilities for consecutive precipitation periods. Wea. Forecasting, 5, 640–650., Wilks, D. S., 1990a

calibration, with a sigmoid calibration and with a non-parametric isotonic calibration. One can observe that the non-
parametric model provides the most accurate probability estimates for samples in the middle, i.e., 0.5.
The following experiment is performed on an artificial dataset for binary classification with 100,000 samples (1,000
of them are used for model fitting) with 20 features. Of the 20 features, only 2 are informative and 10 are redundant.
The figure shows the estimated probabilities obtained with logistic regression, a linear support-vector classifier (SVC),
and linear SVC with both isotonic calibration and sigmoid calibration. The Brier score is a metric which is a combi-
nation of calibration loss and refinement loss, brier_score_loss, reported in the legend (the smaller the better).
Calibration loss is defined as the mean squared deviation from empirical probabilities derived from the slope of ROC
segments. Refinement loss can be defined as the expected optimal loss as measured by the area under the optimal cost
curve.
One can observe here that logistic regression is well calibrated as its curve is nearly diagonal. Linear SVC’s calibration
curve or reliability diagram has a sigmoid curve, which is typical for an under-confident classifier. In the case of
LinearSVC, this is caused by the margin property of the hinge loss, which lets the model focus on hard samples that
are close to the decision boundary (the support vectors). Both kinds of calibration can fix this issue and yield nearly
identical results. The next figure shows the calibration curve of Gaussian naive Bayes on the same data, with both
kinds of calibration and also without calibration.
One can see that Gaussian naive Bayes performs very badly but does so in an other way than linear SVC: While linear
SVC exhibited a sigmoid calibration curve, Gaussian naive Bayes’ calibration curve has a transposed-sigmoid shape.
This is typical for an over-confident classifier. In this case, the classifier’s overconfidence is caused by the redundant
features which violate the naive Bayes assumption of feature-independence.
Calibration of the probabilities of Gaussian naive Bayes with isotonic regression can fix this issue as can be seen from



the nearly diagonal calibration curve. Sigmoid calibration also improves the brier score slightly, albeit not as strongly
as the non-parametric isotonic calibration. This is an intrinsic limitation of sigmoid calibration, whose parametric form
assumes a sigmoid rather than a transposed-sigmoid curve. The non-parametric isotonic calibration model, however,
makes no such strong assumptions and can deal with either shape, provided that there is sufficient calibration data. In
general, sigmoid calibration is preferable in cases where the calibration curve is sigmoid and where there is limited
calibration data, while isotonic calibration is preferable for non-sigmoid calibration curves and in situations where
large amounts of data are available for calibration.
CalibratedClassifierCV can also deal with classification tasks that involve more than two classes if the base
estimator can do so. In this case, the classifier is calibrated first for each class separately in an one-vs-rest fashion.
When predicting probabilities for unseen data, the calibrated probabilities for each class are predicted separately. As
those probabilities do not necessarily sum to one, a postprocessing is performed to normalize them.
The next image illustrates how sigmoid calibration changes predicted probabilities for a 3-class classification problem.
Illustrated is the standard 2-simplex, where the three corners correspond to the three classes. Arrows point from the
probability vectors predicted by an uncalibrated classifier to the probability vectors predicted by the same classifier
after sigmoid calibration on a hold-out validation set. Colors indicate the true class of an instance (red: class 1, green:
class 2, blue: class 3).
The base classifier is a random forest classifier with 25 base estimators (trees). If this classifier is trained on all 800
training datapoints, it is overly confident in its predictions and thus incurs a large log-loss. Calibrating an identical
classifier, which was trained on 600 datapoints, with method=’sigmoid’ on the remaining 200 datapoints reduces the
confidence of the predictions, i.e., moves the probability vectors from the edges of the simplex towards the center:
This calibration results in a lower log-loss. Note that an alternative would have been to increase the number of base


estimators which would have resulted in a similar decrease in log-loss.
References:
• Obtaining calibrated probability estimates from decision trees and naive Bayesian classifiers, B. Zadrozny &
C. Elkan, ICML 2001
• Transforming Classifier Scores into Accurate Multiclass Probability Estimates, B. Zadrozny & C. Elkan,
(KDD 2002)
• Probabilistic Outputs for Support Vector Machines and Comparisons to Regularized Likelihood Methods, J.
Platt, (1999)
3.1.17 Neural network models (supervised)
Warning: This implementation is not intended for large-scale applications. In particular, scikit-learn offers no
GPU support. For much faster, GPU-based implementations, as well as frameworks offering much more flexibility
to build deep learning architectures, see Related Projects.
Multi-layer Perceptron
Multi-layer Perceptron (MLP) is a supervised learning algorithm that learns a function 𝑓 (·) : 𝑅𝑚 → 𝑅𝑜 by training
on a dataset, where 𝑚 is the number of dimensions for input and 𝑜 is the number of dimensions for output. Given a set
of features 𝑋 = 𝑥1 , 𝑥2 , ..., 𝑥𝑚 and a target 𝑦, it can learn a non-linear function approximator for either classification
or regression. It is different from logistic regression, in that between the input and the output layer, there can be one
or more non-linear layers, called hidden layers. Figure 1 shows a one hidden layer MLP with scalar output.
Fig. 3.2: Figure 1 : One hidden layer MLP.
The leftmost layer, known as the input layer, consists of a set of neurons {𝑥𝑖 |𝑥1 , 𝑥2 , ..., 𝑥𝑚 } representing the input
features. Each neuron in the hidden layer transforms the values from the previous layer with a weighted linear sum-
mation 𝑤1 𝑥1 + 𝑤2 𝑥2 + ... + 𝑤𝑚 𝑥𝑚 , followed by a non-linear activation function 𝑔(·) : 𝑅 → 𝑅 - like the hyperbolic
tan function. The output layer receives the values from the last hidden layer and transforms them into output values.

The module contains the public attributes coefs_ and intercepts_. coefs_ is a list of weight matrices, where
weight matrix at index 𝑖 represents the weights between layer 𝑖 and layer 𝑖+1. intercepts_ is a list of bias vectors,
where the vector at index 𝑖 represents the bias values added to layer 𝑖 + 1.
The advantages of Multi-layer Perceptron are:
• Capability to learn non-linear models.
• Capability to learn models in real-time (on-line learning) using partial_fit.
The disadvantages of Multi-layer Perceptron (MLP) include:
• MLP with hidden layers have a non-convex loss function where there exists more than one local minimum.
Therefore different random weight initializations can lead to different validation accuracy.
• MLP requires tuning a number of hyperparameters such as the number of hidden neurons, layers, and iterations.
• MLP is sensitive to feature scaling.
Please see Tips on Practical Use section that addresses some of these disadvantages.
Classification
Class MLPClassifier implements a multi-layer perceptron (MLP) algorithm that trains using Backpropagation.
MLP trains on two arrays: array X of size (n_samples, n_features), which holds the training samples represented as
floating point feature vectors; and array y of size (n_samples,), which holds the target values (class labels) for the
training samples:
>>> from sklearn.neural_network import MLPClassifier

>>> X = [[0., 0.], [1., 1.]]
>>> y = [0, 1]
>>> clf = MLPClassifier(solver='lbfgs', alpha=1e-5,
... hidden_layer_sizes=(5, 2), random_state=1)
...
>>> clf.fit(X, y)
MLPClassifier(activation='relu', alpha=1e-05, batch_size='auto',
beta_1=0.9, beta_2=0.999, early_stopping=False,
epsilon=1e-08, hidden_layer_sizes=(5, 2),
learning_rate='constant', learning_rate_init=0.001,
max_iter=200, momentum=0.9, n_iter_no_change=10,
nesterovs_momentum=True, power_t=0.5, random_state=1,
shuffle=True, solver='lbfgs', tol=0.0001,
validation_fraction=0.1, verbose=False, warm_start=False)
After fitting (training), the model can predict labels for new samples:
>>> clf.predict([[2., 2.], [-1., -2.]])

array([1, 0])
MLP can fit a non-linear model to the training data. clf.coefs_ contains the weight matrices that constitute the
model parameters:
>>> [coef.shape for coef in clf.coefs_]

[(2, 5), (5, 2), (2, 1)]
Currently, MLPClassifier supports only the Cross-Entropy loss function, which allows probability estimates by
running the predict_proba method.

MLP trains using Backpropagation. More precisely, it trains using some form of gradient descent and the gradients
are calculated using Backpropagation. For classification, it minimizes the Cross-Entropy loss function, giving a vector
of probability estimates 𝑃 (𝑦|𝑥) per sample 𝑥:
>>> clf.predict_proba([[2., 2.], [1., 2.]])

array([[1.967...e-04, 9.998...-01],
[1.967...e-04, 9.998...-01]])
MLPClassifier supports multi-class classification by applying Softmax as the output function.

Further, the model supports multi-label classification in which a sample can belong to more than one class. For each
class, the raw output passes through the logistic function. Values larger or equal to 0.5 are rounded to 1, otherwise to
0. For a predicted output of a sample, the indices where the value is 1 represents the assigned classes of that sample:
>>> X = [[0., 0.], [1., 1.]]

>>> y = [[0, 1], [1, 1]]
>>> clf = MLPClassifier(solver='lbfgs', alpha=1e-5,
... hidden_layer_sizes=(15,), random_state=1)
...
>>> clf.fit(X, y)
MLPClassifier(activation='relu', alpha=1e-05, batch_size='auto',
beta_1=0.9, beta_2=0.999, early_stopping=False,
epsilon=1e-08, hidden_layer_sizes=(15,),
learning_rate='constant', learning_rate_init=0.001,
max_iter=200, momentum=0.9, n_iter_no_change=10,
nesterovs_momentum=True, power_t=0.5, random_state=1,
shuffle=True, solver='lbfgs', tol=0.0001,
validation_fraction=0.1, verbose=False, warm_start=False)
>>> clf.predict([[1., 2.]])
array([[1, 1]])
>>> clf.predict([[0., 0.]])
array([[0, 1]])
See the examples below and the doc string of MLPClassifier.fit for further information.
Examples:
• Compare Stochastic learning strategies for MLPClassifier

• Visualization of MLP weights on MNIST
Regression
Class MLPRegressor implements a multi-layer perceptron (MLP) that trains using backpropagation with no activa-
tion function in the output layer, which can also be seen as using the identity function as activation function. Therefore,
it uses the square error as the loss function, and the output is a set of continuous values.
MLPRegressor also supports multi-output regression, in which a sample can have more than one target.
Regularization
Both MLPRegressor and MLPClassifier use parameter alpha for regularization (L2 regularization) term
which helps in avoiding overfitting by penalizing weights with large magnitudes. Following plot displays varying
decision function with value of alpha.

See the examples below for further information.
Examples:
• Varying regularization in Multi-layer Perceptron
Algorithms
MLP trains using Stochastic Gradient Descent, Adam, or L-BFGS. Stochastic Gradient Descent (SGD) updates pa-
rameters using the gradient of the loss function with respect to a parameter that needs adaptation, i.e.
𝜕𝑅(𝑤) 𝜕𝐿𝑜𝑠𝑠
𝑤 ← 𝑤 − 𝜂(𝛼 + )
𝜕𝑤 𝜕𝑤
where 𝜂 is the learning rate which controls the step-size in the parameter space search. 𝐿𝑜𝑠𝑠 is the loss function used
for the network.
More details can be found in the documentation of SGD
Adam is similar to SGD in a sense that it is a stochastic optimizer, but it can automatically adjust the amount to update
parameters based on adaptive estimates of lower-order moments.
With SGD or Adam, training supports online and mini-batch learning.
L-BFGS is a solver that approximates the Hessian matrix which represents the second-order partial derivative of a
function. Further it approximates the inverse of the Hessian matrix to perform parameter updates. The implementation
uses the Scipy version of L-BFGS.
If the selected solver is ‘L-BFGS’, training does not support online nor mini-batch learning.

Complexity
Suppose there are 𝑛 training samples, 𝑚 features, 𝑘 hidden layers, each containing ℎ neurons - for simplicity, and 𝑜
output neurons. The time complexity of backpropagation is 𝑂(𝑛 · 𝑚 · ℎ𝑘 · 𝑜 · 𝑖), where 𝑖 is the number of iterations.
Since backpropagation has a high time complexity, it is advisable to start with smaller number of hidden neurons and
few hidden layers for training.
Given a set of training examples (𝑥1 , 𝑦1 ), (𝑥2 , 𝑦2 ), . . . , (𝑥𝑛 , 𝑦𝑛 ) where 𝑥𝑖 ∈ R𝑛 and 𝑦𝑖 ∈ {0, 1}, a one hidden layer
one hidden neuron MLP learns the function 𝑓 (𝑥) = 𝑊2 𝑔(𝑊1𝑇 𝑥 + 𝑏1 ) + 𝑏2 where 𝑊1 ∈ R𝑚 and 𝑊2 , 𝑏1 , 𝑏2 ∈ R are
model parameters. 𝑊1 , 𝑊2 represent the weights of the input layer and hidden layer, respectively; and 𝑏1 , 𝑏2 represent
the bias added to the hidden layer and the output layer, respectively. 𝑔(·) : 𝑅 → 𝑅 is the activation function, set by
default as the hyperbolic tan. It is given as,
𝑒𝑧 − 𝑒−𝑧
𝑔(𝑧) =
𝑒𝑧 + 𝑒−𝑧
For binary classification, 𝑓 (𝑥) passes through the logistic function 𝑔(𝑧) = 1/(1+𝑒−𝑧 ) to obtain output values between
zero and one. A threshold, set to 0.5, would assign samples of outputs larger or equal 0.5 to the positive class, and the
rest to the negative class.
If there are more than two classes, 𝑓 (𝑥) itself would be a vector of size (n_classes,). Instead of passing through logistic
function, it passes through the softmax function, which is written as,
exp(𝑧𝑖 )
softmax(𝑧)𝑖 = ∑︀𝑘
𝑙=1 exp(𝑧𝑙 )
where 𝑧𝑖 represents the 𝑖 th element of the input to softmax, which corresponds to class 𝑖, and 𝐾 is the number of
classes. The result is a vector containing the probabilities that sample 𝑥 belong to each class. The output is the class
with the highest probability.
In regression, the output remains as 𝑓 (𝑥); therefore, output activation function is just the identity function.
MLP uses different loss functions depending on the problem type. The loss function for classification is Cross-Entropy,
which in binary case is given as,
𝑦 , 𝑦, 𝑊 ) = −𝑦 ln 𝑦ˆ − (1 − 𝑦) ln (1 − 𝑦ˆ) + 𝛼||𝑊 ||22

𝐿𝑜𝑠𝑠(ˆ
where 𝛼||𝑊 ||22 is an L2-regularization term (aka penalty) that penalizes complex models; and 𝛼 > 0 is a non-negative
hyperparameter that controls the magnitude of the penalty.
For regression, MLP uses the Square Error loss function; written as,
1 𝛼
𝐿𝑜𝑠𝑠(ˆ
𝑦 , 𝑦, 𝑊 ) = 𝑦 − 𝑦||22 + ||𝑊 ||22
||ˆ
2 2
Starting from initial random weights, multi-layer perceptron (MLP) minimizes the loss function by repeatedly updating
these weights. After computing the loss, a backward pass propagates it from the output layer to the previous layers,
providing each weight parameter with an update value meant to decrease the loss.
In gradient descent, the gradient ∇𝐿𝑜𝑠𝑠𝑊 of the loss with respect to the weights is computed and deducted from 𝑊 .
More formally, this is expressed as,
𝑊 𝑖+1 = 𝑊 𝑖 − 𝜖∇𝐿𝑜𝑠𝑠𝑖𝑊
where 𝑖 is the iteration step, and 𝜖 is the learning rate with a value larger than 0.
The algorithm stops when it reaches a preset maximum number of iterations; or when the improvement in loss is below
a certain, small number.

• Multi-layer Perceptron is sensitive to feature scaling, so it is highly recommended to scale your data. For
example, scale each attribute on the input vector X to [0, 1] or [-1, +1], or standardize it to have mean 0 and
variance 1. Note that you must apply the same scaling to the test set for meaningful results. You can use
StandardScaler for standardization.
>>> from sklearn.preprocessing import StandardScaler

>>> scaler = StandardScaler()
>>> # Don't cheat - fit only on training data
>>> scaler.fit(X_train)
>>> X_train = scaler.transform(X_train)
>>> # apply same transformation to test data
>>> X_test = scaler.transform(X_test)
An alternative and recommended approach is to use StandardScaler in a Pipeline

• Finding a reasonable regularization parameter 𝛼 is best done using GridSearchCV, usually in the range 10.0
** -np.arange(1, 7).
• Empirically, we observed that L-BFGS converges faster and with better solutions on small datasets. For relatively
large datasets, however, Adam is very robust. It usually converges quickly and gives pretty good performance.
SGD with momentum or nesterov’s momentum, on the other hand, can perform better than those two algorithms
if learning rate is correctly tuned.
More control with warm_start
If you want more control over stopping criteria or learning rate in SGD, or want to do additional monitoring, using
warm_start=True and max_iter=1 and iterating yourself can be helpful:
>>> X = [[0., 0.], [1., 1.]]

>>> y = [0, 1]
>>> clf = MLPClassifier(hidden_layer_sizes=(15,), random_state=1, max_iter=1, warm_
˓→start=True)
>>> for i in range(10):

... clf.fit(X, y)
... # additional monitoring / inspection
MLPClassifier(...
References:
• “Learning representations by back-propagating errors.” Rumelhart, David E., Geoffrey E. Hinton, and Ronald
J. Williams.
• “Stochastic Gradient Descent” L. Bottou - Website, 2010.
• “Backpropagation” Andrew Ng, Jiquan Ngiam, Chuan Yu Foo, Yifan Mai, Caroline Suen - Website, 2011.
• “Efficient BackProp” Y. LeCun, L. Bottou, G. Orr, K. Müller - In Neural Networks: Tricks of the Trade 1998.
• “Adam: A method for stochastic optimization.” Kingma, Diederik, and Jimmy Ba. arXiv preprint
arXiv:1412.6980 (2014).

3.2 Unsupervised learning
3.2.1 Gaussian mixture models
sklearn.mixture is a package which enables one to learn Gaussian Mixture Models (diagonal, spherical, tied
and full covariance matrices supported), sample them, and estimate them from data. Facilities to help determine the
appropriate number of components are also provided.
Fig. 3.3: Two-component Gaussian mixture model: data points, and equi-probability surfaces of the model.
A Gaussian mixture model is a probabilistic model that assumes all the data points are generated from a mixture of a
finite number of Gaussian distributions with unknown parameters. One can think of mixture models as generalizing
k-means clustering to incorporate information about the covariance structure of the data as well as the centers of the
latent Gaussians.
Scikit-learn implements different classes to estimate Gaussian mixture models, that correspond to different estimation
strategies, detailed below.
Gaussian Mixture
The GaussianMixture object implements the expectation-maximization (EM) algorithm for fitting mixture-of-
Gaussian models. It can also draw confidence ellipsoids for multivariate models, and compute the Bayesian Infor-
mation Criterion to assess the number of clusters in the data. A GaussianMixture.fit method is provided that
learns a Gaussian Mixture Model from train data. Given test data, it can assign to each sample the Gaussian it mostly
probably belong to using the GaussianMixture.predict method.
The GaussianMixture comes with different options to constrain the covariance of the difference classes estimated:
spherical, diagonal, tied or full covariance.
Examples:
• See GMM covariances for an example of using the Gaussian mixture as clustering on the iris dataset.
• See Density Estimation for a Gaussian mixture for an example on plotting the density estimation.
3.2. Unsupervised learning 307

Pros and cons of class GaussianMixture
Pros
Speed It is the fastest algorithm for learning mixture models

Agnostic As this algorithm maximizes only the likelihood, it will not bias the means towards zero, or
bias the cluster sizes to have specific structures that might or might not apply.
Cons
Singularities When one has insufficiently many points per mixture, estimating the covariance matrices
becomes difficult, and the algorithm is known to diverge and find solutions with infinite likelihood
unless one regularizes the covariances artificially.
Number of components This algorithm will always use all the components it has access to, needing
held-out data or information theoretical criteria to decide how many components to use in the ab-
sence of external cues.
Selecting the number of components in a classical Gaussian Mixture Model
The BIC criterion can be used to select the number of components in a Gaussian Mixture in an efficient way. In theory,
it recovers the true number of components only in the asymptotic regime (i.e. if much data is available and assuming

that the data was actually generated i.i.d. from a mixture of Gaussian distribution). Note that using a Variational
Bayesian Gaussian mixture avoids the specification of the number of components for a Gaussian mixture model.
Examples:
• See Gaussian Mixture Model Selection for an example of model selection performed with classical Gaussian
mixture.
Estimation algorithm Expectation-maximization
The main difficulty in learning Gaussian mixture models from unlabeled data is that it is one usually doesn’t know
which points came from which latent component (if one has access to this information it gets very easy to fit a separate
Gaussian distribution to each set of points). Expectation-maximization is a well-founded statistical algorithm to get
around this problem by an iterative process. First one assumes random components (randomly centered on data points,
learned from k-means, or even just normally distributed around the origin) and computes for each point a probability
of being generated by each component of the model. Then, one tweaks the parameters to maximize the likelihood of
the data given those assignments. Repeating this process is guaranteed to always converge to a local optimum.
Variational Bayesian Gaussian Mixture
The BayesianGaussianMixture object implements a variant of the Gaussian mixture model with variational
inference algorithms. The API is similar as the one defined by GaussianMixture.
Estimation algorithm: variational inference
Variational inference is an extension of expectation-maximization that maximizes a lower bound on model evidence
(including priors) instead of data likelihood. The principle behind variational methods is the same as expectation-
maximization (that is both are iterative algorithms that alternate between finding the probabilities for each point to

be generated by each mixture and fitting the mixture to these assigned points), but variational methods add regular-
ization by integrating information from prior distributions. This avoids the singularities often found in expectation-
maximization solutions but introduces some subtle biases to the model. Inference is often notably slower, but not
usually as much so as to render usage unpractical.
Due to its Bayesian nature, the variational algorithm needs more hyperparameters than expectation-maximization,
the most important of these being the concentration parameter weight_concentration_prior. Specifying a
low value for the concentration prior will make the model put most of the weight on few components set the remain-
ing components weights very close to zero. High values of the concentration prior will allow a larger number of
components to be active in the mixture.
The parameters implementation of the BayesianGaussianMixture class proposes two types of prior for the
weights distribution: a finite mixture model with Dirichlet distribution and an infinite mixture model with the Dirichlet
Process. In practice Dirichlet Process inference algorithm is approximated and uses a truncated distribution with a fixed
maximum number of components (called the Stick-breaking representation). The number of components actually used
almost always depends on the data.
The next figure compares the results obtained for the different type of the weight concentration prior (parameter
weight_concentration_prior_type) for different values of weight_concentration_prior. Here,
we can see the value of the weight_concentration_prior parameter has a strong impact on the effective
number of active components obtained. We can also notice that large values for the concentration weight prior lead
to more uniform weights when the type of prior is ‘dirichlet_distribution’ while this is not necessarily the case for the
‘dirichlet_process’ type (used by default).

The examples below compare Gaussian mixture models with a fixed number of components, to the variational Gaus-
sian mixture models with a Dirichlet process prior. Here, a classical Gaussian mixture is fitted with 5 components on
a dataset composed of 2 clusters. We can see that the variational Gaussian mixture with a Dirichlet process prior is
able to limit itself to only 2 components whereas the Gaussian mixture fits the data with a fixed number of components
that has to be set a priori by the user. In this case the user has selected n_components=5 which does not match the
true generative distribution of this toy dataset. Note that with very little observations, the variational Gaussian mixture
models with a Dirichlet process prior can take a conservative stand, and fit only one component.
On the following figure we are fitting a dataset not well-depicted by a Gaussian mixture. Adjusting the
weight_concentration_prior, parameter of the BayesianGaussianMixture controls the number of

components used to fit this data. We also present on the last two plots a random sampling generated from the two
resulting mixtures.
Examples:
• See Gaussian Mixture Model Ellipsoids for an example on plotting the confidence ellipsoids for both
GaussianMixture and BayesianGaussianMixture.
• Gaussian Mixture Model Sine Curve shows using GaussianMixture and
BayesianGaussianMixture to fit a sine wave.
• See Concentration Prior Type Analysis of Variation Bayesian Gaussian Mixture for an ex-
ample plotting the confidence ellipsoids for the BayesianGaussianMixture with dif-

ferent weight_concentration_prior_type for different values of the parameter

weight_concentration_prior.
Pros and cons of variational inference with BayesianGaussianMixture
Pros
Automatic selection when weight_concentration_prior is small enough and

n_components is larger than what is found necessary by the model, the Variational Bayesian
mixture model has a natural tendency to set some mixture weights values close to zero. This makes
it possible to let the model choose a suitable number of effective components automatically. Only an
upper bound of this number needs to be provided. Note however that the “ideal” number of active
components is very application specific and is typically ill-defined in a data exploration setting.
Less sensitivity to the number of parameters unlike finite models, which will almost always use
all components as much as they can, and hence will produce wildly different solutions for
different numbers of components, the variational inference with a Dirichlet process prior
(weight_concentration_prior_type='dirichlet_process') won’t change much
with changes to the parameters, leading to more stability and less tuning.
Regularization due to the incorporation of prior information, variational solutions have less pathological
special cases than expectation-maximization solutions.
Cons
Speed the extra parametrization necessary for variational inference make inference slower, although not
by much.
Hyperparameters this algorithm needs an extra hyperparameter that might need experimental tuning via
cross-validation.
Bias there are many implicit biases in the inference algorithms (and also in the Dirichlet process if used),
and whenever there is a mismatch between these biases and the data it might be possible to fit better
models using a finite mixture.
The Dirichlet Process
Here we describe variational inference algorithms on Dirichlet process mixture. The Dirichlet process is a prior
probability distribution on clusterings with an infinite, unbounded, number of partitions. Variational techniques let us
incorporate this prior structure on Gaussian mixture models at almost no penalty in inference time, comparing with a
finite Gaussian mixture model.
An important question is how can the Dirichlet process use an infinite, unbounded number of clusters and still be
consistent. While a full explanation doesn’t fit this manual, one can think of its stick breaking process analogy to help
understanding it. The stick breaking process is a generative story for the Dirichlet process. We start with a unit-length
stick and in each step we break off a portion of the remaining stick. Each time, we associate the length of the piece of
the stick to the proportion of points that falls into a group of the mixture. At the end, to represent the infinite mixture,
we associate the last remaining piece of the stick to the proportion of points that don’t fall into all the other groups. The
length of each piece is a random variable with probability proportional to the concentration parameter. Smaller value
of the concentration will divide the unit-length into larger pieces of the stick (defining more concentrated distribution).
Larger concentration values will create smaller pieces of the stick (increasing the number of components with non
zero weights).

Variational inference techniques for the Dirichlet process still work with a finite approximation to this infinite mixture
model, but instead of having to specify a priori how many components one wants to use, one just specifies the concen-
tration parameter and an upper bound on the number of mixture components (this upper bound, assuming it is higher
than the “true” number of components, affects only algorithmic complexity, not the actual number of components
used).
3.2.2 Manifold learning
Look for the bare necessities

The simple bare necessities
Forget about your worries and your strife
I mean the bare necessities
Old Mother Nature’s recipes
That bring the bare necessities of life
– Baloo’s song [The Jungle Book]
Manifold learning is an approach to non-linear dimensionality reduction. Algorithms for this task are based on the
idea that the dimensionality of many data sets is only artificially high.
Introduction
High-dimensional datasets can be very difficult to visualize. While data in two or three dimensions can be plotted to
show the inherent structure of the data, equivalent high-dimensional plots are much less intuitive. To aid visualization
of the structure of a dataset, the dimension must be reduced in some way.
The simplest way to accomplish this dimensionality reduction is by taking a random projection of the data. Though
this allows some degree of visualization of the data structure, the randomness of the choice leaves much to be desired.
In a random projection, it is likely that the more interesting structure within the data will be lost.

To address this concern, a number of supervised and unsupervised linear dimensionality reduction frameworks have
been designed, such as Principal Component Analysis (PCA), Independent Component Analysis, Linear Discriminant
Analysis, and others. These algorithms define specific rubrics to choose an “interesting” linear projection of the data.
These methods can be powerful, but often miss important non-linear structure in the data.
Manifold Learning can be thought of as an attempt to generalize linear frameworks like PCA to be sensitive to non-
linear structure in data. Though supervised variants exist, the typical manifold learning problem is unsupervised: it
learns the high-dimensional structure of the data from the data itself, without the use of predetermined classifications.
Examples:
• See Manifold learning on handwritten digits: Locally Linear Embedding, Isomap. . . for an example of
dimensionality reduction on handwritten digits.
• See Comparison of Manifold Learning methods for an example of dimensionality reduction on a toy “S-
curve” dataset.
The manifold learning implementations available in scikit-learn are summarized below
Isomap
One of the earliest approaches to manifold learning is the Isomap algorithm, short for Isometric Mapping. Isomap can
be viewed as an extension of Multi-dimensional Scaling (MDS) or Kernel PCA. Isomap seeks a lower-dimensional

embedding which maintains geodesic distances between all points. Isomap can be performed with the object Isomap.
Complexity
The Isomap algorithm comprises three stages:

1. Nearest neighbor search. Isomap uses sklearn.neighbors.BallTree for efficient neighbor search.
The cost is approximately 𝑂[𝐷 log(𝑘)𝑁 log(𝑁 )], for 𝑘 nearest neighbors of 𝑁 points in 𝐷 dimensions.
2. Shortest-path graph search. The most efficient known algorithms for this are Dijkstra’s Algorithm, which is
approximately 𝑂[𝑁 2 (𝑘 + log(𝑁 ))], or the Floyd-Warshall algorithm, which is 𝑂[𝑁 3 ]. The algorithm can be
selected by the user with the path_method keyword of Isomap. If unspecified, the code attempts to choose
the best algorithm for the input data.
3. Partial eigenvalue decomposition. The embedding is encoded in the eigenvectors corresponding to the 𝑑
largest eigenvalues of the 𝑁 × 𝑁 isomap kernel. For a dense solver, the cost is approximately 𝑂[𝑑𝑁 2 ]. This
cost can often be improved using the ARPACK solver. The eigensolver can be specified by the user with the
path_method keyword of Isomap. If unspecified, the code attempts to choose the best algorithm for the
input data.
The overall complexity of Isomap is 𝑂[𝐷 log(𝑘)𝑁 log(𝑁 )] + 𝑂[𝑁 2 (𝑘 + log(𝑁 ))] + 𝑂[𝑑𝑁 2 ].
• 𝑁 : number of training data points
• 𝐷 : input dimension
• 𝑘 : number of nearest neighbors
• 𝑑 : output dimension
References:
• “A global geometric framework for nonlinear dimensionality reduction” Tenenbaum, J.B.; De Silva, V.; &
Langford, J.C. Science 290 (5500)

Locally Linear Embedding
Locally linear embedding (LLE) seeks a lower-dimensional projection of the data which preserves distances within
local neighborhoods. It can be thought of as a series of local Principal Component Analyses which are globally
compared to find the best non-linear embedding.
Locally linear embedding can be performed with function locally_linear_embedding or its object-oriented
counterpart LocallyLinearEmbedding.
Complexity
The standard LLE algorithm comprises three stages:

1. Nearest Neighbors Search. See discussion under Isomap above.
2. Weight Matrix Construction. 𝑂[𝐷𝑁 𝑘 3 ]. The construction of the LLE weight matrix involves the solution of
a 𝑘 × 𝑘 linear equation for each of the 𝑁 local neighborhoods
3. Partial Eigenvalue Decomposition. See discussion under Isomap above.
The overall complexity of standard LLE is 𝑂[𝐷 log(𝑘)𝑁 log(𝑁 )] + 𝑂[𝐷𝑁 𝑘 3 ] + 𝑂[𝑑𝑁 2 ].
References:
• “Nonlinear dimensionality reduction by locally linear embedding” Roweis, S. & Saul, L. Science 290:2323
(2000)
Modified Locally Linear Embedding
One well-known issue with LLE is the regularization problem. When the number of neighbors is greater than the
number of input dimensions, the matrix defining each local neighborhood is rank-deficient. To address this, standard

LLE applies an arbitrary regularization parameter 𝑟, which is chosen relative to the trace of the local weight matrix.
Though it can be shown formally that as 𝑟 → 0, the solution converges to the desired embedding, there is no guarantee
that the optimal solution will be found for 𝑟 > 0. This problem manifests itself in embeddings which distort the
underlying geometry of the manifold.
One method to address the regularization problem is to use multiple weight vectors in each neighborhood.
This is the essence of modified locally linear embedding (MLLE). MLLE can be performed with function
locally_linear_embedding or its object-oriented counterpart LocallyLinearEmbedding, with the key-
word method = 'modified'. It requires n_neighbors > n_components.
Complexity
The MLLE algorithm comprises three stages:

1. Nearest Neighbors Search. Same as standard LLE
2. Weight Matrix Construction. Approximately 𝑂[𝐷𝑁 𝑘 3 ]+𝑂[𝑁 (𝑘 −𝐷)𝑘 2 ]. The first term is exactly equivalent
to that of standard LLE. The second term has to do with constructing the weight matrix from multiple weights.
In practice, the added cost of constructing the MLLE weight matrix is relatively small compared to the cost of
steps 1 and 3.
3. Partial Eigenvalue Decomposition. Same as standard LLE
The overall complexity of MLLE is 𝑂[𝐷 log(𝑘)𝑁 log(𝑁 )] + 𝑂[𝐷𝑁 𝑘 3 ] + 𝑂[𝑁 (𝑘 − 𝐷)𝑘 2 ] + 𝑂[𝑑𝑁 2 ].
References:
• “MLLE: Modified Locally Linear Embedding Using Multiple Weights” Zhang, Z. & Wang, J.

Hessian Eigenmapping
Hessian Eigenmapping (also known as Hessian-based LLE: HLLE) is another method of solving the regularization
problem of LLE. It revolves around a hessian-based quadratic form at each neighborhood which is used to recover
the locally linear structure. Though other implementations note its poor scaling with data size, sklearn imple-
ments some algorithmic improvements which make its cost comparable to that of other LLE variants for small output
dimension. HLLE can be performed with function locally_linear_embedding or its object-oriented counter-
part LocallyLinearEmbedding, with the keyword method = 'hessian'. It requires n_neighbors >
n_components * (n_components + 3) / 2.
Complexity
The HLLE algorithm comprises three stages:

2. Weight Matrix Construction. Approximately 𝑂[𝐷𝑁 𝑘 3 ] + 𝑂[𝑁 𝑑6 ]. The first term reflects a similar cost to
that of standard LLE. The second term comes from a QR decomposition of the local hessian estimator.
The overall complexity of standard HLLE is 𝑂[𝐷 log(𝑘)𝑁 log(𝑁 )] + 𝑂[𝐷𝑁 𝑘 3 ] + 𝑂[𝑁 𝑑6 ] + 𝑂[𝑑𝑁 2 ].
References:
• “Hessian Eigenmaps: Locally linear embedding techniques for high-dimensional data” Donoho, D. &
Grimes, C. Proc Natl Acad Sci USA. 100:5591 (2003)

Spectral Embedding
Spectral Embedding is an approach to calculating a non-linear embedding. Scikit-learn implements Laplacian Eigen-
maps, which finds a low dimensional representation of the data using a spectral decomposition of the graph Laplacian.
The graph generated can be considered as a discrete approximation of the low dimensional manifold in the high dimen-
sional space. Minimization of a cost function based on the graph ensures that points close to each other on the manifold
are mapped close to each other in the low dimensional space, preserving local distances. Spectral embedding can be
performed with the function spectral_embedding or its object-oriented counterpart SpectralEmbedding.
Complexity
The Spectral Embedding (Laplacian Eigenmaps) algorithm comprises three stages:

1. Weighted Graph Construction. Transform the raw input data into graph representation using affinity (adja-
cency) matrix representation.
2. Graph Laplacian Construction. unnormalized Graph Laplacian is constructed as 𝐿 = 𝐷 − 𝐴 for and normal-
1 1
ized one as 𝐿 = 𝐷− 2 (𝐷 − 𝐴)𝐷− 2 .
3. Partial Eigenvalue Decomposition. Eigenvalue decomposition is done on graph Laplacian
The overall complexity of spectral embedding is 𝑂[𝐷 log(𝑘)𝑁 log(𝑁 )] + 𝑂[𝐷𝑁 𝑘 3 ] + 𝑂[𝑑𝑁 2 ].
References:
• “Laplacian Eigenmaps for Dimensionality Reduction and Data Representation” M. Belkin, P. Niyogi, Neural
Computation, June 2003; 15 (6):1373-1396
Local Tangent Space Alignment
Though not technically a variant of LLE, Local tangent space alignment (LTSA) is algorithmically similar enough
to LLE that it can be put in this category. Rather than focusing on preserving neighborhood distances as in LLE,
LTSA seeks to characterize the local geometry at each neighborhood via its tangent space, and performs a global
optimization to align these local tangent spaces to learn the embedding. LTSA can be performed with function
locally_linear_embedding or its object-oriented counterpart LocallyLinearEmbedding, with the key-
word method = 'ltsa'.
Complexity
The LTSA algorithm comprises three stages:

2. Weight Matrix Construction. Approximately 𝑂[𝐷𝑁 𝑘 3 ] + 𝑂[𝑘 2 𝑑]. The first term reflects a similar cost to that
of standard LLE.

The overall complexity of standard LTSA is 𝑂[𝐷 log(𝑘)𝑁 log(𝑁 )] + 𝑂[𝐷𝑁 𝑘 3 ] + 𝑂[𝑘 2 𝑑] + 𝑂[𝑑𝑁 2 ].
References:
• “Principal manifolds and nonlinear dimensionality reduction via tangent space alignment” Zhang, Z. & Zha,
H. Journal of Shanghai Univ. 8:406 (2004)
Multi-dimensional Scaling (MDS)
Multidimensional scaling (MDS) seeks a low-dimensional representation of the data in which the distances respect well
the distances in the original high-dimensional space.
In general, is a technique used for analyzing similarity or dissimilarity data. MDS attempts to model similarity or
dissimilarity data as distances in a geometric spaces. The data can be ratings of similarity between objects, interaction
frequencies of molecules, or trade indices between countries.
There exists two types of MDS algorithm: metric and non metric. In the scikit-learn, the class MDS implements
both. In Metric MDS, the input similarity matrix arises from a metric (and thus respects the triangular inequality), the
distances between output two points are then set to be as close as possible to the similarity or dissimilarity data. In
the non-metric version, the algorithms will try to preserve the order of the distances, and hence seek for a monotonic
relationship between the distances in the embedded space and the similarities/dissimilarities.
Let 𝑆 be the similarity matrix, and 𝑋 the coordinates of the 𝑛 input points. Disparities 𝑑ˆ𝑖𝑗 are transformation of the
similarities chosen in some optimal ways. The objective, called the stress, is then defined by 𝑠𝑢𝑚𝑖<𝑗 𝑑𝑖𝑗 (𝑋) − 𝑑ˆ𝑖𝑗 (𝑋)
Metric MDS
The simplest metric MDS model, called absolute MDS, disparities are defined by 𝑑ˆ𝑖𝑗 = 𝑆𝑖𝑗 . With absolute MDS, the
value 𝑆𝑖𝑗 should then correspond exactly to the distance between point 𝑖 and 𝑗 in the embedding point.

Most commonly, disparities are set to 𝑑ˆ𝑖𝑗 = 𝑏𝑆𝑖𝑗 .
Nonmetric MDS
Non metric MDS focuses on the ordination of the data. If 𝑆𝑖𝑗 < 𝑆𝑘𝑙 , then the embedding should enforce 𝑑𝑖𝑗 < 𝑑𝑗𝑘 .
A simple algorithm to enforce that is to use a monotonic regression of 𝑑𝑖𝑗 on 𝑆𝑖𝑗 , yielding disparities 𝑑ˆ𝑖𝑗 in the same
order as 𝑆𝑖𝑗 .
A trivial solution to this problem is to set all the points on the origin. In order to avoid that, the disparities 𝑑ˆ𝑖𝑗 are
normalized.
References:
• “Modern Multidimensional Scaling - Theory and Applications” Borg, I.; Groenen P. Springer Series in Statis-
tics (1997)
• “Nonmetric multidimensional scaling: a numerical method” Kruskal, J. Psychometrika, 29 (1964)

• “Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis” Kruskal, J. Psychome-

trika, 29, (1964)
t-distributed Stochastic Neighbor Embedding (t-SNE)
t-SNE (TSNE) converts affinities of data points to probabilities. The affinities in the original space are represented by
Gaussian joint probabilities and the affinities in the embedded space are represented by Student’s t-distributions. This
allows t-SNE to be particularly sensitive to local structure and has a few other advantages over existing techniques:
• Revealing the structure at many scales on a single map
• Revealing data that lie in multiple, different, manifolds or clusters
• Reducing the tendency to crowd points together at the center
While Isomap, LLE and variants are best suited to unfold a single continuous low dimensional manifold, t-SNE will
focus on the local structure of the data and will tend to extract clustered local groups of samples as highlighted on the
S-curve example. This ability to group samples based on the local structure might be beneficial to visually disentangle
a dataset that comprises several manifolds at once as is the case in the digits dataset.
The Kullback-Leibler (KL) divergence of the joint probabilities in the original space and the embedded space will
be minimized by gradient descent. Note that the KL divergence is not convex, i.e. multiple restarts with different
initializations will end up in local minima of the KL divergence. Hence, it is sometimes useful to try different seeds
and select the embedding with the lowest KL divergence.
The disadvantages to using t-SNE are roughly:
• t-SNE is computationally expensive, and can take several hours on million-sample datasets where PCA will
finish in seconds or minutes
• The Barnes-Hut t-SNE method is limited to two or three dimensional embeddings.
• The algorithm is stochastic and multiple restarts with different seeds can yield different embeddings. However,
it is perfectly legitimate to pick the embedding with the least error.
• Global structure is not explicitly preserved. This is problem is mitigated by initializing points with PCA (using
init=’pca’).

Optimizing t-SNE
The main purpose of t-SNE is visualization of high-dimensional data. Hence, it works best when the data will be
embedded on two or three dimensions.
Optimizing the KL divergence can be a little bit tricky sometimes. There are five parameters that control the optimiza-
tion of t-SNE and therefore possibly the quality of the resulting embedding:
• perplexity
• early exaggeration factor
• learning rate
• maximum number of iterations
• angle (not used in the exact method)
The perplexity is defined as 𝑘 = 2(𝑆) where 𝑆 is the Shannon entropy of the conditional probability distribution.
The perplexity of a 𝑘-sided die is 𝑘, so that 𝑘 is effectively the number of nearest neighbors t-SNE considers when
generating the conditional probabilities. Larger perplexities lead to more nearest neighbors and less sensitive to small
structure. Conversely a lower perplexity considers a smaller number of neighbors, and thus ignores more global
information in favour of the local neighborhood. As dataset sizes get larger more points will be required to get a
reasonable sample of the local neighborhood, and hence larger perplexities may be required. Similarly noisier datasets
will require larger perplexity values to encompass enough local neighbors to see beyond the background noise.
The maximum number of iterations is usually high enough and does not need any tuning. The optimization consists of
two phases: the early exaggeration phase and the final optimization. During early exaggeration the joint probabilities
in the original space will be artificially increased by multiplication with a given factor. Larger factors result in larger
gaps between natural clusters in the data. If the factor is too high, the KL divergence could increase during this phase.
Usually it does not have to be tuned. A critical parameter is the learning rate. If it is too low gradient descent will get
stuck in a bad local minimum. If it is too high the KL divergence will increase during optimization. More tips can be
found in Laurens van der Maaten’s FAQ (see references). The last parameter, angle, is a tradeoff between performance
and accuracy. Larger angles imply that we can approximate larger regions by a single point, leading to better speed
but less accurate results.
“How to Use t-SNE Effectively” provides a good discussion of the effects of the various parameters, as well as
interactive plots to explore the effects of different parameters.
Barnes-Hut t-SNE
The Barnes-Hut t-SNE that has been implemented here is usually much slower than other manifold learning algo-
rithms. The optimization is quite difficult and the computation of the gradient is 𝑂[𝑑𝑁 𝑙𝑜𝑔(𝑁 )], where 𝑑 is the number
of output dimensions and 𝑁 is the number of samples. The Barnes-Hut method improves on the exact method where
t-SNE complexity is 𝑂[𝑑𝑁 2 ], but has several other notable differences:
• The Barnes-Hut implementation only works when the target dimensionality is 3 or less. The 2D case is typical
when building visualizations.
• Barnes-Hut only works with dense input data. Sparse data matrices can only be embedded with the exact method
or can be approximated by a dense low rank projection for instance using sklearn.decomposition.
TruncatedSVD
• Barnes-Hut is an approximation of the exact method. The approximation is parameterized with the angle pa-
rameter, therefore the angle parameter is unused when method=”exact”
• Barnes-Hut is significantly more scalable. Barnes-Hut can be used to embed hundred of thousands of data points
while the exact method can handle thousands of samples before becoming computationally intractable

For visualization purpose (which is the main use case of t-SNE), using the Barnes-Hut method is strongly recom-
mended. The exact t-SNE method is useful for checking the theoretically properties of the embedding possibly in
higher dimensional space but limit to small datasets due to computational constraints.
Also note that the digits labels roughly match the natural grouping found by t-SNE while the linear 2D projection of
the PCA model yields a representation where label regions largely overlap. This is a strong clue that this data can be
well separated by non linear methods that focus on the local structure (e.g. an SVM with a Gaussian RBF kernel).
However, failing to visualize well separated homogeneously labeled groups with t-SNE in 2D does not necessarily
imply that the data cannot be correctly classified by a supervised model. It might be the case that 2 dimensions are not
low enough to accurately represents the internal structure of the data.
References:
• “Visualizing High-Dimensional Data Using t-SNE” van der Maaten, L.J.P.; Hinton, G. Journal of Machine
Learning Research (2008)
• “t-Distributed Stochastic Neighbor Embedding” van der Maaten, L.J.P.
• “Accelerating t-SNE using Tree-Based Algorithms.” L.J.P. van der Maaten. Journal of Machine Learning
Research 15(Oct):3221-3245, 2014.
Tips on practical use
• Make sure the same scale is used over all features. Because manifold learning methods are based on a nearest-
neighbor search, the algorithm may perform poorly otherwise. See StandardScaler for convenient ways of
scaling heterogeneous data.
• The reconstruction error computed by each routine can be used to choose the optimal output dimension. For a
𝑑-dimensional manifold embedded in a 𝐷-dimensional parameter space, the reconstruction error will decrease
as n_components is increased until n_components == d.
• Note that noisy data can “short-circuit” the manifold, in essence acting as a bridge between parts of the manifold
that would otherwise be well-separated. Manifold learning on noisy and/or incomplete data is an active area of
research.
• Certain input configurations can lead to singular weight matrices, for example when more than two points in the
dataset are identical, or when the data is split into disjointed groups. In this case, solver='arpack' will
fail to find the null space. The easiest way to address this is to use solver='dense' which will work on a
singular matrix, though it may be very slow depending on the number of input points. Alternatively, one can
attempt to understand the source of the singularity: if it is due to disjoint sets, increasing n_neighbors may
help. If it is due to identical points in the dataset, removing these points may help.
See also:
Totally Random Trees Embedding can also be useful to derive non-linear representations of feature space, also it does
not perform dimensionality reduction.
3.2.3 Clustering
Clustering of unlabeled data can be performed with the module sklearn.cluster.

Each clustering algorithm comes in two variants: a class, that implements the fit method to learn the clusters on train
data, and a function, that, given train data, returns an array of integer labels corresponding to the different clusters. For
the class, the labels over the training data can be found in the labels_ attribute.

Input data
One important thing to note is that the algorithms implemented in this module can take different kinds of matrix as
input. All the methods accept standard data matrices of shape [n_samples, n_features]. These can be ob-
tained from the classes in the sklearn.feature_extraction module. For AffinityPropagation,
SpectralClustering and DBSCAN one can also input similarity matrices of shape [n_samples,
n_samples]. These can be obtained from the functions in the sklearn.metrics.pairwise module.
Overview of clustering methods
Fig. 3.4: A comparison of the clustering algorithms in scikit-learn

Method Parameters Scalability Usecase Geometry (metric

name used)
K-Means number of clus- Very large General-purpose, even clus- Distances between
ters n_samples, ter size, flat geometry, not points
medium too many clusters
n_clusters
with MiniBatch code
Affinity propa- damping, sam- Not scalable with Many clusters, uneven clus- Graph distance (e.g.
gation ple preference n_samples ter size, non-flat geometry nearest-neighbor
graph)
Mean-shift bandwidth Not scalable with Many clusters, uneven clus- Distances between
n_samples ter size, non-flat geometry points
Spectral clus- number of clus- Medium Few clusters, even cluster Graph distance (e.g.
tering ters n_samples, small size, non-flat geometry nearest-neighbor
n_clusters graph)
Ward hi- number of clus- Large n_samples Many clusters, possibly con- Distances between
erarchical ters and n_clusters nectivity constraints points
clustering
Agglomerative number of clus- Large n_samples Many clusters, possibly con- Any pairwise distance
clustering ters, linkage and n_clusters nectivity constraints, non
type, distance Euclidean distances
DBSCAN neighborhood Very large Non-flat geometry, uneven Distances between
size n_samples, cluster sizes nearest points
medium
n_clusters
Gaussian mix- many Not scalable Flat geometry, good for den- Mahalanobis dis-
tures sity estimation tances to centers
Birch branching fac- Large Large dataset, outlier re- Euclidean distance
tor, threshold, n_clusters moval, data reduction. between points
optional global and n_samples
clusterer.
Non-flat geometry clustering is useful when the clusters have a specific shape, i.e. a non-flat manifold, and the standard
euclidean distance is not the right metric. This case arises in the two top rows of the figure above.
Gaussian mixture models, useful for clustering, are described in another chapter of the documentation dedicated
to mixture models. KMeans can be seen as a special case of Gaussian mixture model with equal covariance per
component.
K-means
The KMeans algorithm clusters data by trying to separate samples in n groups of equal variance, minimizing a criterion
known as the inertia or within-cluster sum-of-squares. This algorithm requires the number of clusters to be specified.
It scales well to large number of samples and has been used across a large range of application areas in many different
fields.
The k-means algorithm divides a set of 𝑁 samples 𝑋 into 𝐾 disjoint clusters 𝐶, each described by the mean 𝜇𝑗 of
the samples in the cluster. The means are commonly called the cluster “centroids”; note that they are not, in general,
points from 𝑋, although they live in the same space. The K-means algorithm aims to choose centroids that minimise
the inertia, or within-cluster sum of squared criterion:
𝑛
∑︁
min (||𝑥𝑖 − 𝜇𝑗 ||2 )
𝜇𝑗 ∈𝐶
𝑖=0

Inertia, or the within-cluster sum of squares criterion, can be recognized as a measure of how internally coherent
clusters are. It suffers from various drawbacks:
• Inertia makes the assumption that clusters are convex and isotropic, which is not always the case. It responds
poorly to elongated clusters, or manifolds with irregular shapes.
• Inertia is not a normalized metric: we just know that lower values are better and zero is optimal. But in very
high-dimensional spaces, Euclidean distances tend to become inflated (this is an instance of the so-called “curse
of dimensionality”). Running a dimensionality reduction algorithm such as PCA prior to k-means clustering
can alleviate this problem and speed up the computations.
K-
means is often referred to as Lloyd’s algorithm. In basic terms, the algorithm has three steps. The first step chooses
the initial centroids, with the most basic method being to choose 𝑘 samples from the dataset 𝑋. After initialization,
K-means consists of looping between the two other steps. The first step assigns each sample to its nearest centroid.
The second step creates new centroids by taking the mean value of all of the samples assigned to each previous
centroid. The difference between the old and the new centroids are computed and the algorithm repeats these last two
steps until this value is less than a threshold. In other words, it repeats until the centroids do not move significantly.

K-means is equivalent to the expectation-maximization algorithm with a

small, all-equal, diagonal covariance matrix.
The algorithm can also be understood through the concept of Voronoi diagrams. First the Voronoi diagram of the points
is calculated using the current centroids. Each segment in the Voronoi diagram becomes a separate cluster. Secondly,
the centroids are updated to the mean of each segment. The algorithm then repeats this until a stopping criterion is
fulfilled. Usually, the algorithm stops when the relative decrease in the objective function between iterations is less
than the given tolerance value. This is not the case in this implementation: iteration stops when centroids move less
than the tolerance.
Given enough time, K-means will always converge, however this may be to a local minimum. This is highly depen-
dent on the initialization of the centroids. As a result, the computation is often done several times, with different
initializations of the centroids. One method to help address this issue is the k-means++ initialization scheme, which
has been implemented in scikit-learn (use the init='k-means++' parameter). This initializes the centroids to
be (generally) distant from each other, leading to provably better results than random initialization, as shown in the
reference.
The algorithm supports sample weights, which can be given by a parameter sample_weight. This allows to assign
more weight to some samples when computing cluster centers and values of inertia. For example, assigning a weight
of 2 to a sample is equivalent to adding a duplicate of that sample to the dataset 𝑋.
A parameter can be given to allow K-means to be run in parallel, called n_jobs. Giving this parameter a positive
value uses that many processors (default: 1). A value of -1 uses all available processors, with -2 using one less, and so
on. Parallelization generally speeds up computation at the cost of memory (in this case, multiple copies of centroids
need to be stored, one for each job).
Warning: The parallel version of K-Means is broken on OS X when numpy uses the Accelerate Framework. This
is expected behavior: Accelerate can be called after a fork but you need to execv the subprocess with the Python
binary (which multiprocessing does not do under posix).
K-means can be used for vector quantization. This is achieved using the transform method of a trained model of
KMeans.
Examples:
• Demonstration of k-means assumptions: Demonstrating when k-means performs intuitively and when it does
not
• A demo of K-Means clustering on the handwritten digits data: Clustering handwritten digits
References:
• “k-means++: The advantages of careful seeding” Arthur, David, and Sergei Vassilvitskii, Proceedings of

the eighteenth annual ACM-SIAM symposium on Discrete algorithms, Society for Industrial and Applied
Mathematics (2007)
Mini Batch K-Means
The MiniBatchKMeans is a variant of the KMeans algorithm which uses mini-batches to reduce the computation
time, while still attempting to optimise the same objective function. Mini-batches are subsets of the input data, ran-
domly sampled in each training iteration. These mini-batches drastically reduce the amount of computation required
to converge to a local solution. In contrast to other algorithms that reduce the convergence time of k-means, mini-batch
k-means produces results that are generally only slightly worse than the standard algorithm.
The algorithm iterates between two major steps, similar to vanilla k-means. In the first step, 𝑏 samples are drawn
randomly from the dataset, to form a mini-batch. These are then assigned to the nearest centroid. In the second step,
the centroids are updated. In contrast to k-means, this is done on a per-sample basis. For each sample in the mini-batch,
the assigned centroid is updated by taking the streaming average of the sample and all previous samples assigned to
that centroid. This has the effect of decreasing the rate of change for a centroid over time. These steps are performed
until convergence or a predetermined number of iterations is reached.
MiniBatchKMeans converges faster than KMeans, but the quality of the results is reduced. In practice this differ-
ence in quality can be quite small, as shown in the example and cited reference.
Examples:
• Comparison of the K-Means and MiniBatchKMeans clustering algorithms: Comparison of KMeans and
MiniBatchKMeans
• Clustering text documents using k-means: Document clustering using sparse MiniBatchKMeans
• Online learning of a dictionary of parts of faces
References:
• “Web Scale K-Means clustering” D. Sculley, Proceedings of the 19th international conference on World wide
web (2010)

Affinity Propagation
AffinityPropagation creates clusters by sending messages between pairs of samples until convergence. A
dataset is then described using a small number of exemplars, which are identified as those most representative of other
samples. The messages sent between pairs represent the suitability for one sample to be the exemplar of the other,
which is updated in response to the values from other pairs. This updating happens iteratively until convergence, at
which point the final exemplars are chosen, and hence the final clustering is given.
Affinity Propagation can be interesting as it chooses the number of clusters based on the data provided. For this pur-
pose, the two important parameters are the preference, which controls how many exemplars are used, and the damping
factor which damps the responsibility and availability messages to avoid numerical oscillations when updating these
messages.
The main drawback of Affinity Propagation is its complexity. The algorithm has a time complexity of the order
𝑂(𝑁 2 𝑇 ), where 𝑁 is the number of samples and 𝑇 is the number of iterations until convergence. Further, the memory
complexity is of the order 𝑂(𝑁 2 ) if a dense similarity matrix is used, but reducible if a sparse similarity matrix is
used. This makes Affinity Propagation most appropriate for small to medium sized datasets.
Examples:
• Demo of affinity propagation clustering algorithm: Affinity Propagation on a synthetic 2D datasets with 3
classes.
• Visualizing the stock market structure Affinity Propagation on Financial time series to find groups of compa-
nies
Algorithm description: The messages sent between points belong to one of two categories. The first is the responsi-
bility 𝑟(𝑖, 𝑘), which is the accumulated evidence that sample 𝑘 should be the exemplar for sample 𝑖. The second is the
availability 𝑎(𝑖, 𝑘) which is the accumulated evidence that sample 𝑖 should choose sample 𝑘 to be its exemplar, and
considers the values for all other samples that 𝑘 should be an exemplar. In this way, exemplars are chosen by samples
if they are (1) similar enough to many samples and (2) chosen by many samples to be representative of themselves.
More formally, the responsibility of a sample 𝑘 to be the exemplar of sample 𝑖 is given by:
𝑟(𝑖, 𝑘) ← 𝑠(𝑖, 𝑘) − 𝑚𝑎𝑥[𝑎(𝑖, 𝑘 ′ ) + 𝑠(𝑖, 𝑘 ′ )∀𝑘 ′ ̸= 𝑘]
Where 𝑠(𝑖, 𝑘) is the similarity between samples 𝑖 and 𝑘. The availability of sample 𝑘 to be the exemplar of sample 𝑖 is

given by:
∑︁
𝑎(𝑖, 𝑘) ← 𝑚𝑖𝑛[0, 𝑟(𝑘, 𝑘) + 𝑟(𝑖′ , 𝑘)]
𝑖′ 𝑠.𝑡. 𝑖′ ∈{𝑖,𝑘}
/
To begin with, all values for 𝑟 and 𝑎 are set to zero, and the calculation of each iterates until convergence. As discussed
above, in order to avoid numerical oscillations when updating the messages, the damping factor 𝜆 is introduced to
iteration process:
𝑟𝑡+1 (𝑖, 𝑘) = 𝜆 · 𝑟𝑡 (𝑖, 𝑘) + (1 − 𝜆) · 𝑟𝑡+1 (𝑖, 𝑘)
𝑎𝑡+1 (𝑖, 𝑘) = 𝜆 · 𝑎𝑡 (𝑖, 𝑘) + (1 − 𝜆) · 𝑎𝑡+1 (𝑖, 𝑘)

where 𝑡 indicates the iteration times.
Mean Shift
MeanShift clustering aims to discover blobs in a smooth density of samples. It is a centroid based algorithm, which
works by updating candidates for centroids to be the mean of the points within a given region. These candidates are
then filtered in a post-processing stage to eliminate near-duplicates to form the final set of centroids.
Given a candidate centroid 𝑥𝑖 for iteration 𝑡, the candidate is updated according to the following equation:
𝑥𝑡+1
𝑖 = 𝑚(𝑥𝑡𝑖 )
Where 𝑁 (𝑥𝑖 ) is the neighborhood of samples within a given distance around 𝑥𝑖 and 𝑚 is the mean shift vector that
is computed for each centroid that points towards a region of the maximum increase in the density of points. This
is computed using the following equation, effectively updating a centroid to be the mean of the samples within its
neighborhood:
∑︀
𝑥 ∈𝑁 (𝑥𝑖 ) 𝐾(𝑥𝑗 − 𝑥𝑖 )𝑥𝑗
𝑚(𝑥𝑖 ) = ∑︀𝑗
𝑥𝑗 ∈𝑁 (𝑥𝑖 ) 𝐾(𝑥𝑗 − 𝑥𝑖 )
The algorithm automatically sets the number of clusters, instead of relying on a parameter bandwidth, which dictates
the size of the region to search through. This parameter can be set manually, but can be estimated using the provided
estimate_bandwidth function, which is called if the bandwidth is not set.
The algorithm is not highly scalable, as it requires multiple nearest neighbor searches during the execution of the
algorithm. The algorithm is guaranteed to converge, however the algorithm will stop iterating when the change in
centroids is small.
Labelling a new sample is performed by finding the nearest centroid for a given sample.
Examples:
• A demo of the mean-shift clustering algorithm: Mean Shift clustering on a synthetic 2D datasets with 3
classes.
References:
• “Mean shift: A robust approach toward feature space analysis.” D. Comaniciu and P. Meer, IEEE Transactions
on Pattern Analysis and Machine Intelligence (2002)

Spectral clustering
SpectralClustering does a low-dimension embedding of the affinity matrix between samples, followed by a
KMeans in the low dimensional space. It is especially efficient if the affinity matrix is sparse and the pyamg module
is installed. SpectralClustering requires the number of clusters to be specified. It works well for a small number of
clusters but is not advised when using many clusters.
For two clusters, it solves a convex relaxation of the normalised cuts problem on the similarity graph: cutting the
graph in two so that the weight of the edges cut is small compared to the weights of the edges inside each cluster. This
criteria is especially interesting when working on images: graph vertices are pixels, and edges of the similarity graph
are a function of the gradient of the image.
Warning: Transforming distance to well-behaved similarities

Note that if the values of your similarity matrix are not well distributed, e.g. with negative values or with a distance
matrix rather than a similarity, the spectral problem will be singular and the problem not solvable. In which case
it is advised to apply a transformation to the entries of the matrix. For instance, in the case of a signed distance
matrix, is common to apply a heat kernel:
similarity = np.exp(-beta * distance / distance.std())
See the examples for such an application.

Examples:
• Spectral clustering for image segmentation: Segmenting objects from a noisy background using spectral
clustering.
• Segmenting the picture of greek coins in regions: Spectral clustering to split the image of coins in regions.
Different label assignment strategies
Different label assignment strategies can be used, corresponding to the assign_labels parameter of
SpectralClustering. The "kmeans" strategy can match finer details of the data, but it can be more unsta-
ble. In particular, unless you control the random_state, it may not be reproducible from run-to-run, as it depends
on a random initialization. On the other hand, the "discretize" strategy is 100% reproducible, but it tends to
create parcels of fairly even and geometrical shape.
assign_labels="kmeans" assign_labels="discretize"
Spectral Clustering Graphs
Spectral Clustering can also be used to cluster graphs by their spectral embeddings. In this case, the affinity matrix is
the adjacency matrix of the graph, and SpectralClustering is initialized with affinity=’precomputed’:
>>> from sklearn.cluster import SpectralClustering

>>> sc = SpectralClustering(3, affinity='precomputed', n_init=100,
... assign_labels='discretize')
>>> sc.fit_predict(adjacency_matrix)
References:
• “A Tutorial on Spectral Clustering” Ulrike von Luxburg, 2007

• “Normalized cuts and image segmentation” Jianbo Shi, Jitendra Malik, 2000
• “A Random Walks View of Spectral Segmentation” Marina Meila, Jianbo Shi, 2001
• “On Spectral Clustering: Analysis and an algorithm” Andrew Y. Ng, Michael I. Jordan, Yair Weiss, 2001
Hierarchical clustering
Hierarchical clustering is a general family of clustering algorithms that build nested clusters by merging or splitting
them successively. This hierarchy of clusters is represented as a tree (or dendrogram). The root of the tree is the unique
cluster that gathers all the samples, the leaves being the clusters with only one sample. See the Wikipedia page for
more details.
The AgglomerativeClustering object performs a hierarchical clustering using a bottom up approach: each
observation starts in its own cluster, and clusters are successively merged together. The linkage criteria determines the
metric used for the merge strategy:
• Ward minimizes the sum of squared differences within all clusters. It is a variance-minimizing approach and in
this sense is similar to the k-means objective function but tackled with an agglomerative hierarchical approach.
• Maximum or complete linkage minimizes the maximum distance between observations of pairs of clusters.
• Average linkage minimizes the average of the distances between all observations of pairs of clusters.
• Single linkage minimizes the distance between the closest observations of pairs of clusters.
AgglomerativeClustering can also scale to large number of samples when it is used jointly with a connectivity
matrix, but is computationally expensive when no connectivity constraints are added between samples: it considers at
each step all the possible merges.
FeatureAgglomeration
The FeatureAgglomeration uses agglomerative clustering to group together features that look very similar,
thus decreasing the number of features. It is a dimensionality reduction tool, see Unsupervised dimensionality
reduction.
Different linkage type: Ward, complete, average, and single linkage
AgglomerativeClustering supports Ward, single, average, and complete linkage strategies.

Ag-
glomerative cluster has a “rich get richer” behavior that leads to uneven cluster sizes. In this regard, single linkage is
the worst strategy, and Ward gives the most regular sizes. However, the affinity (or distance used in clustering) cannot
be varied with Ward, thus for non Euclidean metrics, average linkage is a good alternative. Single linkage, while not
robust to noisy data, can be computed very efficiently and can therefore be useful to provide hierarchical clustering of
larger datasets. Single linkage can also perform well on non-globular data.
Examples:
• Various Agglomerative Clustering on a 2D embedding of digits: exploration of the different linkage strategies
in a real dataset.

Adding connectivity constraints
An interesting aspect of AgglomerativeClustering is that connectivity constraints can be added to this al-
gorithm (only adjacent clusters can be merged together), through a connectivity matrix that defines for each sample
the neighboring samples following a given structure of the data. For instance, in the swiss-roll example below, the
connectivity constraints forbid the merging of points that are not adjacent on the swiss roll, and thus avoid forming
clusters that extend across overlapping folds of the roll.
These constraint are useful to impose a certain local structure, but they also make the algorithm faster, especially when
the number of the samples is high.
The connectivity constraints are imposed via an connectivity matrix: a scipy sparse matrix that has elements only
at the intersection of a row and a column with indices of the dataset that should be connected. This matrix can
be constructed from a-priori information: for instance, you may wish to cluster web pages by only merging pages
with a link pointing from one to another. It can also be learned from the data, for instance using sklearn.
neighbors.kneighbors_graph to restrict merging to nearest neighbors as in this example, or using sklearn.
feature_extraction.image.grid_to_graph to enable only merging of neighboring pixels on an image,
as in the coin example.
Examples:
• A demo of structured Ward hierarchical clustering on an image of coins: Ward clustering to split the image
of coins in regions.
• Hierarchical clustering: structured vs unstructured ward: Example of Ward algorithm on a swiss-roll, com-
parison of structured approaches versus unstructured approaches.
• Feature agglomeration vs. univariate selection: Example of dimensionality reduction with feature agglomer-
ation based on Ward hierarchical clustering.
• Agglomerative clustering with and without structure
Warning: Connectivity constraints with single, average and complete linkage

Connectivity constraints and single, complete or average linkage can enhance the ‘rich getting richer’ aspect of
agglomerative clustering, particularly so if they are built with sklearn.neighbors.kneighbors_graph.
In the limit of a small number of clusters, they tend to give a few macroscopically occupied clusters and almost
empty ones. (see the discussion in Agglomerative clustering with and without structure). Single linkage is the
most brittle linkage option with regard to this issue.

Varying the metric
Single, average and complete linkage can be used with a variety of distances (or affinities), in particular Euclidean
distance (l2), Manhattan distance (or Cityblock, or l1), cosine distance, or any precomputed affinity matrix.
• l1 distance is often good for sparse features, or sparse noise: i.e. many of the features are zero, as in text mining
using occurrences of rare words.
• cosine distance is interesting because it is invariant to global scalings of the signal.
The guidelines for choosing a metric is to use one that maximizes the dis-
tance between samples in different classes, and minimizes that within each class.
Examples:
• Agglomerative clustering with different metrics
DBSCAN
The DBSCAN algorithm views clusters as areas of high density separated by areas of low density. Due to this rather
generic view, clusters found by DBSCAN can be any shape, as opposed to k-means which assumes that clusters are
convex shaped. The central component to the DBSCAN is the concept of core samples, which are samples that are in
areas of high density. A cluster is therefore a set of core samples, each close to each other (measured by some distance
measure) and a set of non-core samples that are close to a core sample (but are not themselves core samples). There
are two parameters to the algorithm, min_samples and eps, which define formally what we mean when we say
dense. Higher min_samples or lower eps indicate higher density necessary to form a cluster.
More formally, we define a core sample as being a sample in the dataset such that there exist min_samples other
samples within a distance of eps, which are defined as neighbors of the core sample. This tells us that the core sample
is in a dense area of the vector space. A cluster is a set of core samples that can be built by recursively taking a core
sample, finding all of its neighbors that are core samples, finding all of their neighbors that are core samples, and so
on. A cluster also has a set of non-core samples, which are samples that are neighbors of a core sample in the cluster
but are not themselves core samples. Intuitively, these samples are on the fringes of a cluster.

Any core sample is part of a cluster, by definition. Any sample that is not a core sample, and is at least eps in distance
from any core sample, is considered an outlier by the algorithm.
In the figure below, the color indicates cluster membership, with large circles indicating core samples found by the
algorithm. Smaller circles are non-core samples that are still part of a cluster. Moreover, the outliers are indicated by
black points below.
Examples:
• Demo of DBSCAN clustering algorithm
Implementation
The DBSCAN algorithm is deterministic, always generating the same clusters when given the same data in the
same order. However, the results can differ when data is provided in a different order. First, even though the core
samples will always be assigned to the same clusters, the labels of those clusters will depend on the order in which
those samples are encountered in the data. Second and more importantly, the clusters to which non-core samples
are assigned can differ depending on the data order. This would happen when a non-core sample has a distance
lower than eps to two core samples in different clusters. By the triangular inequality, those two core samples must
be more distant than eps from each other, or they would be in the same cluster. The non-core sample is assigned
to whichever cluster is generated first in a pass through the data, and so the results will depend on the data ordering.
The current implementation uses ball trees and kd-trees to determine the neighborhood of points, which avoids
calculating the full distance matrix (as was done in scikit-learn versions before 0.14). The possibility to use custom
metrics is retained; for details, see NearestNeighbors.
Memory consumption for large sample sizes
This implementation is by default not memory efficient because it constructs a full pairwise similarity matrix in the
case where kd-trees or ball-trees cannot be used (e.g. with sparse matrices). This matrix will consume n^2 floats.
A couple of mechanisms for getting around this are:
• A sparse radius neighborhood graph (where missing entries are presumed to be out of eps) can be precom-
puted in a memory-efficient way and dbscan can be run over this with metric='precomputed'. See
sklearn.neighbors.NearestNeighbors.radius_neighbors_graph.

• The dataset can be compressed, either by removing exact duplicates if these occur in your data, or by using
BIRCH. Then you only have a relatively small number of representatives for a large number of points. You
can then provide a sample_weight when fitting DBSCAN.
References:
• “A Density-Based Algorithm for Discovering Clusters in Large Spatial Databases with Noise” Ester, M., H. P.
Kriegel, J. Sander, and X. Xu, In Proceedings of the 2nd International Conference on Knowledge Discovery
and Data Mining, Portland, OR, AAAI Press, pp. 226–231. 1996
Birch
The Birch builds a tree called the Characteristic Feature Tree (CFT) for the given data. The data is essentially lossy
compressed to a set of Characteristic Feature nodes (CF Nodes). The CF Nodes have a number of subclusters called
Characteristic Feature subclusters (CF Subclusters) and these CF Subclusters located in the non-terminal CF Nodes
can have CF Nodes as children.
The CF Subclusters hold the necessary information for clustering which prevents the need to hold the entire input data
in memory. This information includes:
• Number of samples in a subcluster.
• Linear Sum - A n-dimensional vector holding the sum of all samples
• Squared Sum - Sum of the squared L2 norm of all samples.
• Centroids - To avoid recalculation linear sum / n_samples.
• Squared norm of the centroids.
The Birch algorithm has two parameters, the threshold and the branching factor. The branching factor limits the
number of subclusters in a node and the threshold limits the distance between the entering sample and the existing
subclusters.
This algorithm can be viewed as an instance or data reduction method, since it reduces the input data to a set of
subclusters which are obtained directly from the leaves of the CFT. This reduced data can be further processed by
feeding it into a global clusterer. This global clusterer can be set by n_clusters. If n_clusters is set to None,
the subclusters from the leaves are directly read off, otherwise a global clustering step labels these subclusters into
global clusters (labels) and the samples are mapped to the global label of the nearest subcluster.
Algorithm description:
• A new sample is inserted into the root of the CF Tree which is a CF Node. It is then merged with the subcluster of
the root, that has the smallest radius after merging, constrained by the threshold and branching factor conditions.
If the subcluster has any child node, then this is done repeatedly till it reaches a leaf. After finding the nearest
subcluster in the leaf, the properties of this subcluster and the parent subclusters are recursively updated.
• If the radius of the subcluster obtained by merging the new sample and the nearest subcluster is greater than
the square of the threshold and if the number of subclusters is greater than the branching factor, then a space is
temporarily allocated to this new sample. The two farthest subclusters are taken and the subclusters are divided
into two groups on the basis of the distance between these subclusters.
• If this split node has a parent subcluster and there is room for a new subcluster, then the parent is split into two.
If there is no room, then this node is again split into two and the process is continued recursively, till it reaches
the root.
Birch or MiniBatchKMeans?

• Birch does not scale very well to high dimensional data. As a rule of thumb if n_features is greater than
twenty, it is generally better to use MiniBatchKMeans.
• If the number of instances of data needs to be reduced, or if one wants a large number of subclusters either as a
preprocessing step or otherwise, Birch is more useful than MiniBatchKMeans.
How to use partial_fit?
To avoid the computation of global clustering, for every call of partial_fit the user is advised
1. To set n_clusters=None initially
2. Train all data by multiple calls to partial_fit.
3. Set n_clusters to a required value using brc.set_params(n_clusters=n_clusters).
4. Call partial_fit finally with no arguments, i.e. brc.partial_fit() which performs the global clus-
tering.
References:
• Tian Zhang, Raghu Ramakrishnan, Maron Livny BIRCH: An efficient data clustering method for large
databases. http://www.cs.sfu.ca/CourseCentral/459/han/papers/zhang96.pdf
• Roberto Perdisci JBirch - Java implementation of BIRCH clustering algorithm https://code.google.com/
archive/p/jbirch
Clustering performance evaluation
Evaluating the performance of a clustering algorithm is not as trivial as counting the number of errors or the precision
and recall of a supervised classification algorithm. In particular any evaluation metric should not take the absolute
values of the cluster labels into account but rather if this clustering define separations of the data similar to some
ground truth set of classes or satisfying some assumption such that members belong to the same class are more similar
that members of different classes according to some similarity metric.
Adjusted Rand index
Given the knowledge of the ground truth class assignments labels_true and our clustering algorithm assignments
of the same samples labels_pred, the adjusted Rand index is a function that measures the similarity of the two
assignments, ignoring permutations and with chance normalization:


>>> labels_true = [0, 0, 0, 1, 1, 1]
>>> labels_pred = [0, 0, 1, 1, 2, 2]
>>> metrics.adjusted_rand_score(labels_true, labels_pred)

0.24...
One can permute 0 and 1 in the predicted labels, rename 2 to 3, and get the same score:
>>> labels_pred = [1, 1, 0, 0, 3, 3]

0.24...
Furthermore, adjusted_rand_score is symmetric: swapping the argument does not change the score. It can
thus be used as a consensus measure:
>>> metrics.adjusted_rand_score(labels_pred, labels_true)

0.24...
Perfect labeling is scored 1.0:
>>> labels_pred = labels_true[:]

1.0
Bad (e.g. independent labelings) have negative or close to 0.0 scores:
>>> labels_true = [0, 1, 2, 0, 3, 4, 5, 1]

>>> labels_pred = [1, 1, 0, 0, 2, 2, 2, 2]
-0.12...
Advantages
• Random (uniform) label assignments have a ARI score close to 0.0 for any value of n_clusters and
n_samples (which is not the case for raw Rand index or the V-measure for instance).
• Bounded range [-1, 1]: negative values are bad (independent labelings), similar clusterings have a positive ARI,
1.0 is the perfect match score.
• No assumption is made on the cluster structure: can be used to compare clustering algorithms such as k-
means which assumes isotropic blob shapes with results of spectral clustering algorithms which can find cluster
with “folded” shapes.
Drawbacks
• Contrary to inertia, ARI requires knowledge of the ground truth classes while is almost never available in
practice or requires manual assignment by human annotators (as in the supervised learning setting).
However ARI can also be useful in a purely unsupervised setting as a building block for a Consensus Index that
can be used for clustering model selection (TODO).

Examples:
• Adjustment for chance in clustering performance evaluation: Analysis of the impact of the dataset size on the
value of clustering measures for random assignments.
If C is a ground truth class assignment and K the clustering, let us define 𝑎 and 𝑏 as:
• 𝑎, the number of pairs of elements that are in the same set in C and in the same set in K
• 𝑏, the number of pairs of elements that are in different sets in C and in different sets in K
The raw (unadjusted) Rand index is then given by:
𝑎+𝑏
RI = 𝑛
𝐶2 𝑠𝑎𝑚𝑝𝑙𝑒𝑠
𝑛
Where 𝐶2 𝑠𝑎𝑚𝑝𝑙𝑒𝑠 is the total number of possible pairs in the dataset (without ordering).
However the RI score does not guarantee that random label assignments will get a value close to zero (esp. if the
number of clusters is in the same order of magnitude as the number of samples).
To counter this effect we can discount the expected RI 𝐸[RI] of random labelings by defining the adjusted Rand index
as follows:
RI − 𝐸[RI]
ARI =
max(RI) − 𝐸[RI]
References
• Comparing Partitions L. Hubert and P. Arabie, Journal of Classification 1985

• Wikipedia entry for the adjusted Rand index
Mutual Information based scores
Given the knowledge of the ground truth class assignments labels_true and our clustering algorithm assignments
of the same samples labels_pred, the Mutual Information is a function that measures the agreement of the two
assignments, ignoring permutations. Two different normalized versions of this measure are available, Normalized
Mutual Information (NMI) and Adjusted Mutual Information (AMI). NMI is often used in the literature, while
AMI was proposed more recently and is normalized against chance:
>>> labels_true = [0, 0, 0, 1, 1, 1]
>>> labels_pred = [0, 0, 1, 1, 2, 2]
>>> metrics.adjusted_mutual_info_score(labels_true, labels_pred)

0.22504...
One can permute 0 and 1 in the predicted labels, rename 2 to 3 and get the same score:
>>> labels_pred = [1, 1, 0, 0, 3, 3]
0.22504...

All, mutual_info_score, adjusted_mutual_info_score and normalized_mutual_info_score

are symmetric: swapping the argument does not change the score. Thus they can be used as a consensus measure:
>>> metrics.adjusted_mutual_info_score(labels_pred, labels_true)

0.22504...

1.0
>>> metrics.normalized_mutual_info_score(labels_true, labels_pred)

1.0
This is not true for mutual_info_score, which is therefore harder to judge:
>>> metrics.mutual_info_score(labels_true, labels_pred)

0.69...
Bad (e.g. independent labelings) have non-positive scores:
>>> labels_true = [0, 1, 2, 0, 3, 4, 5, 1]

>>> labels_pred = [1, 1, 0, 0, 2, 2, 2, 2]
-0.10526...
Advantages
• Random (uniform) label assignments have a AMI score close to 0.0 for any value of n_clusters and
n_samples (which is not the case for raw Mutual Information or the V-measure for instance).
• Upper bound of 1: Values close to zero indicate two label assignments that are largely independent, while
values close to one indicate significant agreement. Further, an AMI of exactly 1 indicates that the two label
assignments are equal (with or without permutation).
Drawbacks
• Contrary to inertia, MI-based measures require the knowledge of the ground truth classes while almost
never available in practice or requires manual assignment by human annotators (as in the supervised learning
setting).
However MI-based measures can also be useful in purely unsupervised setting as a building block for a Consen-
sus Index that can be used for clustering model selection.
• NMI and MI are not adjusted against chance.
Examples:
value of clustering measures for random assignments. This example also includes the Adjusted Rand Index.

Assume two label assignments (of the same N objects), 𝑈 and 𝑉 . Their entropy is the amount of uncertainty for a
partition set, defined by:
|𝑈 |
∑︁
𝐻(𝑈 ) = − 𝑃 (𝑖) log(𝑃 (𝑖))
𝑖=1
where 𝑃 (𝑖) = |𝑈𝑖 |/𝑁 is the probability that an object picked at random from 𝑈 falls into class 𝑈𝑖 . Likewise for 𝑉 :
|𝑉 |
∑︁
𝐻(𝑉 ) = − 𝑃 ′ (𝑗) log(𝑃 ′ (𝑗))
𝑗=1
With 𝑃 ′ (𝑗) = |𝑉𝑗 |/𝑁 . The mutual information (MI) between 𝑈 and 𝑉 is calculated by:
|𝑈 | |𝑉 | (︂ )︂
∑︁ ∑︁ 𝑃 (𝑖, 𝑗)
MI(𝑈, 𝑉 ) = 𝑃 (𝑖, 𝑗) log
𝑖=1 𝑗=1
𝑃 (𝑖)𝑃 ′ (𝑗)
where 𝑃 (𝑖, 𝑗) = |𝑈𝑖 ∩ 𝑉𝑗 |/𝑁 is the probability that an object picked at random falls into both classes 𝑈𝑖 and 𝑉𝑗 .
It also can be expressed in set cardinality formulation:
|𝑈 | |𝑉 | (︂ )︂
∑︁ ∑︁ |𝑈𝑖 ∩ 𝑉𝑗 | 𝑁 |𝑈𝑖 ∩ 𝑉𝑗 |
MI(𝑈, 𝑉 ) = log
𝑖=1 𝑗=1
𝑁 |𝑈𝑖 ||𝑉𝑗 |
The normalized mutual information is defined as

MI(𝑈, 𝑉 )
NMI(𝑈, 𝑉 ) =
mean(𝐻(𝑈 ), 𝐻(𝑉 ))
This value of the mutual information and also the normalized variant is not adjusted for chance and will tend to increase
as the number of different labels (clusters) increases, regardless of the actual amount of “mutual information” between
the label assignments.
The expected value for the mutual information can be calculated using the following equation [VEB2009]. In this
equation, 𝑎𝑖 = |𝑈𝑖 | (the number of elements in 𝑈𝑖 ) and 𝑏𝑗 = |𝑉𝑗 | (the number of elements in 𝑉𝑗 ).
|𝑈 | |𝑉 | min(𝑎𝑖 ,𝑏𝑗 ) (︂ )︂
∑︁ ∑︁ ∑︁ 𝑛𝑖𝑗 𝑁.𝑛𝑖𝑗 𝑎𝑖 !𝑏𝑗 !(𝑁 − 𝑎𝑖 )!(𝑁 − 𝑏𝑗 )!
𝐸[MI(𝑈, 𝑉 )] = log
𝑖=1 𝑗=1 𝑛𝑖𝑗 =(𝑎𝑖 +𝑏𝑗 −𝑁 )+
𝑁 𝑎𝑖 𝑏𝑗 𝑁 !𝑛𝑖𝑗 !(𝑎𝑖 − 𝑛𝑖𝑗 )!(𝑏𝑗 − 𝑛𝑖𝑗 )!(𝑁 − 𝑎𝑖 − 𝑏𝑗 + 𝑛𝑖𝑗 )!
Using the expected value, the adjusted mutual information can then be calculated using a similar form to that of the
adjusted Rand index:
MI − 𝐸[MI]
AMI =
mean(𝐻(𝑈 ), 𝐻(𝑉 )) − 𝐸[MI]
For normalized mutual information and adjusted mutual information, the normalizing value is typically some gener-
alized mean of the entropies of each clustering. Various generalized means exist, and no firm rules exist for preferring
one over the others. The decision is largely a field-by-field basis; for instance, in community detection, the arithmetic
mean is most common. Each normalizing method provides “qualitatively similar behaviours” [YAT2016]. In our
implementation, this is controlled by the average_method parameter.
Vinh et al. (2010) named variants of NMI and AMI by their averaging method [VEB2010]. Their ‘sqrt’ and ‘sum’
averages are the geometric and arithmetic means; we use these more broadly common names.

References
• Strehl, Alexander, and Joydeep Ghosh (2002). “Cluster ensembles – a knowledge reuse frame-
work for combining multiple partitions”. Journal of Machine Learning Research 3: 583–617.
doi:10.1162/153244303321897735.
• [VEB2009] Vinh, Epps, and Bailey, (2009). “Information theoretic measures for clusterings compar-
ison”. Proceedings of the 26th Annual International Conference on Machine Learning - ICML ‘09.
doi:10.1145/1553374.1553511. ISBN 9781605585161.
• [VEB2010] Vinh, Epps, and Bailey, (2010). “Information Theoretic Measures for Clusterings Comparison:
Variants, Properties, Normalization and Correction for Chance”. JMLR <http://jmlr.csail.mit.edu/papers/
volume11/vinh10a/vinh10a.pdf>
• Wikipedia entry for the (normalized) Mutual Information
• Wikipedia entry for the Adjusted Mutual Information
• [YAT2016] Yang, Algesheimer, and Tessone, (2016). “A comparative analysis of community detection algo-
rithms on artificial networks”. Scientific Reports 6: 30750. doi:10.1038/srep30750.
Homogeneity, completeness and V-measure
Given the knowledge of the ground truth class assignments of the samples, it is possible to define some intuitive metric
using conditional entropy analysis.
In particular Rosenberg and Hirschberg (2007) define the following two desirable objectives for any cluster assign-
ment:
• homogeneity: each cluster contains only members of a single class.
• completeness: all members of a given class are assigned to the same cluster.
We can turn those concept as scores homogeneity_score and completeness_score. Both are bounded
below by 0.0 and above by 1.0 (higher is better):

>>> labels_true = [0, 0, 0, 1, 1, 1]
>>> labels_pred = [0, 0, 1, 1, 2, 2]
>>> metrics.homogeneity_score(labels_true, labels_pred)

0.66...
>>> metrics.completeness_score(labels_true, labels_pred)

0.42...
Their harmonic mean called V-measure is computed by v_measure_score:
>>> metrics.v_measure_score(labels_true, labels_pred)

0.51...
The V-measure is actually equivalent to the mutual information (NMI) discussed above, with the aggregation function
being the arithmetic mean [B2011].
Homogeneity, completeness and V-measure can be computed at once using
homogeneity_completeness_v_measure as follows:

>>> metrics.homogeneity_completeness_v_measure(labels_true, labels_pred)

...
(0.66..., 0.42..., 0.51...)
The following clustering assignment is slightly better, since it is homogeneous but not complete:
>>> labels_pred = [0, 0, 0, 1, 2, 2]

>>> metrics.homogeneity_completeness_v_measure(labels_true, labels_pred)
...
(1.0, 0.68..., 0.81...)
Note: v_measure_score is symmetric: it can be used to evaluate the agreement of two independent assignments
on the same dataset.
This is not the case for completeness_score and homogeneity_score: both are bound by the relationship:
homogeneity_score(a, b) == completeness_score(b, a)
Advantages
• Bounded scores: 0.0 is as bad as it can be, 1.0 is a perfect score.

• Intuitive interpretation: clustering with bad V-measure can be qualitatively analyzed in terms of homogeneity
and completeness to better feel what ‘kind’ of mistakes is done by the assignment.
Drawbacks
• The previously introduced metrics are not normalized with regards to random labeling: this means that
depending on the number of samples, clusters and ground truth classes, a completely random labeling will
not always yield the same values for homogeneity, completeness and hence v-measure. In particular random
labeling won’t yield zero scores especially when the number of clusters is large.
This problem can safely be ignored when the number of samples is more than a thousand and the number of
clusters is less than 10. For smaller sample sizes or larger number of clusters it is safer to use an adjusted
index such as the Adjusted Rand Index (ARI).
• These metrics require the knowledge of the ground truth classes while almost never available in practice or
requires manual assignment by human annotators (as in the supervised learning setting).
Examples:
value of clustering measures for random assignments.


Homogeneity and completeness scores are formally given by:

𝐻(𝐶|𝐾)
ℎ=1−
𝐻(𝐶)
𝐻(𝐾|𝐶)
𝑐=1−
𝐻(𝐾)
where 𝐻(𝐶|𝐾) is the conditional entropy of the classes given the cluster assignments and is given by:
|𝐶| |𝐾| (︂ )︂
∑︁ ∑︁ 𝑛𝑐,𝑘 𝑛𝑐,𝑘
𝐻(𝐶|𝐾) = − · log
𝑐=1 𝑘=1
𝑛 𝑛𝑘
and 𝐻(𝐶) is the entropy of the classes and is given by:

|𝐶|
∑︁ 𝑛𝑐 (︁ 𝑛 )︁
𝑐
𝐻(𝐶) = − · log
𝑐=1
𝑛 𝑛
with 𝑛 the total number of samples, 𝑛𝑐 and 𝑛𝑘 the number of samples respectively belonging to class 𝑐 and cluster 𝑘,
and finally 𝑛𝑐,𝑘 the number of samples from class 𝑐 assigned to cluster 𝑘.
The conditional entropy of clusters given class 𝐻(𝐾|𝐶) and the entropy of clusters 𝐻(𝐾) are defined in a sym-
metric manner.
Rosenberg and Hirschberg further define V-measure as the harmonic mean of homogeneity and completeness:
ℎ·𝑐
𝑣 =2·
ℎ+𝑐
References
• V-Measure: A conditional entropy-based external cluster evaluation measure Andrew Rosenberg and Julia
Hirschberg, 2007
Fowlkes-Mallows scores
The Fowlkes-Mallows index (sklearn.metrics.fowlkes_mallows_score) can be used when the ground

truth class assignments of the samples is known. The Fowlkes-Mallows score FMI is defined as the geometric mean
of the pairwise precision and recall:
TP
FMI = √︀
(TP + FP)(TP + FN)
Where TP is the number of True Positive (i.e. the number of pair of points that belong to the same clusters in both the
true labels and the predicted labels), FP is the number of False Positive (i.e. the number of pair of points that belong
to the same clusters in the true labels and not in the predicted labels) and FN is the number of False Negative (i.e the
number of pair of points that belongs in the same clusters in the predicted labels and not in the true labels).
The score ranges from 0 to 1. A high value indicates a good similarity between two clusters.
>>> labels_true = [0, 0, 0, 1, 1, 1]
>>> labels_pred = [0, 0, 1, 1, 2, 2]

>>> metrics.fowlkes_mallows_score(labels_true, labels_pred)

0.47140...
One can permute 0 and 1 in the predicted labels, rename 2 to 3 and get the same score:
>>> labels_pred = [1, 1, 0, 0, 3, 3]

0.47140...

1.0
Bad (e.g. independent labelings) have zero scores:
>>> labels_true = [0, 1, 2, 0, 3, 4, 5, 1]

>>> labels_pred = [1, 1, 0, 0, 2, 2, 2, 2]
0.0
Advantages
• Random (uniform) label assignments have a FMI score close to 0.0 for any value of n_clusters and
n_samples (which is not the case for raw Mutual Information or the V-measure for instance).
• Upper-bounded at 1: Values close to zero indicate two label assignments that are largely independent, while
values close to one indicate significant agreement. Further, values of exactly 0 indicate purely independent
label assignments and a FMI of exactly 1 indicates that the two label assignments are equal (with or without
permutation).
Drawbacks
• Contrary to inertia, FMI-based measures require the knowledge of the ground truth classes while almost
never available in practice or requires manual assignment by human annotators (as in the supervised learning
setting).
References
• E. B. Fowkles and C. L. Mallows, 1983. “A method for comparing two hierarchical clusterings”. Journal of
the American Statistical Association. http://wildfire.stat.ucla.edu/pdflibrary/fowlkes.pdf
• Wikipedia entry for the Fowlkes-Mallows Index

Silhouette Coefficient
If the ground truth labels are not known, evaluation must be performed using the model itself. The Silhouette Coeffi-
cient (sklearn.metrics.silhouette_score) is an example of such an evaluation, where a higher Silhouette
Coefficient score relates to a model with better defined clusters. The Silhouette Coefficient is defined for each sample
and is composed of two scores:
• a: The mean distance between a sample and all other points in the same class.
• b: The mean distance between a sample and all other points in the next nearest cluster.
The Silhouette Coefficient s for a single sample is then given as:
𝑏−𝑎
𝑠=
𝑚𝑎𝑥(𝑎, 𝑏)
The Silhouette Coefficient for a set of samples is given as the mean of the Silhouette Coefficient for each sample.

>>> from sklearn.metrics import pairwise_distances
>>> dataset = datasets.load_iris()
>>> X = dataset.data
>>> y = dataset.target
In normal usage, the Silhouette Coefficient is applied to the results of a cluster analysis.

>>> from sklearn.cluster import KMeans
>>> kmeans_model = KMeans(n_clusters=3, random_state=1).fit(X)
>>> labels = kmeans_model.labels_
>>> metrics.silhouette_score(X, labels, metric='euclidean')
...
0.55...
References
• Peter J. Rousseeuw (1987). “Silhouettes: a Graphical Aid to the Interpretation and Validation of Cluster
Analysis”. Computational and Applied Mathematics 20: 53–65. doi:10.1016/0377-0427(87)90125-7.
Advantages
• The score is bounded between -1 for incorrect clustering and +1 for highly dense clustering. Scores around zero
indicate overlapping clusters.
• The score is higher when clusters are dense and well separated, which relates to a standard concept of a cluster.
Drawbacks
• The Silhouette Coefficient is generally higher for convex clusters than other concepts of clusters, such as density
based clusters like those obtained through DBSCAN.

Examples:
• Selecting the number of clusters with silhouette analysis on KMeans clustering : In this example the silhouette
analysis is used to choose an optimal value for n_clusters.
Calinski-Harabaz Index
If the ground truth labels are not known, the Calinski-Harabaz index (sklearn.metrics.
calinski_harabaz_score) - also known as the Variance Ratio Criterion - can be used to evaluate the
model, where a higher Calinski-Harabaz score relates to a model with better defined clusters.
For 𝑘 clusters, the Calinski-Harabaz score 𝑠 is given as the ratio of the between-clusters dispersion mean and the
within-cluster dispersion:
Tr(𝐵𝑘 ) 𝑁 −𝑘
𝑠(𝑘) = ×
Tr(𝑊𝑘 ) 𝑘−1
where 𝐵𝐾 is the between group dispersion matrix and 𝑊𝐾 is the within-cluster dispersion matrix defined by:
𝑘 ∑︁
∑︁
𝑊𝑘 = (𝑥 − 𝑐𝑞 )(𝑥 − 𝑐𝑞 )𝑇
𝑞=1 𝑥∈𝐶𝑞
∑︁
𝐵𝑘 = 𝑛𝑞 (𝑐𝑞 − 𝑐)(𝑐𝑞 − 𝑐)𝑇
𝑞
with 𝑁 be the number of points in our data, 𝐶𝑞 be the set of points in cluster 𝑞, 𝑐𝑞 be the center of cluster 𝑞, 𝑐 be the
center of 𝐸, 𝑛𝑞 be the number of points in cluster 𝑞.

>>> dataset = datasets.load_iris()
>>> X = dataset.data
>>> y = dataset.target
In normal usage, the Calinski-Harabaz index is applied to the results of a cluster analysis.

>>> kmeans_model = KMeans(n_clusters=3, random_state=1).fit(X)
>>> labels = kmeans_model.labels_
>>> metrics.calinski_harabaz_score(X, labels)
561.62...
Advantages
• The score is higher when clusters are dense and well separated, which relates to a standard concept of a cluster.
• The score is fast to compute

Drawbacks
• The Calinski-Harabaz index is generally higher for convex clusters than other concepts of clusters, such as
density based clusters like those obtained through DBSCAN.
References
• Caliński, T., & Harabasz, J. (1974). “A dendrite method for cluster analysis”. Communications in Statistics-
theory and Methods 3: 1-27. doi:10.1080/03610926.2011.560741.
Davies-Bouldin Index
If the ground truth labels are not known, the Davies-Bouldin index (sklearn.metrics.
davies_bouldin_score) can be used to evaluate the model, where a lower Davies-Bouldin index relates
to a model with better separation between the clusters.
The index is defined as the average similarity between each cluster 𝐶𝑖 for 𝑖 = 1, ..., 𝑘 and its most similar one 𝐶𝑗 . In
the context of this index, similarity is defined as a measure 𝑅𝑖𝑗 that trades off:
• 𝑠𝑖 , the average distance between each point of cluster 𝑖 and the centroid of that cluster – also know as cluster
diameter.
• 𝑑𝑖𝑗 , the distance between cluster centroids 𝑖 and 𝑗.
A simple choice to construct 𝑅𝑖 𝑗 so that it is nonnegative and symmetric is:
𝑠𝑖 + 𝑠𝑗
𝑅𝑖𝑗 =
𝑑𝑖𝑗
Then the Davies-Bouldin index is defined as:

𝑘
1 ∑︁
𝐷𝐵 = max 𝑅𝑖𝑗
𝑘 𝑖=1 𝑖̸=𝑗
Zero is the lowest possible score. Values closer to zero indicate a better partition.
In normal usage, the Davies-Bouldin index is applied to the results of a cluster analysis as follows:

>>> X = iris.data
>>> from sklearn.metrics import davies_bouldin_score
>>> kmeans = KMeans(n_clusters=3, random_state=1).fit(X)
>>> labels = kmeans.labels_
>>> davies_bouldin_score(X, labels)
0.6619...
Advantages
• The computation of Davies-Bouldin is simpler than that of Silhouette scores.

• The index is computed only quantities and features inherent to the dataset.

Drawbacks
• The Davies-Boulding index is generally higher for convex clusters than other concepts of clusters, such as
density based clusters like those obtained from DBSCAN.
• The usage of centroid distance limits the distance metric to Euclidean space.
• A good value reported by this method does not imply the best information retrieval.
References
• Davies, David L.; Bouldin, Donald W. (1979). “A Cluster Separation Measure” IEEE Transactions on Pattern
Analysis and Machine Intelligence. PAMI-1 (2): 224-227. doi:10.1109/TPAMI.1979.4766909.
• Halkidi, Maria; Batistakis, Yannis; Vazirgiannis, Michalis (2001). “On Clustering Validation Techniques”
Journal of Intelligent Information Systems, 17(2-3), 107-145. doi:10.1023/A:1012801612483.
• Wikipedia entry for Davies-Bouldin index.
Contingency Matrix
Contingency matrix (sklearn.metrics.cluster.contingency_matrix) reports the intersection cardi-

nality for every true/predicted cluster pair. The contingency matrix provides sufficient statistics for all clustering met-
rics where the samples are independent and identically distributed and one doesn’t need to account for some instances
not being clustered.
Here is an example:
>>> from sklearn.metrics.cluster import contingency_matrix

>>> x = ["a", "a", "a", "b", "b", "b"]
>>> y = [0, 0, 1, 1, 2, 2]
>>> contingency_matrix(x, y)
array([[2, 1, 0],
[0, 1, 2]])
The first row of output array indicates that there are three samples whose true cluster is “a”. Of them, two are in
predicted cluster 0, one is in 1, and none is in 2. And the second row indicates that there are three samples whose true
cluster is “b”. Of them, none is in predicted cluster 0, one is in 1 and two are in 2.
A confusion matrix for classification is a square contingency matrix where the order of rows and columns correspond
to a list of classes.
Advantages
• Allows to examine the spread of each true cluster across predicted clusters and vice versa.
• The contingency table calculated is typically utilized in the calculation of a similarity statistic (like the others
listed in this document) between the two clusterings.
Drawbacks
• Contingency matrix is easy to interpret for a small number of clusters, but becomes very hard to interpret for a
large number of clusters.

• It doesn’t give a single metric to use as an objective for clustering optimisation.
References
• Wikipedia entry for contingency matrix
3.2.4 Biclustering
Biclustering can be performed with the module sklearn.cluster.bicluster. Biclustering algorithms simul-
taneously cluster rows and columns of a data matrix. These clusters of rows and columns are known as biclusters.
Each determines a submatrix of the original data matrix with some desired properties.
For instance, given a matrix of shape (10, 10), one possible bicluster with three rows and two columns induces a
submatrix of shape (3, 2):
>>> data = np.arange(100).reshape(10, 10)
>>> rows = np.array([0, 2, 3])[:, np.newaxis]
>>> columns = np.array([1, 2])
>>> data[rows, columns]
array([[ 1, 2],
[21, 22],
[31, 32]])
For visualization purposes, given a bicluster, the rows and columns of the data matrix may be rearranged to make the
bicluster contiguous.
Algorithms differ in how they define biclusters. Some of the common types include:
• constant values, constant rows, or constant columns
• unusually high or low values
• submatrices with low variance
• correlated rows or columns
Algorithms also differ in how rows and columns may be assigned to biclusters, which leads to different bicluster
structures. Block diagonal or checkerboard structures occur when rows and columns are divided into partitions.
If each row and each column belongs to exactly one bicluster, then rearranging the rows and columns of the data matrix
reveals the biclusters on the diagonal. Here is an example of this structure where biclusters have higher average values
than the other rows and columns:
In the checkerboard case, each row belongs to all column clusters, and each column belongs to all row clusters. Here
is an example of this structure where the variance of the values within each bicluster is small:
After fitting a model, row and column cluster membership can be found in the rows_ and columns_ attributes.
rows_[i] is a binary vector with nonzero entries corresponding to rows that belong to bicluster i. Similarly,
columns_[i] indicates which columns belong to bicluster i.
Some models also have row_labels_ and column_labels_ attributes. These models partition the rows and
columns, such as in the block diagonal and checkerboard bicluster structures.
Note: Biclustering has many other names in different fields including co-clustering, two-mode clustering, two-way
clustering, block clustering, coupled two-way clustering, etc. The names of some algorithms, such as the Spectral
Co-Clustering algorithm, reflect these alternate names.

Fig. 3.5: An example of biclusters formed by partitioning rows and columns.
Fig. 3.6: An example of checkerboard biclusters.

Spectral Co-Clustering
The SpectralCoclustering algorithm finds biclusters with values higher than those in the corresponding other
rows and columns. Each row and each column belongs to exactly one bicluster, so rearranging the rows and columns
to make partitions contiguous reveals these high values along the diagonal:
Note: The algorithm treats the input data matrix as a bipartite graph: the rows and columns of the matrix correspond
to the two sets of vertices, and each entry corresponds to an edge between a row and a column. The algorithm
approximates the normalized cut of this graph to find heavy subgraphs.
An approximate solution to the optimal normalized cut may be found via the generalized eigenvalue decomposition of
the Laplacian of the graph. Usually this would mean working directly with the Laplacian matrix. If the original data
matrix 𝐴 has shape 𝑚 × 𝑛, the Laplacian matrix for the corresponding bipartite graph has shape (𝑚 + 𝑛) × (𝑚 + 𝑛).
However, in this case it is possible to work directly with 𝐴, which is smaller and more efficient.
The input matrix 𝐴 is preprocessed as follows:
𝐴𝑛 = 𝑅−1/2 𝐴𝐶 −1/2
∑︀
Where
∑︀ 𝑅 is the diagonal matrix with entry 𝑖 equal to 𝑗 𝐴𝑖𝑗 and 𝐶 is the diagonal matrix with entry 𝑗 equal to
𝑖 𝐴𝑖𝑗 .
The singular value decomposition, 𝐴𝑛 = 𝑈 Σ𝑉 ⊤ , provides the partitions of the rows and columns of 𝐴. A subset of
the left singular vectors gives the row partitions, and a subset of the right singular vectors gives the column partitions.
The ℓ = ⌈log2 𝑘⌉ singular vectors, starting from the second, provide the desired partitioning information. They are
used to form the matrix 𝑍:
⎡ −1/2 ⎤
𝑅 𝑈
𝑍=⎣ ⎦
−1/2
𝐶 𝑉
where the columns of 𝑈 are 𝑢2 , . . . , 𝑢ℓ+1 , and similarly for 𝑉 .

Then the rows of 𝑍 are clustered using k-means. The first n_rows labels provide the row partitioning, and the
remaining n_columns labels provide the column partitioning.
Examples:
• A demo of the Spectral Co-Clustering algorithm: A simple example showing how to generate a data matrix
with biclusters and apply this method to it.
• Biclustering documents with the Spectral Co-clustering algorithm: An example of finding biclusters in the
twenty newsgroup dataset.
References:
• Dhillon, Inderjit S, 2001. Co-clustering documents and words using bipartite spectral graph partitioning.

Spectral Biclustering
The SpectralBiclustering algorithm assumes that the input data matrix has a hidden checkerboard structure.
The rows and columns of a matrix with this structure may be partitioned so that the entries of any bicluster in the
Cartesian product of row clusters and column clusters are approximately constant. For instance, if there are two row
partitions and three column partitions, each row will belong to three biclusters, and each column will belong to two
biclusters.
The algorithm partitions the rows and columns of a matrix so that a corresponding blockwise-constant checkerboard
matrix provides a good approximation to the original matrix.
The input matrix 𝐴 is first normalized to make the checkerboard pattern more obvious. There are three possible
methods:
1. Independent row and column normalization, as in Spectral Co-Clustering. This method makes the rows sum to
a constant and the columns sum to a different constant.
2. Bistochastization: repeated row and column normalization until convergence. This method makes both rows
and columns sum to the same constant.
3. Log normalization: the log of the data matrix is computed: 𝐿 = log 𝐴. Then the column mean 𝐿𝑖· , row mean
𝐿·𝑗 , and overall mean 𝐿·· of 𝐿 are computed. The final matrix is computed according to the formula
𝐾𝑖𝑗 = 𝐿𝑖𝑗 − 𝐿𝑖· − 𝐿·𝑗 + 𝐿··
After normalizing, the first few singular vectors are computed, just as in the Spectral Co-Clustering algorithm.
If log normalization was used, all the singular vectors are meaningful. However, if independent normalization or
bistochastization were used, the first singular vectors, 𝑢1 and 𝑣1 . are discarded. From now on, the “first” singular
vectors refers to 𝑢2 . . . 𝑢𝑝+1 and 𝑣2 . . . 𝑣𝑝+1 except in the case of log normalization.
Given these singular vectors, they are ranked according to which can be best approximated by a piecewise-constant
vector. The approximations for each vector are found using one-dimensional k-means and scored using the Euclidean
distance. Some subset of the best left and right singular vector are selected. Next, the data is projected to this best
subset of singular vectors and clustered.
For instance, if 𝑝 singular vectors were calculated, the 𝑞 best are found as described, where 𝑞 < 𝑝. Let 𝑈 be the matrix
with columns the 𝑞 best left singular vectors, and similarly 𝑉 for the right. To partition the rows, the rows of 𝐴 are
projected to a 𝑞 dimensional space: 𝐴 * 𝑉 . Treating the 𝑚 rows of this 𝑚 × 𝑞 matrix as samples and clustering using
k-means yields the row labels. Similarly, projecting the columns to 𝐴⊤ * 𝑈 and clustering this 𝑛 × 𝑞 matrix yields the
column labels.
Examples:
• A demo of the Spectral Biclustering algorithm: a simple example showing how to generate a checkerboard
matrix and bicluster it.
References:
• Kluger, Yuval, et. al., 2003. Spectral biclustering of microarray data: coclustering genes and conditions.

Biclustering evaluation
There are two ways of evaluating a biclustering result: internal and external. Internal measures, such as cluster
stability, rely only on the data and the result themselves. Currently there are no internal bicluster measures in scikit-
learn. External measures refer to an external source of information, such as the true solution. When working with
real data the true solution is usually unknown, but biclustering artificial data may be useful for evaluating algorithms
precisely because the true solution is known.
To compare a set of found biclusters to the set of true biclusters, two similarity measures are needed: a similarity
measure for individual biclusters, and a way to combine these individual similarities into an overall score.
To compare individual biclusters, several measures have been used. For now, only the Jaccard index is implemented:
|𝐴 ∩ 𝐵|
𝐽(𝐴, 𝐵) =
|𝐴| + |𝐵| − |𝐴 ∩ 𝐵|
where 𝐴 and 𝐵 are biclusters, |𝐴 ∩ 𝐵| is the number of elements in their intersection. The Jaccard index achieves its
minimum of 0 when the biclusters to not overlap at all and its maximum of 1 when they are identical.
Several methods have been developed to compare two sets of biclusters. For now, only consensus_score (Hochre-
iter et. al., 2010) is available:
1. Compute bicluster similarities for pairs of biclusters, one in each set, using the Jaccard index or a similar
measure.
2. Assign biclusters from one set to another in a one-to-one fashion to maximize the sum of their similarities. This
step is performed using the Hungarian algorithm.
3. The final sum of similarities is divided by the size of the larger set.
The minimum consensus score, 0, occurs when all pairs of biclusters are totally dissimilar. The maximum score, 1,
occurs when both sets are identical.
References:
• Hochreiter, Bodenhofer, et. al., 2010. FABIA: factor analysis for bicluster acquisition.
3.2.5 Decomposing signals in components (matrix factorization problems)
Principal component analysis (PCA)
Exact PCA and probabilistic interpretation
PCA is used to decompose a multivariate dataset in a set of successive orthogonal components that explain a maximum
amount of the variance. In scikit-learn, PCA is implemented as a transformer object that learns 𝑛 components in its
fit method, and can be used on new data to project it on these components.
The optional parameter whiten=True makes it possible to project the data onto the singular space while scaling
each component to unit variance. This is often useful if the models down-stream make strong assumptions on the
isotropy of the signal: this is for example the case for Support Vector Machines with the RBF kernel and the K-Means
clustering algorithm.
Below is an example of the iris dataset, which is comprised of 4 features, projected on the 2 dimensions that explain
most variance:
The PCA object also provides a probabilistic interpretation of the PCA that can give a likelihood of data based on the
amount of variance it explains. As such it implements a score method that can be used in cross-validation:


Examples:
• Comparison of LDA and PCA 2D projection of Iris dataset

• Model selection with Probabilistic PCA and Factor Analysis (FA)
Incremental PCA
The PCA object is very useful, but has certain limitations for large datasets. The biggest limitation is that PCA only sup-
ports batch processing, which means all of the data to be processed must fit in main memory. The IncrementalPCA
object uses a different form of processing and allows for partial computations which almost exactly match the results
of PCA while processing the data in a minibatch fashion. IncrementalPCA makes it possible to implement out-of-
core Principal Component Analysis either by:
• Using its partial_fit method on chunks of data fetched sequentially from the local hard drive or a network
database.
• Calling its fit method on a memory mapped file using numpy.memmap.
IncrementalPCA only stores estimates of component and noise variances, in order update
explained_variance_ratio_ incrementally. This is why memory usage depends on the number of
samples per batch, rather than the number of samples to be processed in the dataset.
Examples:
• Incremental PCA
PCA using randomized SVD
It is often interesting to project data to a lower-dimensional space that preserves most of the variance, by dropping the
singular vector of components associated with lower singular values.
For instance, if we work with 64x64 pixel gray-level pictures for face recognition, the dimensionality of the data is
4096 and it is slow to train an RBF support vector machine on such wide data. Furthermore we know that the intrinsic
dimensionality of the data is much lower than 4096 since all pictures of human faces look somewhat alike. The
samples lie on a manifold of much lower dimension (say around 200 for instance). The PCA algorithm can be used to
linearly transform the data while both reducing the dimensionality and preserve most of the explained variance at the
same time.
The class PCA used with the optional parameter svd_solver='randomized' is very useful in that case: since
we are going to drop most of the singular vectors it is much more efficient to limit the computation to an approximated
estimate of the singular vectors we will keep to actually perform the transform.
For instance, the following shows 16 sample portraits (centered around 0.0) from the Olivetti dataset. On the right
hand side are the first 16 singular vectors reshaped as portraits. Since we only require the top 16 singular vectors of a
dataset with size 𝑛𝑠𝑎𝑚𝑝𝑙𝑒𝑠 = 400 and 𝑛𝑓 𝑒𝑎𝑡𝑢𝑟𝑒𝑠 = 64 × 64 = 4096, the computation time is less than 1s:



If we note 𝑛max = max(𝑛samples , 𝑛features ) and 𝑛min = min(𝑛samples , 𝑛features ), the time complexity of the random-
ized PCA is 𝑂(𝑛2max · 𝑛components ) instead of 𝑂(𝑛2max · 𝑛min ) for the exact method implemented in PCA.
The memory footprint of randomized PCA is also proportional to 2 · 𝑛max · 𝑛components instead of 𝑛max · 𝑛min for the
exact method.
Note: the implementation of inverse_transform in PCA with svd_solver='randomized' is not the exact
inverse transform of transform even when whiten=False (default).
Examples:
• Faces recognition example using eigenfaces and SVMs

• Faces dataset decompositions
References:
• “Finding structure with randomness: Stochastic algorithms for constructing approximate matrix decomposi-
tions” Halko, et al., 2009

Kernel PCA
KernelPCA is an extension of PCA which achieves non-linear dimensionality reduction through the use of
kernels (see Pairwise metrics, Affinities and Kernels). It has many applications including denoising, compres-
sion and structured prediction (kernel dependency estimation). KernelPCA supports both transform and
inverse_transform.
Examples:
• Kernel PCA
Sparse principal components analysis (SparsePCA and MiniBatchSparsePCA)
SparsePCA is a variant of PCA, with the goal of extracting the set of sparse components that best reconstruct the
data.
Mini-batch sparse PCA (MiniBatchSparsePCA) is a variant of SparsePCA that is faster but less accurate. The
increased speed is reached by iterating over small chunks of the set of features, for a given number of iterations.
Principal component analysis (PCA) has the disadvantage that the components extracted by this method have exclu-
sively dense expressions, i.e. they have non-zero coefficients when expressed as linear combinations of the original
variables. This can make interpretation difficult. In many cases, the real underlying components can be more naturally
imagined as sparse vectors; for example in face recognition, components might naturally map to parts of faces.
Sparse principal components yields a more parsimonious, interpretable representation, clearly emphasizing which of
the original features contribute to the differences between samples.
The following example illustrates 16 components extracted using sparse PCA from the Olivetti faces dataset. It can
be seen how the regularization term induces many zeros. Furthermore, the natural structure of the data causes the
non-zero coefficients to be vertically adjacent. The model does not enforce this mathematically: each component is

a vector ℎ ∈ R4096 , and there is no notion of vertical adjacency except during the human-friendly visualization as
64x64 pixel images. The fact that the components shown below appear local is the effect of the inherent structure of
the data, which makes such local patterns minimize reconstruction error. There exist sparsity-inducing norms that take
into account adjacency and different kinds of structure; see [Jen09] for a review of such methods. For more details on
how to use Sparse PCA, see the Examples section, below.
Note that there are many different formulations for the Sparse PCA problem. The one implemented here is based
on [Mrl09] . The optimization problem solved is a PCA problem (dictionary learning) with an ℓ1 penalty on the
components:
1
(𝑈 * , 𝑉 * ) = arg min ||𝑋 − 𝑈 𝑉 ||22 + 𝛼||𝑉 ||1
𝑈,𝑉 2
subject to ||𝑈𝑘 ||2 = 1 for all 0 ≤ 𝑘 < 𝑛𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠
The sparsity-inducing ℓ1 norm also prevents learning components from noise when few training samples are available.
The degree of penalization (and thus sparsity) can be adjusted through the hyperparameter alpha. Small values lead
to a gently regularized factorization, while larger values shrink many coefficients to zero.
Note: While in the spirit of an online algorithm, the class MiniBatchSparsePCA does not implement
partial_fit because the algorithm is online along the features direction, not the samples direction.

Examples:
References:
Truncated singular value decomposition and latent semantic analysis
TruncatedSVD implements a variant of singular value decomposition (SVD) that only computes the 𝑘 largest
singular values, where 𝑘 is a user-specified parameter.
When truncated SVD is applied to term-document matrices (as returned by CountVectorizer or
TfidfVectorizer), this transformation is known as latent semantic analysis (LSA), because it transforms such
matrices to a “semantic” space of low dimensionality. In particular, LSA is known to combat the effects of synonymy
and polysemy (both of which roughly mean there are multiple meanings per word), which cause term-document ma-
trices to be overly sparse and exhibit poor similarity under measures such as cosine similarity.
Note: LSA is also known as latent semantic indexing, LSI, though strictly that refers to its use in persistent indexes
for information retrieval purposes.
Mathematically, truncated SVD applied to training samples 𝑋 produces a low-rank approximation 𝑋:
𝑋 ≈ 𝑋𝑘 = 𝑈𝑘 Σ𝑘 𝑉𝑘⊤
After this operation, 𝑈𝑘 Σ⊤

𝑘 is the transformed training set with 𝑘 features (called n_components in the API).
To also transform a test set 𝑋, we multiply it with 𝑉𝑘 :
𝑋 ′ = 𝑋𝑉𝑘
Note: Most treatments of LSA in the natural language processing (NLP) and information retrieval (IR) literature
swap the axes of the matrix 𝑋 so that it has shape n_features × n_samples. We present LSA in a different way
that matches the scikit-learn API better, but the singular values found are the same.
TruncatedSVD is very similar to PCA, but differs in that it works on sample matrices 𝑋 directly instead of their
covariance matrices. When the columnwise (per-feature) means of 𝑋 are subtracted from the feature values, truncated
SVD on the resulting matrix is equivalent to PCA. In practical terms, this means that the TruncatedSVD transformer
accepts scipy.sparse matrices without the need to densify them, as densifying may fill up memory even for
medium-sized document collections.
While the TruncatedSVD transformer works with any (sparse) feature matrix, using it on tf–idf matrices is recom-
mended over raw frequency counts in an LSA/document processing setting. In particular, sublinear scaling and inverse
document frequency should be turned on (sublinear_tf=True, use_idf=True) to bring the feature values
closer to a Gaussian distribution, compensating for LSA’s erroneous assumptions about textual data.
Examples:
• Clustering text documents using k-means

References:
• Christopher D. Manning, Prabhakar Raghavan and Hinrich Schütze (2008), Introduction to Information Re-
trieval, Cambridge University Press, chapter 18: Matrix decompositions & latent semantic indexing
Dictionary Learning
Sparse coding with a precomputed dictionary
The SparseCoder object is an estimator that can be used to transform signals into sparse linear combination of
atoms from a fixed, precomputed dictionary such as a discrete wavelet basis. This object therefore does not implement
a fit method. The transformation amounts to a sparse coding problem: finding a representation of the data as a linear
combination of as few dictionary atoms as possible. All variations of dictionary learning implement the following
transform methods, controllable via the transform_method initialization parameter:
• Orthogonal matching pursuit (Orthogonal Matching Pursuit (OMP))
• Least-angle regression (Least Angle Regression)
• Lasso computed by least-angle regression
• Lasso using coordinate descent (Lasso)
• Thresholding
Thresholding is very fast but it does not yield accurate reconstructions. They have been shown useful in literature for
classification tasks. For image reconstruction tasks, orthogonal matching pursuit yields the most accurate, unbiased
reconstruction.
The dictionary learning objects offer, via the split_code parameter, the possibility to separate the positive and
negative values in the results of sparse coding. This is useful when dictionary learning is used for extracting features
that will be used for supervised learning, because it allows the learning algorithm to assign different weights to negative
loadings of a particular atom, from to the corresponding positive loading.
The split code for a single sample has length 2 * n_components and is constructed using the following rule:
First, the regular code of length n_components is computed. Then, the first n_components entries of the
split_code are filled with the positive part of the regular code vector. The second half of the split code is filled
with the negative part of the code vector, only with a positive sign. Therefore, the split_code is non-negative.
Examples:
• Sparse coding with a precomputed dictionary
Generic dictionary learning
Dictionary learning (DictionaryLearning) is a matrix factorization problem that amounts to finding a (usually
overcomplete) dictionary that will perform well at sparsely encoding the fitted data.
Representing data as sparse combinations of atoms from an overcomplete dictionary is suggested to be the way the
mammalian primary visual cortex works. Consequently, dictionary learning applied on image patches has been shown
to give good results in image processing tasks such as image completion, inpainting and denoising, as well as for
supervised recognition tasks.

Dictionary learning is an optimization problem solved by alternatively updating the sparse code, as a solution to
multiple Lasso problems, considering the dictionary fixed, and then updating the dictionary to best fit the sparse code.
1
(𝑈 * , 𝑉 * ) = arg min ||𝑋 − 𝑈 𝑉 ||22 + 𝛼||𝑈 ||1
𝑈,𝑉 2
subject to ||𝑉𝑘 ||2 = 1 for all 0 ≤ 𝑘 < 𝑛atoms
After using such a procedure to fit the dictionary, the transform is simply a sparse coding step that shares the same
implementation with all dictionary learning objects (see Sparse coding with a precomputed dictionary).
It is also possible to constrain the dictionary and/or code to be positive to match constraints that may be present in the
data. Below are the faces with different positivity constraints applied. Red indicates negative values, blue indicates
positive values, and white represents zeros.


The following image shows how a dictionary learned from 4x4 pixel image patches extracted from part of the image
of a raccoon face looks like.
Examples:
• Image denoising using dictionary learning
References:
• “Online dictionary learning for sparse coding” J. Mairal, F. Bach, J. Ponce, G. Sapiro, 2009
Mini-batch dictionary learning
MiniBatchDictionaryLearning implements a faster, but less accurate version of the dictionary learning algo-
rithm that is better suited for large datasets.
By default, MiniBatchDictionaryLearning divides the data into mini-batches and optimizes in an online
manner by cycling over the mini-batches for the specified number of iterations. However, at the moment it does not
implement a stopping condition.

The estimator also implements partial_fit, which updates the dictionary by iterating only once over a mini-batch.
This can be used for online learning when the data is not readily available from the start, or for when the data does not
fit into the memory.
Clustering for dictionary learning
Note that when using dictionary learning to extract a representation (e.g. for sparse coding) clustering can be a
good proxy to learn the dictionary. For instance the MiniBatchKMeans estimator is computationally efficient
and implements on-line learning with a partial_fit method.
Example: Online learning of a dictionary of parts of faces
Factor Analysis
In unsupervised learning we only have a dataset 𝑋 = {𝑥1 , 𝑥2 , . . . , 𝑥𝑛 }. How can this dataset be described mathemat-
ically? A very simple continuous latent variable model for 𝑋 is
𝑥𝑖 = 𝑊 ℎ𝑖 + 𝜇 + 𝜖
The vector ℎ𝑖 is called “latent” because it is unobserved. 𝜖 is considered a noise term distributed according to a
Gaussian with mean 0 and covariance Ψ (i.e. 𝜖 ∼ 𝒩 (0, Ψ)), 𝜇 is some arbitrary offset vector. Such a model is called
“generative” as it describes how 𝑥𝑖 is generated from ℎ𝑖 . If we use all the 𝑥𝑖 ‘s as columns to form a matrix X and all
the ℎ𝑖 ‘s as columns of a matrix H then we can write (with suitably defined M and E):
X = 𝑊H + M + E
In other words, we decomposed matrix X.

If ℎ𝑖 is given, the above equation automatically implies the following probabilistic interpretation:
𝑝(𝑥𝑖 |ℎ𝑖 ) = 𝒩 (𝑊 ℎ𝑖 + 𝜇, Ψ)
For a complete probabilistic model we also need a prior distribution for the latent variable ℎ. The most straightforward
assumption (based on the nice properties of the Gaussian distribution) is ℎ ∼ 𝒩 (0, I). This yields a Gaussian as the
marginal distribution of 𝑥:
𝑝(𝑥) = 𝒩 (𝜇, 𝑊 𝑊 𝑇 + Ψ)
Now, without any further assumptions the idea of having a latent variable ℎ would be superfluous – 𝑥 can be com-
pletely modelled with a mean and a covariance. We need to impose some more specific structure on one of these two
parameters. A simple additional assumption regards the structure of the error covariance Ψ:
• Ψ = 𝜎 2 I: This assumption leads to the probabilistic model of PCA.

• Ψ = diag(𝜓1 , 𝜓2 , . . . , 𝜓𝑛 ): This model is called FactorAnalysis, a classical statistical model. The matrix
W is sometimes called the “factor loading matrix”.
Both models essentially estimate a Gaussian with a low-rank covariance matrix. Because both models are probabilistic
they can be integrated in more complex models, e.g. Mixture of Factor Analysers. One gets very different models (e.g.
FastICA) if non-Gaussian priors on the latent variables are assumed.
Factor analysis can produce similar components (the columns of its loading matrix) to PCA. However, one can not
make any general statements about these components (e.g. whether they are orthogonal):
The main advantage for Factor Analysis over PCA is that it can model the variance in every direction of the input space
independently (heteroscedastic noise):
This allows better model selection than probabilistic PCA in the presence of heteroscedastic noise:
Examples:


Independent component analysis (ICA)
Independent component analysis separates a multivariate signal into additive subcomponents that are maximally in-
dependent. It is implemented in scikit-learn using the Fast ICA algorithm. Typically, ICA is not used for reducing
dimensionality but for separating superimposed signals. Since the ICA model does not include a noise term, for the
model to be correct, whitening must be applied. This can be done internally using the whiten argument or manually
using one of the PCA variants.
It is classically used to separate mixed signals (a problem known as blind source separation), as in the example below:
ICA can also be used as yet another non linear decomposition that finds components with some sparsity:

Examples:
• Blind source separation using FastICA

• FastICA on 2D point clouds
Non-negative matrix factorization (NMF or NNMF)
NMF with the Frobenius norm
NMF 1 is an alternative approach to decomposition that assumes that the data and the components are non-negative.
NMF can be plugged in instead of PCA or its variants, in the cases where the data matrix does not contain negative
values. It finds a decomposition of samples 𝑋 into two matrices 𝑊 and 𝐻 of non-negative elements, by optimizing the
distance 𝑑 between 𝑋 and the matrix product 𝑊 𝐻. The most widely used distance function is the squared Frobenius
norm, which is an obvious extension of the Euclidean norm to matrices:
1 1 ∑︁
𝑑Fro (𝑋, 𝑌 ) = ||𝑋 − 𝑌 ||2Fro = (𝑋𝑖𝑗 − 𝑌𝑖𝑗 )2
2 2 𝑖,𝑗
Unlike PCA, the representation of a vector is obtained in an additive fashion, by superimposing the components,
without subtracting. Such additive models are efficient for representing images and text.
It has been observed in [Hoyer, 2004]2 that, when carefully constrained, NMF can produce a parts-based representation
of the dataset, resulting in interpretable models. The following example displays 16 sparse components found by NMF
from the images in the Olivetti faces dataset, in comparison with the PCA eigenfaces.
1 “Learning the parts of objects by non-negative matrix factorization” D. Lee, S. Seung, 1999
2 “Non-negative Matrix Factorization with Sparseness Constraints” P. Hoyer, 2004

The init attribute determines the initialization method applied, which has a great impact on the performance of the
method. NMF implements the method Nonnegative Double Singular Value Decomposition. NNDSVD4 is based on
two SVD processes, one approximating the data matrix, the other approximating positive sections of the resulting
partial SVD factors utilizing an algebraic property of unit rank matrices. The basic NNDSVD algorithm is better fit
for sparse factorization. Its variants NNDSVDa (in which all zeros are set equal to the mean of all elements of the
data), and NNDSVDar (in which the zeros are set to random perturbations less than the mean of the data divided by
100) are recommended in the dense case.
Note that the Multiplicative Update (‘mu’) solver cannot update zeros present in the initialization, so it leads to poorer
results when used jointly with the basic NNDSVD algorithm which introduces a lot of zeros; in this case, NNDSVDa
or NNDSVDar should be preferred.
NMF can also be initialized with correctly scaled random non-negative matrices by setting init="random". An
integer seed or a RandomState can also be passed to random_state to control reproducibility.
In NMF, L1 and L2 priors can be added to the loss function in order to regularize the model. The L2 prior uses the
Frobenius norm, while the L1 prior uses an elementwise L1 norm. As in ElasticNet, we control the combination
of L1 and L2 with the l1_ratio (𝜌) parameter, and the intensity of the regularization with the alpha (𝛼) parameter.
Then the priors terms are:
𝛼(1 − 𝜌) 𝛼(1 − 𝜌)
𝛼𝜌||𝑊 ||1 + 𝛼𝜌||𝐻||1 + ||𝑊 ||2Fro + ||𝐻||2Fro
2 2
4 “SVD based initialization: A head start for nonnegative matrix factorization” C. Boutsidis, E. Gallopoulos, 2008

and the regularized objective function is:
𝛼(1 − 𝜌) 𝛼(1 − 𝜌)
𝑑Fro (𝑋, 𝑊 𝐻) + 𝛼𝜌||𝑊 ||1 + 𝛼𝜌||𝐻||1 + ||𝑊 ||2Fro + ||𝐻||2Fro
2 2
NMF regularizes both W and H. The public function non_negative_factorization allows a finer control
through the regularization attribute, and may regularize only W, only H, or both.
NMF with a beta-divergence
As described previously, the most widely used distance function is the squared Frobenius norm, which is an obvious
extension of the Euclidean norm to matrices:
1 1 ∑︁
𝑑Fro (𝑋, 𝑌 ) = ||𝑋 − 𝑌 ||2𝐹 𝑟𝑜 = (𝑋𝑖𝑗 − 𝑌𝑖𝑗 )2
2 2 𝑖,𝑗
Other distance functions can be used in NMF as, for example, the (generalized) Kullback-Leibler (KL) divergence,
also referred as I-divergence:
∑︁ 𝑋𝑖𝑗
𝑑𝐾𝐿 (𝑋, 𝑌 ) = (𝑋𝑖𝑗 log( ) − 𝑋𝑖𝑗 + 𝑌𝑖𝑗 )
𝑖,𝑗
𝑌𝑖𝑗
Or, the Itakura-Saito (IS) divergence:

∑︁ 𝑋𝑖𝑗 𝑋𝑖𝑗
𝑑𝐼𝑆 (𝑋, 𝑌 ) = ( − log( ) − 1)
𝑖,𝑗
𝑌𝑖𝑗 𝑌𝑖𝑗
These three distances are special cases of the beta-divergence family, with 𝛽 = 2, 1, 0 respectively6 . The beta-
divergence are defined by :
∑︁ 1
𝑑𝛽 (𝑋, 𝑌 ) = (𝑋 𝛽 + (𝛽 − 1)𝑌𝑖𝑗𝛽 − 𝛽𝑋𝑖𝑗 𝑌𝑖𝑗𝛽−1 )
𝑖,𝑗
𝛽(𝛽 − 1) 𝑖𝑗
Note that this definition is not valid if 𝛽 ∈ (0; 1), yet it can be continuously extended to the definitions of 𝑑𝐾𝐿 and
𝑑𝐼𝑆 respectively.
NMF implements two solvers, using Coordinate Descent (‘cd’)5 , and Multiplicative Update (‘mu’)6 . The ‘mu’ solver
can optimize every beta-divergence, including of course the Frobenius norm (𝛽 = 2), the (generalized) Kullback-
Leibler divergence (𝛽 = 1) and the Itakura-Saito divergence (𝛽 = 0). Note that for 𝛽 ∈ (1; 2), the ‘mu’ solver is
significantly faster than for other values of 𝛽. Note also that with a negative (or 0, i.e. ‘itakura-saito’) 𝛽, the input
matrix cannot contain zero values.
The ‘cd’ solver can only optimize the Frobenius norm. Due to the underlying non-convexity of NMF, the different
solvers may converge to different minima, even when optimizing the same distance function.
NMF is best used with the fit_transform method, which returns the matrix W. The matrix H is stored into the
fitted model in the components_ attribute; the method transform will decompose a new matrix X_new based on
these stored components:

>>> X = np.array([[1, 1], [2, 1], [3, 1.2], [4, 1], [5, 0.8], [6, 1]])
>>> from sklearn.decomposition import NMF
>>> model = NMF(n_components=2, init='random', random_state=0)
6 “Algorithms for nonnegative matrix factorization with the beta-divergence” C. Fevotte, J. Idier, 2011
5 “Fast local algorithms for large scale nonnegative matrix and tensor factorizations.” A. Cichocki, A. Phan, 2009

>>> W = model.fit_transform(X)
>>> H = model.components_
>>> X_new = np.array([[1, 0], [1, 6.1], [1, 0], [1, 4], [3.2, 1], [0, 4]])
>>> W_new = model.transform(X_new)
Examples:

• Topic extraction with Non-negative Matrix Factorization and Latent Dirichlet Allocation
• Beta-divergence loss functions
References:
Latent Dirichlet Allocation (LDA)
Latent Dirichlet Allocation is a generative probabilistic model for collections of discrete dataset such as text corpora.
It is also a topic model that is used for discovering abstract topics from a collection of documents.
The graphical model of LDA is a three-level generative model:

Note on notations presented in the graphical model above, which can be found in Hoffman et al. (2013):
• The corpus is a collection of 𝐷 documents.
• A document is a sequence of 𝑁 words.
• There are 𝐾 topics in the corpus.
• The boxes represent repeated sampling.
In the graphical model, each node is a random variable and has a role in the generative process. A shaded node
indicates an observed variable and an unshaded node indicates a hidden (latent) variable. In this case, words in the
corpus are the only data that we observe. The latent variables determine the random mixture of topics in the corpus
and the distribution of words in the documents. The goal of LDA is to use the observed words to infer the hidden topic
structure.
When modeling text corpora, the model assumes the following generative process for a corpus with 𝐷 documents and
𝐾 topics, with 𝐾 corresponding to n_components in the API:
1. For each topic 𝑘 ∈ 𝐾, draw 𝛽𝑘 ∼ Dirichlet(𝜂). This provides a distribution over the words, i.e. the
probability of a word appearing in topic 𝑘. 𝜂 corresponds to topic_word_prior.
2. For each document 𝑑 ∈ 𝐷, draw the topic proportions 𝜃𝑑 ∼ Dirichlet(𝛼). 𝛼 corresponds to
doc_topic_prior.
3. For each word 𝑖 in document 𝑑:
1. Draw the topic assignment 𝑧𝑑𝑖 ∼ Multinomial(𝜃𝑑 )
2. Draw the observed word 𝑤𝑖𝑗 ∼ Multinomial(𝛽𝑧𝑑𝑖 )
For parameter estimation, the posterior distribution is:
𝑝(𝑧, 𝜃, 𝛽|𝛼, 𝜂)
𝑝(𝑧, 𝜃, 𝛽|𝑤, 𝛼, 𝜂) =
𝑝(𝑤|𝛼, 𝜂)
Since the posterior is intractable, variational Bayesian method uses a simpler distribution 𝑞(𝑧, 𝜃, 𝛽|𝜆, 𝜑, 𝛾) to approx-
imate it, and those variational parameters 𝜆, 𝜑, 𝛾 are optimized to maximize the Evidence Lower Bound (ELBO):
△
log 𝑃 (𝑤|𝛼, 𝜂) ≥ 𝐿(𝑤, 𝜑, 𝛾, 𝜆) = 𝐸𝑞 [log 𝑝(𝑤, 𝑧, 𝜃, 𝛽|𝛼, 𝜂)] − 𝐸𝑞 [log 𝑞(𝑧, 𝜃, 𝛽)]
Maximizing ELBO is equivalent to minimizing the Kullback-Leibler(KL) divergence between 𝑞(𝑧, 𝜃, 𝛽) and the true
posterior 𝑝(𝑧, 𝜃, 𝛽|𝑤, 𝛼, 𝜂).
LatentDirichletAllocation implements the online variational Bayes algorithm and supports both online and
batch update methods. While the batch method updates variational variables after each full pass through the data, the
online method updates variational variables from mini-batch data points.
Note: Although the online method is guaranteed to converge to a local optimum point, the quality of the optimum
point and the speed of convergence may depend on mini-batch size and attributes related to learning rate setting.

When LatentDirichletAllocation is applied on a “document-term” matrix, the matrix will be decomposed

into a “topic-term” matrix and a “document-topic” matrix. While “topic-term” matrix is stored as components_ in
the model, “document-topic” matrix can be calculated from transform method.
LatentDirichletAllocation also implements partial_fit method. This is used when data can be fetched
sequentially.
Examples:
References:
• “Latent Dirichlet Allocation” D. Blei, A. Ng, M. Jordan, 2003

• “Online Learning for Latent Dirichlet Allocation” M. Hoffman, D. Blei, F. Bach, 2010
• “Stochastic Variational Inference” M. Hoffman, D. Blei, C. Wang, J. Paisley, 2013
3.2.6 Covariance estimation
Many statistical problems require the estimation of a population’s covariance matrix, which can be seen as an estima-
tion of data set scatter plot shape. Most of the time, such an estimation has to be done on a sample whose properties
(size, structure, homogeneity) have a large influence on the estimation’s quality. The sklearn.covariance package
provides tools for accurately estimating a population’s covariance matrix under various settings.
We assume that the observations are independent and identically distributed (i.i.d.).
Empirical covariance
The covariance matrix of a data set is known to be well approximated by the classical maximum likelihood estimator
(or “empirical covariance”), provided the number of observations is large enough compared to the number of features
(the variables describing the observations). More precisely, the Maximum Likelihood Estimator of a sample is an
unbiased estimator of the corresponding population’s covariance matrix.
The empirical covariance matrix of a sample can be computed using the empirical_covariance function of the
package, or by fitting an EmpiricalCovariance object to the data sample with the EmpiricalCovariance.
fit method. Be careful that results depend on whether the data are centered, so one may want to use the
assume_centered parameter accurately. More precisely, if assume_centered=False, then the test set
is supposed to have the same mean vector as the training set. If not, both should be centered by the user, and
assume_centered=True should be used.
Examples:
• See Shrinkage covariance estimation: LedoitWolf vs OAS and max-likelihood for an example on how to fit an
EmpiricalCovariance object to data.

Shrunk Covariance
Basic shrinkage
Despite being an unbiased estimator of the covariance matrix, the Maximum Likelihood Estimator is not a good esti-
mator of the eigenvalues of the covariance matrix, so the precision matrix obtained from its inversion is not accurate.
Sometimes, it even occurs that the empirical covariance matrix cannot be inverted for numerical reasons. To avoid
such an inversion problem, a transformation of the empirical covariance matrix has been introduced: the shrinkage.
In scikit-learn, this transformation (with a user-defined shrinkage coefficient) can be directly applied to a pre-computed
covariance with the shrunk_covariance method. Also, a shrunk estimator of the covariance can be fitted to data
with a ShrunkCovariance object and its ShrunkCovariance.fit method. Again, results depend on whether
the data are centered, so one may want to use the assume_centered parameter accurately.
Mathematically, this shrinkage consists in reducing the ratio between the smallest and the largest eigenvalues of the
empirical covariance matrix. It can be done by simply shifting every eigenvalue according to a given offset, which is
equivalent of finding the l2-penalized Maximum Likelihood Estimator of the covariance matrix. In practice, shrinkage
^
boils down to a simple a convex transformation : Σshrunk = (1 − 𝛼)Σ̂ + 𝛼 Tr𝑝Σ Id.
Choosing the amount of shrinkage, 𝛼 amounts to setting a bias/variance trade-off, and is discussed below.
Examples:
• See Shrinkage covariance estimation: LedoitWolf vs OAS and max-likelihood for an example on how to fit a
ShrunkCovariance object to data.
Ledoit-Wolf shrinkage
In their 2004 paper1 , O. Ledoit and M. Wolf propose a formula to compute the optimal shrinkage coefficient 𝛼 that
minimizes the Mean Squared Error between the estimated and the real covariance matrix.
The Ledoit-Wolf estimator of the covariance matrix can be computed on a sample with the ledoit_wolf function of
the sklearn.covariance package, or it can be otherwise obtained by fitting a LedoitWolf object to the same sample.
Note: Case when population covariance matrix is isotropic

It is important to note that when the number of samples is much larger than the number of features, one would expect
that no shrinkage would be necessary. The intuition behind this is that if the population covariance is full rank, when
the number of sample grows, the sample covariance will also become positive definite. As a result, no shrinkage would
necessary and the method should automatically do this.
This, however, is not the case in the Ledoit-Wolf procedure when the population covariance happens to be a multiple of
the identity matrix. In this case, the Ledoit-Wolf shrinkage estimate approaches 1 as the number of samples increases.
This indicates that the optimal estimate of the covariance matrix in the Ledoit-Wolf sense is multiple of the identity.
Since the population covariance is already a multiple of the identity matrix, the Ledoit-Wolf solution is indeed a
reasonable estimate.
Examples:
1 O. Ledoit and M. Wolf, “A Well-Conditioned Estimator for Large-Dimensional Covariance Matrices”, Journal of Multivariate Analysis, Vol-
ume 88, Issue 2, February 2004, pages 365-411.

• See Shrinkage covariance estimation: LedoitWolf vs OAS and max-likelihood for an example on how to fit a
LedoitWolf object to data and for visualizing the performances of the Ledoit-Wolf estimator in terms of
likelihood.
References:
Oracle Approximating Shrinkage
Under the assumption that the data are Gaussian distributed, Chen et al.2 derived a formula aimed at choosing a
shrinkage coefficient that yields a smaller Mean Squared Error than the one given by Ledoit and Wolf’s formula. The
resulting estimator is known as the Oracle Shrinkage Approximating estimator of the covariance.
The OAS estimator of the covariance matrix can be computed on a sample with the oas function of the
sklearn.covariance package, or it can be otherwise obtained by fitting an OAS object to the same sample.
Fig. 3.7: Bias-variance trade-off when setting the shrinkage: comparing the choices of Ledoit-Wolf and OAS estima-
tors
References:
Examples:
• See Shrinkage covariance estimation: LedoitWolf vs OAS and max-likelihood for an example on how to fit an
OAS object to data.
2 Chen et al., “Shrinkage Algorithms for MMSE Covariance Estimation”, IEEE Trans. on Sign. Proc., Volume 58, Issue 10, October 2010.

• See Ledoit-Wolf vs OAS estimation to visualize the Mean Squared Error difference between a LedoitWolf
and an OAS estimator of the covariance.
Sparse inverse covariance
The matrix inverse of the covariance matrix, often called the precision matrix, is proportional to the partial correlation
matrix. It gives the partial independence relationship. In other words, if two features are independent conditionally on
the others, the corresponding coefficient in the precision matrix will be zero. This is why it makes sense to estimate
a sparse precision matrix: the estimation of the covariance matrix is better conditioned by learning independence
relations from the data. This is known as covariance selection.
In the small-samples situation, in which n_samples is on the order of n_features or smaller, sparse inverse
covariance estimators tend to work better than shrunk covariance estimators. However, in the opposite situation, or for
very correlated data, they can be numerically unstable. In addition, unlike shrinkage estimators, sparse estimators are
able to recover off-diagonal structure.
The GraphicalLasso estimator uses an l1 penalty to enforce sparsity on the precision matrix: the higher its
alpha parameter, the more sparse the precision matrix. The corresponding GraphicalLassoCV object uses
cross-validation to automatically set the alpha parameter.
Note: Structure recovery

Recovering a graphical structure from correlations in the data is a challenging thing. If you are interested in such
recovery keep in mind that:
• Recovery is easier from a correlation matrix than a covariance matrix: standardize your observations before
running GraphicalLasso
• If the underlying graph has nodes with much more connections than the average node, the algorithm will miss
some of these connections.

Fig. 3.8: A comparison of maximum likelihood, shrinkage and sparse estimates of the covariance and precision matrix
in the very small samples settings.
• If your number of observations is not large compared to the number of edges in your underlying graph, you will
not recover it.
• Even if you are in favorable recovery conditions, the alpha parameter chosen by cross-validation (e.g. using the
GraphicalLassoCV object) will lead to selecting too many edges. However, the relevant edges will have
heavier weights than the irrelevant ones.
The mathematical formulation is the following:
ˆ = argmin𝐾 tr𝑆𝐾 − logdet𝐾 + 𝛼‖𝐾‖1

(︀ )︀
𝐾
Where 𝐾 is the precision matrix to be estimated, and 𝑆 is the sample covariance matrix. ‖𝐾‖1 is the sum of the abso-
lute values of off-diagonal coefficients of 𝐾. The algorithm employed to solve this problem is the GLasso algorithm,
from the Friedman 2008 Biostatistics paper. It is the same algorithm as in the R glasso package.
Examples:
• Sparse inverse covariance estimation: example on synthetic data showing some recovery of a structure, and
comparing to other covariance estimators.
• Visualizing the stock market structure: example on real stock market data, finding which symbols are most
linked.
References:
• Friedman et al, “Sparse inverse covariance estimation with the graphical lasso”, Biostatistics 9, pp 432, 2008

Robust Covariance Estimation
Real data sets are often subject to measurement or recording errors. Regular but uncommon observations may also
appear for a variety of reasons. Observations which are very uncommon are called outliers. The empirical covariance
estimator and the shrunk covariance estimators presented above are very sensitive to the presence of outliers in the data.
Therefore, one should use robust covariance estimators to estimate the covariance of its real data sets. Alternatively,
robust covariance estimators can be used to perform outlier detection and discard/downweight some observations
according to further processing of the data.
The sklearn.covariance package implements a robust estimator of covariance, the Minimum Covariance De-
terminant3 .
Minimum Covariance Determinant
The Minimum Covariance Determinant estimator is a robust estimator of a data set’s covariance introduced by P.J.
Rousseeuw in3 . The idea is to find a given proportion (h) of “good” observations which are not outliers and compute
their empirical covariance matrix. This empirical covariance matrix is then rescaled to compensate the performed
selection of observations (“consistency step”). Having computed the Minimum Covariance Determinant estimator,
one can give weights to observations according to their Mahalanobis distance, leading to a reweighted estimate of the
covariance matrix of the data set (“reweighting step”).
Rousseeuw and Van Driessen4 developed the FastMCD algorithm in order to compute the Minimum Covariance
Determinant. This algorithm is used in scikit-learn when fitting an MCD object to data. The FastMCD algorithm also
computes a robust estimate of the data set location at the same time.
Raw estimates can be accessed as raw_location_ and raw_covariance_ attributes of a MinCovDet robust
covariance estimator object.
References:
Examples:
• See Robust vs Empirical covariance estimate for an example on how to fit a MinCovDet object to data and
see how the estimate remains accurate despite the presence of outliers.
• See Robust covariance estimation and Mahalanobis distances relevance to visualize the difference between
EmpiricalCovariance and MinCovDet covariance estimators in terms of Mahalanobis distance (so
we get a better estimate of the precision matrix too).
3P. J. Rousseeuw. Least median of squares regression. J. Am Stat Ass, 79:871, 1984.
4A Fast Algorithm for the Minimum Covariance Determinant Estimator, 1999, American Statistical Association and the American Society for
Quality, TECHNOMETRICS.

Influence of outliers on location and covariance Separating inliers from outliers using a Mahalanobis
estimates distance
3.2.7 Novelty and Outlier Detection
Many applications require being able to decide whether a new observation belongs to the same distribution as existing
observations (it is an inlier), or should be considered as different (it is an outlier). Often, this ability is used to clean
real data sets. Two important distinctions must be made:
outlier detection The training data contains outliers which are defined as observations that are far from
the others. Outlier detection estimators thus try to fit the regions where the training data is the most
concentrated, ignoring the deviant observations.
novelty detection The training data is not polluted by outliers and we are interested in detecting whether
a new observation is an outlier. In this context an outlier is also called a novelty.
Outlier detection and novelty detection are both used for anomaly detection, where one is interested in detecting
abnormal or unusual observations. Outlier detection is then also known as unsupervised anomaly detection and novelty
detection as semi-supervised anomaly detection. In the context of outlier detection, the outliers/anomalies cannot form
a dense cluster as available estimators assume that the outliers/anomalies are located in low density regions. On the
contrary, in the context of novelty detection, novelties/anomalies can form a dense cluster as long as they are in a low
density region of the training data, considered as normal in this context.
The scikit-learn project provides a set of machine learning tools that can be used both for novelty or outlier detection.
This strategy is implemented with objects learning in an unsupervised way from the data:
estimator.fit(X_train)
new observations can then be sorted as inliers or outliers with a predict method:
estimator.predict(X_test)
Inliers are labeled 1, while outliers are labeled -1. The predict method makes use of a threshold on the raw scoring
function computed by the estimator. This scoring function is accessible through the score_samples method, while
the threshold can be controlled by the contamination parameter.
The decision_function method is also defined from the scoring function, in such a way that negative values are
outliers and non-negative ones are inliers:
estimator.decision_function(X_test)

Note that neighbors.LocalOutlierFactor does not support predict, decision_function and

score_samples methods by default but only a fit_predict method, as this estimator was originally meant
to be applied for outlier detection. The scores of abnormality of the training samples are accessible through the
negative_outlier_factor_ attribute.
If you really want to use neighbors.LocalOutlierFactor for novelty detection, i.e. predict labels or compute
the score of abnormality of new unseen data, you can instantiate the estimator with the novelty parameter set to
True before fitting the estimator. In this case, fit_predict is not available.
Warning: Novelty detection with Local Outlier Factor

When novelty is set to True be aware that you must only use predict, decision_function
and score_samples on new unseen data and not on the training samples as this would lead to
wrong results. The scores of abnormality of the training samples are always accessible through the
The behavior of neighbors.LocalOutlierFactor is summarized in the following table.
Method Outlier detection Novelty detection

fit_predict OK Not available
predict Not available Use only on new data
decision_function Not available Use only on new data
score_samples Use negative_outlier_factor_ Use only on new data
Overview of outlier detection methods
A comparison of the outlier detection algorithms in scikit-learn. Local Outlier Factor (LOF) does not show a decision
boundary in black as it has no predict method to be applied on new data when it is used for outlier detection.
ensemble.IsolationForest and neighbors.LocalOutlierFactor perform reasonably well on the
data sets considered here. The svm.OneClassSVM is known to be sensitive to outliers and thus does not perform
very well for outlier detection. Finally, covariance.EllipticEnvelope assumes the data is Gaussian and
learns an ellipse. For more details on the different estimators refer to the example Comparing anomaly detection
algorithms for outlier detection on toy datasets and the sections hereunder.
Examples:
• See Comparing anomaly detection algorithms for outlier detection on toy datasets for a com-
parison of the svm.OneClassSVM , the ensemble.IsolationForest, the neighbors.
LocalOutlierFactor and covariance.EllipticEnvelope.
Novelty Detection
Consider a data set of 𝑛 observations from the same distribution described by 𝑝 features. Consider now that we add one
more observation to that data set. Is the new observation so different from the others that we can doubt it is regular?
(i.e. does it come from the same distribution?) Or on the contrary, is it so similar to the other that we cannot distinguish
it from the original observations? This is the question addressed by the novelty detection tools and methods.
In general, it is about to learn a rough, close frontier delimiting the contour of the initial observations distribution,
plotted in embedding 𝑝-dimensional space. Then, if further observations lay within the frontier-delimited subspace,


they are considered as coming from the same population than the initial observations. Otherwise, if they lay outside
the frontier, we can say that they are abnormal with a given confidence in our assessment.
The One-Class SVM has been introduced by Schölkopf et al. for that purpose and implemented in the Support Vector
Machines module in the svm.OneClassSVM object. It requires the choice of a kernel and a scalar parameter to
define a frontier. The RBF kernel is usually chosen although there exists no exact formula or algorithm to set its
bandwidth parameter. This is the default in the scikit-learn implementation. The 𝜈 parameter, also known as the
margin of the One-Class SVM, corresponds to the probability of finding a new, but regular, observation outside the
frontier.
References:
• Estimating the support of a high-dimensional distribution Schölkopf, Bernhard, et al. Neural computation
13.7 (2001): 1443-1471.
Examples:
• See One-class SVM with non-linear kernel (RBF) for visualizing the frontier learned around some data by a
svm.OneClassSVM object.
• Species distribution modeling
Outlier Detection
Outlier detection is similar to novelty detection in the sense that the goal is to separate a core of regular observa-
tions from some polluting ones, called outliers. Yet, in the case of outlier detection, we don’t have a clean data set
representing the population of regular observations that can be used to train any tool.

Fitting an elliptic envelope
One common way of performing outlier detection is to assume that the regular data come from a known distribution
(e.g. data are Gaussian distributed). From this assumption, we generally try to define the “shape” of the data, and can
define outlying observations as observations which stand far enough from the fit shape.
The scikit-learn provides an object covariance.EllipticEnvelope that fits a robust covariance estimate to
the data, and thus fits an ellipse to the central data points, ignoring points outside the central mode.
For instance, assuming that the inlier data are Gaussian distributed, it will estimate the inlier location and covariance
in a robust way (i.e. without being influenced by outliers). The Mahalanobis distances obtained from this estimate is
used to derive a measure of outlyingness. This strategy is illustrated below.
Examples:
• See Robust covariance estimation and Mahalanobis distances relevance for an illustration of the dif-
ference between using a standard (covariance.EmpiricalCovariance) or a robust estimate
(covariance.MinCovDet) of location and covariance to assess the degree of outlyingness of an ob-
servation.
References:
• Rousseeuw, P.J., Van Driessen, K. “A fast algorithm for the minimum covariance determinant estimator”
Technometrics 41(3), 212 (1999)

Isolation Forest
One efficient way of performing outlier detection in high-dimensional datasets is to use random forests. The
ensemble.IsolationForest ‘isolates’ observations by randomly selecting a feature and then randomly se-
lecting a split value between the maximum and minimum values of the selected feature.
Since recursive partitioning can be represented by a tree structure, the number of splittings required to isolate a sample
is equivalent to the path length from the root node to the terminating node.
This path length, averaged over a forest of such random trees, is a measure of normality and our decision function.
Random partitioning produces noticeably shorter paths for anomalies. Hence, when a forest of random trees collec-
tively produce shorter path lengths for particular samples, they are highly likely to be anomalies.
This strategy is illustrated below.
Examples:
• See IsolationForest example for an illustration of the use of IsolationForest.

• See Comparing anomaly detection algorithms for outlier detection on toy datasets for a comparison of
ensemble.IsolationForest with neighbors.LocalOutlierFactor, svm.OneClassSVM
(tuned to perform like an outlier detection method) and a covariance-based outlier detection with
covariance.EllipticEnvelope.
References:
• Liu, Fei Tony, Ting, Kai Ming and Zhou, Zhi-Hua. “Isolation forest.” Data Mining, 2008. ICDM‘08. Eighth
IEEE International Conference on.

Local Outlier Factor
Another efficient way to perform outlier detection on moderately high dimensional datasets is to use the Local Outlier
Factor (LOF) algorithm.
The neighbors.LocalOutlierFactor (LOF) algorithm computes a score (called local outlier factor) reflect-
ing the degree of abnormality of the observations. It measures the local density deviation of a given data point with
respect to its neighbors. The idea is to detect the samples that have a substantially lower density than their neighbors.
In practice the local density is obtained from the k-nearest neighbors. The LOF score of an observation is equal to the
ratio of the average local density of his k-nearest neighbors, and its own local density: a normal instance is expected
to have a local density similar to that of its neighbors, while abnormal data are expected to have much smaller local
density.
The number k of neighbors considered, (alias parameter n_neighbors) is typically chosen 1) greater than the minimum
number of objects a cluster has to contain, so that other objects can be local outliers relative to this cluster, and 2)
smaller than the maximum number of close by objects that can potentially be local outliers. In practice, such informa-
tions are generally not available, and taking n_neighbors=20 appears to work well in general. When the proportion of
outliers is high (i.e. greater than 10 %, as in the example below), n_neighbors should be greater (n_neighbors=35 in
the example below).
The strength of the LOF algorithm is that it takes both local and global properties of datasets into consideration: it can
perform well even in datasets where abnormal samples have different underlying densities. The question is not, how
isolated the sample is, but how isolated it is with respect to the surrounding neighborhood.
When applying LOF for outlier detection, there are no predict, decision_function and score_samples
methods but only a fit_predict method. The scores of abnormality of the training samples are accessi-
ble through the negative_outlier_factor_ attribute. Note that predict, decision_function and
score_samples can be used on new unseen data when LOF is applied for novelty detection, i.e. when the
novelty parameter is set to True. See Novelty detection with Local Outlier Factor.
This strategy is illustrated below.

Examples:
• See Outlier detection with Local Outlier Factor (LOF) for an illustration of the use of neighbors.
LocalOutlierFactor.
• See Comparing anomaly detection algorithms for outlier detection on toy datasets for a comparison with
other anomaly detection methods.
References:
• Breunig, Kriegel, Ng, and Sander (2000) LOF: identifying density-based local outliers. Proc. ACM SIGMOD
Novelty detection with Local Outlier Factor
To use neighbors.LocalOutlierFactor for novelty detection, i.e. predict labels or compute the score of
abnormality of new unseen data, you need to instantiate the estimator with the novelty parameter set to True
before fitting the estimator:
lof = LocalOutlierFactor(novelty=True)
lof.fit(X_train)
Note that fit_predict is not available in this case.
Warning: Novelty detection with Local Outlier Factor‘

When novelty is set to True be aware that you must only use predict, decision_function
and score_samples on new unseen data and not on the training samples as this would lead to
wrong results. The scores of abnormality of the training samples are always accessible through the
Novelty detection with Local Outlier Factor is illustrated below.
3.2.8 Density Estimation
Density estimation walks the line between unsupervised learning, feature engineering, and data modeling. Some of
the most popular and useful density estimation techniques are mixture models such as Gaussian Mixtures (sklearn.
mixture.GaussianMixture), and neighbor-based approaches such as the kernel density estimate (sklearn.
neighbors.KernelDensity). Gaussian Mixtures are discussed more fully in the context of clustering, because
the technique is also useful as an unsupervised clustering scheme.
Density estimation is a very simple concept, and most people are already familiar with one common density estimation
technique: the histogram.
Density Estimation: Histograms
A histogram is a simple visualization of data where bins are defined, and the number of data points within each bin is
tallied. An example of a histogram can be seen in the upper-left panel of the following figure:

A major problem with histograms, however, is that the choice of binning can have a disproportionate effect on the
resulting visualization. Consider the upper-right panel of the above figure. It shows a histogram over the same data,
with the bins shifted right. The results of the two visualizations look entirely different, and might lead to different
interpretations of the data.
Intuitively, one can also think of a histogram as a stack of blocks, one block per point. By stacking the blocks in the
appropriate grid space, we recover the histogram. But what if, instead of stacking the blocks on a regular grid, we

center each block on the point it represents, and sum the total height at each location? This idea leads to the lower-left
visualization. It is perhaps not as clean as a histogram, but the fact that the data drive the block locations mean that it
is a much better representation of the underlying data.
This visualization is an example of a kernel density estimation, in this case with a top-hat kernel (i.e. a square block
at each point). We can recover a smoother distribution by using a smoother kernel. The bottom-right plot shows a
Gaussian kernel density estimate, in which each point contributes a Gaussian curve to the total. The result is a smooth
density estimate which is derived from the data, and functions as a powerful non-parametric model of the distribution
of points.
Kernel Density Estimation
Kernel density estimation in scikit-learn is implemented in the sklearn.neighbors.KernelDensity esti-

mator, which uses the Ball Tree or KD Tree for efficient queries (see Nearest Neighbors for a discussion of these).
Though the above example uses a 1D data set for simplicity, kernel density estimation can be performed in any number
of dimensions, though in practice the curse of dimensionality causes its performance to degrade in high dimensions.
In the following figure, 100 points are drawn from a bimodal distribution, and the kernel density estimates are shown
for three choices of kernels:
It’s clear how the kernel shape affects the smoothness of the resulting distribution. The scikit-learn kernel density
estimator can be used as follows:
>>> from sklearn.neighbors.kde import KernelDensity

>>> X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]])
>>> kde = KernelDensity(kernel='gaussian', bandwidth=0.2).fit(X)
>>> kde.score_samples(X)
array([-0.41075698, -0.41075698, -0.41076071, -0.41075698, -0.41075698,
-0.41076071])

Here we have used kernel='gaussian', as seen above. Mathematically, a kernel is a positive function 𝐾(𝑥; ℎ)
which is controlled by the bandwidth parameter ℎ. Given this kernel form, the density estimate at a point 𝑦 within a
group of points 𝑥𝑖 ; 𝑖 = 1 · · · 𝑁 is given by:
𝑁
∑︁
𝜌𝐾 (𝑦) = 𝐾((𝑦 − 𝑥𝑖 )/ℎ)
𝑖=1
The bandwidth here acts as a smoothing parameter, controlling the tradeoff between bias and variance in the result. A
large bandwidth leads to a very smooth (i.e. high-bias) density distribution. A small bandwidth leads to an unsmooth
(i.e. high-variance) density distribution.
sklearn.neighbors.KernelDensity implements several common kernel forms, which are shown in the
following figure:
The form of these kernels is as follows:

• Gaussian kernel (kernel = 'gaussian')
2
𝑥
𝐾(𝑥; ℎ) ∝ exp(− 2ℎ 2)
• Tophat kernel (kernel = 'tophat')

𝐾(𝑥; ℎ) ∝ 1 if 𝑥 < ℎ
• Epanechnikov kernel (kernel = 'epanechnikov')
𝑥2
𝐾(𝑥; ℎ) ∝ 1 − ℎ2
• Exponential kernel (kernel = 'exponential')

𝐾(𝑥; ℎ) ∝ exp(−𝑥/ℎ)
• Linear kernel (kernel = 'linear')
𝐾(𝑥; ℎ) ∝ 1 − 𝑥/ℎ if 𝑥 < ℎ

• Cosine kernel (kernel = 'cosine')

𝐾(𝑥; ℎ) ∝ cos( 𝜋𝑥
2ℎ ) if 𝑥 < ℎ
The kernel density estimator can be used with any of the valid distance metrics (see sklearn.neighbors.
DistanceMetric for a list of available metrics), though the results are properly normalized only for the Euclidean
metric. One particularly useful metric is the Haversine distance which measures the angular distance between points
on a sphere. Here is an example of using a kernel density estimate for a visualization of geospatial data, in this case
the distribution of observations of two different species on the South American continent:
One other useful application of kernel density estimation is to learn a non-parametric generative model of a dataset in
order to efficiently draw new samples from this generative model. Here is an example of using this process to create a
new set of hand-written digits, using a Gaussian kernel learned on a PCA projection of the data:

The “new” data consists of linear combinations of the input data, with weights probabilistically drawn given the KDE
model.
Examples:
• Simple 1D Kernel Density Estimation: computation of simple kernel density estimates in one dimension.
• Kernel Density Estimation: an example of using Kernel Density estimation to learn a generative model of the
hand-written digits data, and drawing new samples from this model.
• Kernel Density Estimate of Species Distributions: an example of Kernel Density estimation using the Haver-
sine distance metric to visualize geospatial data
3.2.9 Neural network models (unsupervised)
Restricted Boltzmann machines
Restricted Boltzmann machines (RBM) are unsupervised nonlinear feature learners based on a probabilistic model.
The features extracted by an RBM or a hierarchy of RBMs often give good results when fed into a linear classifier
such as a linear SVM or a perceptron.
The model makes assumptions regarding the distribution of inputs. At the moment, scikit-learn only provides
BernoulliRBM , which assumes the inputs are either binary values or values between 0 and 1, each encoding the
probability that the specific feature would be turned on.
The RBM tries to maximize the likelihood of the data using a particular graphical model. The parameter learning
algorithm used (Stochastic Maximum Likelihood) prevents the representations from straying far from the input data,
which makes them capture interesting regularities, but makes the model less useful for small datasets, and usually not
useful for density estimation.

The method gained popularity for initializing deep neural networks with the weights of independent RBMs. This
method is known as unsupervised pre-training.
Examples:
• Restricted Boltzmann Machine features for digit classification
Graphical model and parametrization
The graphical model of an RBM is a fully-connected bipartite graph.

The nodes are random variables whose states depend on the state of the other nodes they are connected to. The model
is therefore parameterized by the weights of the connections, as well as one intercept (bias) term for each visible and
hidden unit, omitted from the image for simplicity.
The energy function measures the quality of a joint assignment:
∑︁ ∑︁ ∑︁ ∑︁
𝐸(v, h) = − 𝑤𝑖𝑗 𝑣𝑖 ℎ𝑗 − 𝑏𝑖 𝑣 𝑖 − 𝑐𝑗 ℎ𝑗
𝑖 𝑗 𝑖 𝑗
In the formula above, b and c are the intercept vectors for the visible and hidden layers, respectively. The joint
probability of the model is defined in terms of the energy:
𝑒−𝐸(v,h)
𝑃 (v, h) =
𝑍
The word restricted refers to the bipartite structure of the model, which prohibits direct interaction between hidden
units, or between visible units. This means that the following conditional independencies are assumed:
ℎ𝑖 ⊥ℎ𝑗 |v
𝑣𝑖 ⊥𝑣𝑗 |h
The bipartite structure allows for the use of efficient block Gibbs sampling for inference.
Bernoulli Restricted Boltzmann machines
In the BernoulliRBM , all units are binary stochastic units. This means that the input data should either be binary, or
real-valued between 0 and 1 signifying the probability that the visible unit would turn on or off. This is a good model
for character recognition, where the interest is on which pixels are active and which aren’t. For images of natural
scenes it no longer fits because of background, depth and the tendency of neighbouring pixels to take the same values.
The conditional probability distribution of each unit is given by the logistic sigmoid activation function of the input it
receives:
∑︁
𝑃 (𝑣𝑖 = 1|h) = 𝜎( 𝑤𝑖𝑗 ℎ𝑗 + 𝑏𝑖 )
𝑗
∑︁
𝑃 (ℎ𝑖 = 1|v) = 𝜎( 𝑤𝑖𝑗 𝑣𝑖 + 𝑐𝑗 )
𝑖

where 𝜎 is the logistic sigmoid function:

1
𝜎(𝑥) =
1 + 𝑒−𝑥
Stochastic Maximum Likelihood learning
The training algorithm implemented in BernoulliRBM is known as Stochastic Maximum Likelihood (SML) or
Persistent Contrastive Divergence (PCD). Optimizing maximum likelihood directly is infeasible because of the form
of the data likelihood:
∑︁ ∑︁
log 𝑃 (𝑣) = log 𝑒−𝐸(𝑣,ℎ) − log 𝑒−𝐸(𝑥,𝑦)
ℎ 𝑥,𝑦
For simplicity the equation above is written for a single training example. The gradient with respect to the weights is
formed of two terms corresponding to the ones above. They are usually known as the positive gradient and the negative
gradient, because of their respective signs. In this implementation, the gradients are estimated over mini-batches of
samples.
In maximizing the log-likelihood, the positive gradient makes the model prefer hidden states that are compatible with
the observed training data. Because of the bipartite structure of RBMs, it can be computed efficiently. The negative
gradient, however, is intractable. Its goal is to lower the energy of joint states that the model prefers, therefore making
it stay true to the data. It can be approximated by Markov chain Monte Carlo using block Gibbs sampling by iteratively
sampling each of 𝑣 and ℎ given the other, until the chain mixes. Samples generated in this way are sometimes referred
as fantasy particles. This is inefficient and it is difficult to determine whether the Markov chain mixes.
The Contrastive Divergence method suggests to stop the chain after a small number of iterations, 𝑘, usually even 1.
This method is fast and has low variance, but the samples are far from the model distribution.
Persistent Contrastive Divergence addresses this. Instead of starting a new chain each time the gradient is needed, and
performing only one Gibbs sampling step, in PCD we keep a number of chains (fantasy particles) that are updated 𝑘
Gibbs steps after each weight update. This allows the particles to explore the space more thoroughly.
References:
• “A fast learning algorithm for deep belief nets” G. Hinton, S. Osindero, Y.-W. Teh, 2006
• “Training Restricted Boltzmann Machines using Approximations to the Likelihood Gradient” T. Tieleman,
2008
3.3 Model selection and evaluation
3.3.1 Cross-validation: evaluating estimator performance
Learning the parameters of a prediction function and testing it on the same data is a methodological mistake: a model
that would just repeat the labels of the samples that it has just seen would have a perfect score but would fail to predict
anything useful on yet-unseen data. This situation is called overfitting. To avoid it, it is common practice when
performing a (supervised) machine learning experiment to hold out part of the available data as a test set X_test,
y_test. Note that the word “experiment” is not intended to denote academic use only, because even in commercial
settings machine learning usually starts out experimentally. Here is a flowchart of typical cross validation workflow in
model training. The best parameters can be determined by grid search techniques.

In scikit-learn a random split into training and test sets can be quickly computed with the train_test_split
helper function. Let’s load the iris data set to fit a linear support vector machine on it:

>>> from sklearn.model_selection import train_test_split

>>> iris.data.shape, iris.target.shape
((150, 4), (150,))
We can now quickly sample a training set while holding out 40% of the data for testing (evaluating) our classifier:
>>> X_train, X_test, y_train, y_test = train_test_split(

... iris.data, iris.target, test_size=0.4, random_state=0)
>>> X_train.shape, y_train.shape

((90, 4), (90,))
>>> X_test.shape, y_test.shape
((60, 4), (60,))
>>> clf = svm.SVC(kernel='linear', C=1).fit(X_train, y_train)

0.96...
When evaluating different settings (“hyperparameters”) for estimators, such as the C setting that must be manually set
for an SVM, there is still a risk of overfitting on the test set because the parameters can be tweaked until the estimator
performs optimally. This way, knowledge about the test set can “leak” into the model and evaluation metrics no longer
report on generalization performance. To solve this problem, yet another part of the dataset can be held out as a so-
called “validation set”: training proceeds on the training set, after which evaluation is done on the validation set, and
when the experiment seems to be successful, final evaluation can be done on the test set.
However, by partitioning the available data into three sets, we drastically reduce the number of samples which can be
used for learning the model, and the results can depend on a particular random choice for the pair of (train, validation)
sets.
A solution to this problem is a procedure called cross-validation (CV for short). A test set should still be held out for
final evaluation, but the validation set is no longer needed when doing CV. In the basic approach, called k-fold CV, the
training set is split into k smaller sets (other approaches are described below, but generally follow the same principles).
The following procedure is followed for each of the k “folds”:
3.3. Model selection and evaluation 403

• A model is trained using 𝑘 − 1 of the folds as training data;

• the resulting model is validated on the remaining part of the data (i.e., it is used as a test set to compute a
performance measure such as accuracy).
The performance measure reported by k-fold cross-validation is then the average of the values computed in the loop.
This approach can be computationally expensive, but does not waste too much data (as is the case when fixing an
arbitrary validation set), which is a major advantage in problems such as inverse inference where the number of
samples is very small.
Computing cross-validated metrics
The simplest way to use cross-validation is to call the cross_val_score helper function on the estimator and the
dataset.
The following example demonstrates how to estimate the accuracy of a linear kernel support vector machine on the
iris dataset by splitting the data, fitting a model and computing the score 5 consecutive times (with different splits each
time):

>>> clf = svm.SVC(kernel='linear', C=1)
>>> scores = cross_val_score(clf, iris.data, iris.target, cv=5)
>>> scores
array([0.96..., 1. ..., 0.96..., 0.96..., 1. ])
The mean score and the 95% confidence interval of the score estimate are hence given by:
>>> print("Accuracy: %0.2f (+/- %0.2f)" % (scores.mean(), scores.std() * 2))

Accuracy: 0.98 (+/- 0.03)
By default, the score computed at each CV iteration is the score method of the estimator. It is possible to change
this by using the scoring parameter:

>>> scores = cross_val_score(
... clf, iris.data, iris.target, cv=5, scoring='f1_macro')

>>> scores
array([0.96..., 1. ..., 0.96..., 0.96..., 1. ])
See The scoring parameter: defining model evaluation rules for details. In the case of the Iris dataset, the samples are
balanced across target classes hence the accuracy and the F1-score are almost equal.
When the cv argument is an integer, cross_val_score uses the KFold or StratifiedKFold strategies by
default, the latter being used if the estimator derives from ClassifierMixin.
It is also possible to use other cross validation strategies by passing a cross validation iterator instead, for instance:
>>> from sklearn.model_selection import ShuffleSplit

>>> n_samples = iris.data.shape[0]
>>> cv = ShuffleSplit(n_splits=5, test_size=0.3, random_state=0)
>>> cross_val_score(clf, iris.data, iris.target, cv=cv)
array([0.977..., 0.977..., 1. ..., 0.955..., 1. ])
Another option is to use an iterable yielding (train, test) splits as arrays of indices, for example:
>>> def custom_cv_2folds(X):

... n = X.shape[0]
... i = 1
... while i <= 2:
... idx = np.arange(n * (i - 1) / 2, n * i / 2, dtype=int)
... yield idx, idx
... i += 1
...
>>> custom_cv = custom_cv_2folds(iris.data)
>>> cross_val_score(clf, iris.data, iris.target, cv=custom_cv)
array([1. , 0.973...])
Data transformation with held out data
Just as it is important to test a predictor on data held-out from training, preprocessing (such as standardization,
feature selection, etc.) and similar data transformations similarly should be learnt from a training set and applied
to held-out data for prediction:
>>> from sklearn import preprocessing
... iris.data, iris.target, test_size=0.4, random_state=0)
>>> scaler = preprocessing.StandardScaler().fit(X_train)
>>> X_train_transformed = scaler.transform(X_train)
>>> clf = svm.SVC(C=1).fit(X_train_transformed, y_train)
>>> X_test_transformed = scaler.transform(X_test)
>>> clf.score(X_test_transformed, y_test)
0.9333...
A Pipeline makes it easier to compose estimators, providing this behavior under cross-validation:
>>> from sklearn.pipeline import make_pipeline
>>> clf = make_pipeline(preprocessing.StandardScaler(), svm.SVC(C=1))
>>> cross_val_score(clf, iris.data, iris.target, cv=cv)
...
array([0.977..., 0.933..., 0.955..., 0.933..., 0.977...])
See Pipelines and composite estimators.

The cross_validate function and multiple metric evaluation
The cross_validate function differs from cross_val_score in two ways -

• It allows specifying multiple metrics for evaluation.
• It returns a dict containing fit-times, score-times (and optionally training scores as well as fitted estimators) in
addition to the test score.
For single metric evaluation, where the scoring parameter is a string, callable or None, the keys will be -
['test_score', 'fit_time', 'score_time']
And for multiple metric evaluation, the return value is a dict with the following keys -
['test_<scorer1_name>', 'test_<scorer2_name>', 'test_<scorer...>', 'fit_time',
'score_time']
return_train_score is set to True by default. It adds train score keys for all the scorers. If train scores are not
needed, this should be set to False explicitly.
You may also retain the estimator fitted on each training set by setting return_estimator=True.
The multiple metrics can be specified either as a list, tuple or set of predefined scorer names:
>>> from sklearn.model_selection import cross_validate

>>> from sklearn.metrics import recall_score
>>> scoring = ['precision_macro', 'recall_macro']
>>> clf = svm.SVC(kernel='linear', C=1, random_state=0)
>>> scores = cross_validate(clf, iris.data, iris.target, scoring=scoring,
... cv=5, return_train_score=False)
>>> sorted(scores.keys())
['fit_time', 'score_time', 'test_precision_macro', 'test_recall_macro']
>>> scores['test_recall_macro']
array([0.96..., 1. ..., 0.96..., 0.96..., 1. ])
Or as a dict mapping scorer name to a predefined or custom scoring function:
>>> from sklearn.metrics.scorer import make_scorer

>>> scoring = {'prec_macro': 'precision_macro',
... 'rec_micro': make_scorer(recall_score, average='macro')}
>>> scores = cross_validate(clf, iris.data, iris.target, scoring=scoring,
... cv=5, return_train_score=True)
['fit_time', 'score_time', 'test_prec_macro', 'test_rec_micro',
'train_prec_macro', 'train_rec_micro']
>>> scores['train_rec_micro']
array([0.97..., 0.97..., 0.99..., 0.98..., 0.98...])
Here is an example of cross_validate using a single metric:
>>> scores = cross_validate(clf, iris.data, iris.target,

... scoring='precision_macro', cv=5,
... return_estimator=True)
['estimator', 'fit_time', 'score_time', 'test_score', 'train_score']

Obtaining predictions by cross-validation
The function cross_val_predict has a similar interface to cross_val_score, but returns, for each element
in the input, the prediction that was obtained for that element when it was in the test set. Only cross-validation
strategies that assign all elements to a test set exactly once can be used (otherwise, an exception is raised).
Warning: Note on inappropriate usage of cross_val_predict

The result of cross_val_predict may be different from those obtained using cross_val_score as the
elements are grouped in different ways. The function cross_val_score takes an average over cross-validation
folds, whereas cross_val_predict simply returns the labels (or probabilities) from several distinct models
undistinguished. Thus, cross_val_predict is not an appropriate measure of generalisation error.
The function cross_val_predict is appropriate for:

• Visualization of predictions obtained from different models.
• Model blending: When predictions of one supervised estimator are used to train another estimator in
ensemble methods.
The available cross validation iterators are introduced in the following section.
Examples
• Receiver Operating Characteristic (ROC) with cross validation,

• Recursive feature elimination with cross-validation,
• Parameter estimation using grid search with cross-validation,
• Sample pipeline for text feature extraction and evaluation,
• Plotting Cross-Validated Predictions,
• Nested versus non-nested cross-validation.
Cross validation iterators
The following sections list utilities to generate indices that can be used to generate dataset splits according to different
cross validation strategies.
Cross-validation iterators for i.i.d. data
Assuming that some data is Independent and Identically Distributed (i.i.d.) is making the assumption that all samples
stem from the same generative process and that the generative process is assumed to have no memory of past generated
samples.
The following cross-validators can be used in such cases.
NOTE
While i.i.d. data is a common assumption in machine learning theory, it rarely holds in practice. If one knows that
the samples have been generated using a time-dependent process, it’s safer to use a time-series aware cross-validation
scheme Similarly if we know that the generative process has a group structure (samples from collected from different
subjects, experiments, measurement devices) it safer to use group-wise cross-validation.

K-fold
KFold divides all the samples in 𝑘 groups of samples, called folds (if 𝑘 = 𝑛, this is equivalent to the Leave One Out
strategy), of equal sizes (if possible). The prediction function is learned using 𝑘 − 1 folds, and the fold left out is used
for test.
Example of 2-fold cross-validation on a dataset with 4 samples:

>>> from sklearn.model_selection import KFold
>>> X = ["a", "b", "c", "d"]

>>> kf = KFold(n_splits=2)
>>> for train, test in kf.split(X):
... print("%s %s" % (train, test))
[2 3] [0 1]
[0 1] [2 3]
Here is a visualization of the cross-validation behavior. Note that KFold is not affected by classes or groups.
Each fold is constituted by two arrays: the first one is related to the training set, and the second one to the test set.
Thus, one can create the training/test sets using numpy indexing:
>>> X = np.array([[0., 0.], [1., 1.], [-1., -1.], [2., 2.]])

>>> y = np.array([0, 1, 0, 1])
>>> X_train, X_test, y_train, y_test = X[train], X[test], y[train], y[test]
Repeated K-Fold
RepeatedKFold repeats K-Fold n times. It can be used when one requires to run KFold n times, producing
different splits in each repetition.
Example of 2-fold K-Fold repeated 2 times:

>>> from sklearn.model_selection import RepeatedKFold
>>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4]])
>>> random_state = 12883823
>>> rkf = RepeatedKFold(n_splits=2, n_repeats=2, random_state=random_state)
>>> for train, test in rkf.split(X):


...
[2 3] [0 1]
[0 1] [2 3]
[0 2] [1 3]
[1 3] [0 2]
Similarly, RepeatedStratifiedKFold repeats Stratified K-Fold n times with different randomization in each
repetition.
Leave One Out (LOO)
LeaveOneOut (or LOO) is a simple cross-validation. Each learning set is created by taking all the samples except
one, the test set being the sample left out. Thus, for 𝑛 samples, we have 𝑛 different training sets and 𝑛 different tests
set. This cross-validation procedure does not waste much data as only one sample is removed from the training set:
>>> from sklearn.model_selection import LeaveOneOut
>>> X = [1, 2, 3, 4]
>>> loo = LeaveOneOut()
>>> for train, test in loo.split(X):
[1 2 3] [0]
[0 2 3] [1]
[0 1 3] [2]
[0 1 2] [3]
Potential users of LOO for model selection should weigh a few known caveats. When compared with 𝑘-fold cross
validation, one builds 𝑛 models from 𝑛 samples instead of 𝑘 models, where 𝑛 > 𝑘. Moreover, each is trained on 𝑛 − 1
samples rather than (𝑘 − 1)𝑛/𝑘. In both ways, assuming 𝑘 is not too large and 𝑘 < 𝑛, LOO is more computationally
expensive than 𝑘-fold cross validation.
In terms of accuracy, LOO often results in high variance as an estimator for the test error. Intuitively, since 𝑛 − 1 of
the 𝑛 samples are used to build each model, models constructed from folds are virtually identical to each other and to
the model built from the entire training set.
However, if the learning curve is steep for the training size in question, then 5- or 10- fold cross validation can
overestimate the generalization error.
As a general rule, most authors, and empirical evidence, suggest that 5- or 10- fold cross validation should be preferred
to LOO.
References:
• http://www.faqs.org/faqs/ai-faq/neural-nets/part3/section-12.html;
• T. Hastie, R. Tibshirani, J. Friedman, The Elements of Statistical Learning, Springer 2009
• L. Breiman, P. Spector Submodel selection and evaluation in regression: The X-random case, International
Statistical Review 1992;
• R. Kohavi, A Study of Cross-Validation and Bootstrap for Accuracy Estimation and Model Selection, Intl.
Jnt. Conf. AI
• R. Bharat Rao, G. Fung, R. Rosales, On the Dangers of Cross-Validation. An Experimental Evaluation, SIAM
2008;

• G. James, D. Witten, T. Hastie, R Tibshirani, An Introduction to Statistical Learning, Springer 2013.
Leave P Out (LPO)
LeavePOut is very similar to LeaveOneOut as it(︀creates all the possible training/test sets by removing 𝑝 samples
from the complete set. For 𝑛 samples, this produces 𝑛𝑝 train-test pairs. Unlike LeaveOneOut and KFold, the test
)︀
sets will overlap for 𝑝 > 1.

Example of Leave-2-Out on a dataset with 4 samples:
>>> from sklearn.model_selection import LeavePOut
>>> X = np.ones(4)
>>> lpo = LeavePOut(p=2)
>>> for train, test in lpo.split(X):
[2 3] [0 1]
[1 3] [0 2]
[1 2] [0 3]
[0 3] [1 2]
[0 2] [1 3]
[0 1] [2 3]
Random permutations cross-validation a.k.a. Shuffle & Split
ShuffleSplit
The ShuffleSplit iterator will generate a user defined number of independent train / test dataset splits. Samples
are first shuffled and then split into a pair of train and test sets.
It is possible to control the randomness for reproducibility of the results by explicitly seeding the random_state
pseudo random number generator.
Here is a usage example:

>>> X = np.arange(10)
>>> ss = ShuffleSplit(n_splits=5, test_size=0.25,
... random_state=0)
>>> for train_index, test_index in ss.split(X):
... print("%s %s" % (train_index, test_index))
[9 1 6 7 3 0 5] [2 8 4]
[2 9 8 0 6 7 4] [3 5 1]
[4 5 1 0 6 9 7] [2 3 8]
[2 7 5 8 0 3 4] [6 1 9]
[4 1 0 6 8 9 3] [5 2 7]
Here is a visualization of the cross-validation behavior. Note that ShuffleSplit is not affected by classes or
groups.
ShuffleSplit is thus a good alternative to KFold cross validation that allows a finer control on the number of
iterations and the proportion of samples on each side of the train / test split.

Cross-validation iterators with stratification based on class labels.
Some classification problems can exhibit a large imbalance in the distribution of the target classes: for instance there
could be several times more negative samples than positive samples. In such cases it is recommended to use stratified
sampling as implemented in StratifiedKFold and StratifiedShuffleSplit to ensure that relative class
frequencies is approximately preserved in each train and validation fold.
Stratified k-fold
StratifiedKFold is a variation of k-fold which returns stratified folds: each set contains approximately the same
percentage of samples of each target class as the complete set.
Example of stratified 3-fold cross-validation on a dataset with 10 samples from two slightly unbalanced classes:
>>> from sklearn.model_selection import StratifiedKFold
>>> X = np.ones(10)
>>> y = [0, 0, 0, 0, 1, 1, 1, 1, 1, 1]
>>> skf = StratifiedKFold(n_splits=3)
>>> for train, test in skf.split(X, y):
[2 3 6 7 8 9] [0 1 4 5]
[0 1 3 4 5 8 9] [2 6 7]
[0 1 2 4 5 6 7] [3 8 9]
Here is a visualization of the cross-validation behavior.

RepeatedStratifiedKFold can be used to repeat Stratified K-Fold n times with different randomization in each
repetition.
Stratified Shuffle Split
StratifiedShuffleSplit is a variation of ShuffleSplit, which returns stratified splits, i.e which creates splits
by preserving the same percentage for each target class as in the complete set.


Cross-validation iterators for grouped data.
The i.i.d. assumption is broken if the underlying generative process yield groups of dependent samples.
Such a grouping of data is domain specific. An example would be when there is medical data collected from multiple
patients, with multiple samples taken from each patient. And such data is likely to be dependent on the individual
group. In our example, the patient id for each sample will be its group identifier.
In this case we would like to know if a model trained on a particular set of groups generalizes well to the unseen
groups. To measure this, we need to ensure that all the samples in the validation fold come from groups that are not
represented at all in the paired training fold.
The following cross-validation splitters can be used to do that. The grouping identifier for the samples is specified via
the groups parameter.
Group k-fold
GroupKFold is a variation of k-fold which ensures that the same group is not represented in both testing and training
sets. For example if the data is obtained from different subjects with several samples per-subject and if the model is
flexible enough to learn from highly person specific features it could fail to generalize to new subjects. GroupKFold
makes it possible to detect this kind of overfitting situations.
Imagine you have three subjects, each with an associated number from 1 to 3:
>>> from sklearn.model_selection import GroupKFold
>>> X = [0.1, 0.2, 2.2, 2.4, 2.3, 4.55, 5.8, 8.8, 9, 10]
>>> y = ["a", "b", "b", "b", "c", "c", "c", "d", "d", "d"]
>>> groups = [1, 1, 1, 2, 2, 2, 3, 3, 3, 3]
>>> gkf = GroupKFold(n_splits=3)

>>> for train, test in gkf.split(X, y, groups=groups):
[0 1 2 3 4 5] [6 7 8 9]
[0 1 2 6 7 8 9] [3 4 5]
[3 4 5 6 7 8 9] [0 1 2]
Each subject is in a different testing fold, and the same subject is never in both testing and training. Notice that the
folds do not have exactly the same size due to the imbalance in the data.

Leave One Group Out
LeaveOneGroupOut is a cross-validation scheme which holds out the samples according to a third-party provided
array of integer groups. This group information can be used to encode arbitrary domain specific pre-defined cross-
validation folds.
Each training set is thus constituted by all the samples except the ones related to a specific group.
For example, in the cases of multiple experiments, LeaveOneGroupOut can be used to create a cross-validation
based on the different experiments: we create a training set using the samples of all the experiments except one:
>>> from sklearn.model_selection import LeaveOneGroupOut
>>> X = [1, 5, 10, 50, 60, 70, 80]

>>> y = [0, 1, 1, 2, 2, 2, 2]
>>> groups = [1, 1, 2, 2, 3, 3, 3]
>>> logo = LeaveOneGroupOut()
>>> for train, test in logo.split(X, y, groups=groups):
[2 3 4 5 6] [0 1]
[0 1 4 5 6] [2 3]
[0 1 2 3] [4 5 6]
Another common application is to use time information: for instance the groups could be the year of collection of the
samples and thus allow for cross-validation against time-based splits.
Leave P Groups Out
LeavePGroupsOut is similar as LeaveOneGroupOut, but removes samples related to 𝑃 groups for each train-
ing/test set.
Example of Leave-2-Group Out:
>>> from sklearn.model_selection import LeavePGroupsOut
>>> X = np.arange(6)
>>> y = [1, 1, 1, 2, 2, 2]
>>> groups = [1, 1, 2, 2, 3, 3]
>>> lpgo = LeavePGroupsOut(n_groups=2)
>>> for train, test in lpgo.split(X, y, groups=groups):
[4 5] [0 1 2 3]
[2 3] [0 1 4 5]
[0 1] [2 3 4 5]
Group Shuffle Split
The GroupShuffleSplit iterator behaves as a combination of ShuffleSplit and LeavePGroupsOut, and

generates a sequence of randomized partitions in which a subset of groups are held out for each split.
Here is a usage example:
>>> from sklearn.model_selection import GroupShuffleSplit
>>> X = [0.1, 0.2, 2.2, 2.4, 2.3, 4.55, 5.8, 0.001]

>>> y = ["a", "b", "b", "b", "c", "c", "c", "a"]

>>> groups = [1, 1, 2, 2, 3, 3, 4, 4]

>>> gss = GroupShuffleSplit(n_splits=4, test_size=0.5, random_state=0)
>>> for train, test in gss.split(X, y, groups=groups):
...
[0 1 2 3] [4 5 6 7]
[2 3 6 7] [0 1 4 5]
[2 3 4 5] [0 1 6 7]
[4 5 6 7] [0 1 2 3]
This class is useful when the behavior of LeavePGroupsOut is desired, but the number of groups is large enough
that generating all possible partitions with 𝑃 groups withheld would be prohibitively expensive. In such a sce-
nario, GroupShuffleSplit provides a random sample (with replacement) of the train / test splits generated by
LeavePGroupsOut.
Predefined Fold-Splits / Validation-Sets
For some datasets, a pre-defined split of the data into training- and validation fold or into several cross-validation folds
already exists. Using PredefinedSplit it is possible to use these folds e.g. when searching for hyperparameters.
For example, when using a validation set, set the test_fold to 0 for all samples that are part of the validation set,
and to -1 for all other samples.
Cross validation of time series data
Time series data is characterised by the correlation between observations that are near in time (autocorrelation). How-
ever, classical cross-validation techniques such as KFold and ShuffleSplit assume the samples are independent
and identically distributed, and would result in unreasonable correlation between training and testing instances (yield-
ing poor estimates of generalisation error) on time series data. Therefore, it is very important to evaluate our model
for time series data on the “future” observations least like those that are used to train the model. To achieve this, one
solution is provided by TimeSeriesSplit.

Time Series Split
TimeSeriesSplit is a variation of k-fold which returns first 𝑘 folds as train set and the (𝑘 + 1) th fold as test set.
Note that unlike standard cross-validation methods, successive training sets are supersets of those that come before
them. Also, it adds all surplus data to the first training partition, which is always used to train the model.
This class can be used to cross-validate time series data samples that are observed at fixed time intervals.
Example of 3-split time series cross-validation on a dataset with 6 samples:
>>> from sklearn.model_selection import TimeSeriesSplit
>>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4], [1, 2], [3, 4]])
>>> y = np.array([1, 2, 3, 4, 5, 6])
>>> tscv = TimeSeriesSplit(n_splits=3)
>>> print(tscv)
TimeSeriesSplit(max_train_size=None, n_splits=3)
>>> for train, test in tscv.split(X):
[0 1 2] [3]
[0 1 2 3] [4]
[0 1 2 3 4] [5]
A note on shuffling
If the data ordering is not arbitrary (e.g. samples with the same class label are contiguous), shuffling it first may
be essential to get a meaningful cross- validation result. However, the opposite may be true if the samples are not
independently and identically distributed. For example, if samples correspond to news articles, and are ordered by
their time of publication, then shuffling the data will likely lead to a model that is overfit and an inflated validation
score: it will be tested on samples that are artificially similar (close in time) to training samples.
Some cross validation iterators, such as KFold, have an inbuilt option to shuffle the data indices before splitting them.
Note that:
• This consumes less memory than shuffling the data directly.
• By default no shuffling occurs, including for the (stratified) K fold cross- validation performed by specifying
cv=some_integer to cross_val_score, grid search, etc. Keep in mind that train_test_split
still returns a random split.

• The random_state parameter defaults to None, meaning that the shuffling will be different every time
KFold(..., shuffle=True) is iterated. However, GridSearchCV will use the same shuffling for
each set of parameters validated by a single call to its fit method.
• To get identical results for each split, set random_state to an integer.
Cross validation and model selection
Cross validation iterators can also be used to directly perform model selection using Grid Search for the optimal
hyperparameters of the model. This is the topic of the next section: Tuning the hyper-parameters of an estimator.
3.3.2 Tuning the hyper-parameters of an estimator
Hyper-parameters are parameters that are not directly learnt within estimators. In scikit-learn they are passed as
arguments to the constructor of the estimator classes. Typical examples include C, kernel and gamma for Support
Vector Classifier, alpha for Lasso, etc.
It is possible and recommended to search the hyper-parameter space for the best cross validation score.
Any parameter provided when constructing an estimator may be optimized in this manner. Specifically, to find the
names and current values for all parameters for a given estimator, use:
estimator.get_params()
A search consists of:

• an estimator (regressor or classifier such as sklearn.svm.SVC());
• a parameter space;
• a method for searching or sampling candidates;
• a cross-validation scheme; and
• a score function.
Some models allow for specialized, efficient parameter search strategies, outlined below. Two generic approaches to
sampling search candidates are provided in scikit-learn: for given values, GridSearchCV exhaustively considers all
parameter combinations, while RandomizedSearchCV can sample a given number of candidates from a parameter
space with a specified distribution. After describing these tools we detail best practice applicable to both approaches.
Note that it is common that a small subset of those parameters can have a large impact on the predictive or computation
performance of the model while others can be left to their default values. It is recommended to read the docstring of
the estimator class to get a finer understanding of their expected behavior, possibly by reading the enclosed reference
to the literature.
Exhaustive Grid Search
The grid search provided by GridSearchCV exhaustively generates candidates from a grid of parameter values
specified with the param_grid parameter. For instance, the following param_grid:
param_grid = [
{'C': [1, 10, 100, 1000], 'kernel': ['linear']},
{'C': [1, 10, 100, 1000], 'gamma': [0.001, 0.0001], 'kernel': ['rbf']},
]

specifies that two grids should be explored: one with a linear kernel and C values in [1, 10, 100, 1000], and the second
one with an RBF kernel, and the cross-product of C values ranging in [1, 10, 100, 1000] and gamma values in [0.001,
0.0001].
The GridSearchCV instance implements the usual estimator API: when “fitting” it on a dataset all the possible
combinations of parameter values are evaluated and the best combination is retained.
Examples:
• See Parameter estimation using grid search with cross-validation for an example of Grid Search computation
on the digits dataset.
• See Sample pipeline for text feature extraction and evaluation for an example of Grid Search coupling pa-
rameters from a text documents feature extractor (n-gram count vectorizer and TF-IDF transformer) with a
classifier (here a linear SVM trained with SGD with either elastic net or L2 penalty) using a pipeline.
Pipeline instance.
• See Nested versus non-nested cross-validation for an example of Grid Search within a cross validation loop
on the iris dataset. This is the best practice for evaluating the performance of a model with grid search.
• See Demonstration of multi-metric evaluation on cross_val_score and GridSearchCV for an example of
GridSearchCV being used to evaluate multiple metrics simultaneously.
Randomized Parameter Optimization
While using a grid of parameter settings is currently the most widely used method for parameter optimization, other
search methods have more favourable properties. RandomizedSearchCV implements a randomized search over
parameters, where each setting is sampled from a distribution over possible parameter values. This has two main
benefits over an exhaustive search:
• A budget can be chosen independent of the number of parameters and possible values.
• Adding parameters that do not influence the performance does not decrease efficiency.
Specifying how parameters should be sampled is done using a dictionary, very similar to specifying parameters for
GridSearchCV . Additionally, a computation budget, being the number of sampled candidates or sampling itera-
tions, is specified using the n_iter parameter. For each parameter, either a distribution over possible values or a list
of discrete choices (which will be sampled uniformly) can be specified:
{'C': scipy.stats.expon(scale=100), 'gamma': scipy.stats.expon(scale=.1),

'kernel': ['rbf'], 'class_weight':['balanced', None]}
This example uses the scipy.stats module, which contains many useful distributions for sampling parameters,
such as expon, gamma, uniform or randint. In principle, any function can be passed that provides a rvs
(random variate sample) method to sample a value. A call to the rvs function should provide independent random
samples from possible parameter values on consecutive calls.
Warning: The distributions in scipy.stats prior to version scipy 0.16 do not allow specifying a
random state. Instead, they use the global numpy random state, that can be seeded via np.random.
seed or set using np.random.set_state. However, beginning scikit-learn 0.18, the sklearn.
model_selection module sets the random state provided by the user if scipy >= 0.16 is also
available.

For continuous parameters, such as C above, it is important to specify a continuous distribution to take full advantage
of the randomization. This way, increasing n_iter will always lead to a finer search.
Examples:
• Comparing randomized search and grid search for hyperparameter estimation compares the usage and effi-
ciency of randomized search and grid search.
References:
• Bergstra, J. and Bengio, Y., Random search for hyper-parameter optimization, The Journal of Machine Learn-
ing Research (2012)
Tips for parameter search
Specifying an objective metric
By default, parameter search uses the score function of the estimator to evaluate a parameter setting. These are
the sklearn.metrics.accuracy_score for classification and sklearn.metrics.r2_score for regres-
sion. For some applications, other scoring functions are better suited (for example in unbalanced classification, the
accuracy score is often uninformative). An alternative scoring function can be specified via the scoring parameter
to GridSearchCV , RandomizedSearchCV and many of the specialized cross-validation tools described below.
See The scoring parameter: defining model evaluation rules for more details.
Specifying multiple metrics for evaluation
GridSearchCV and RandomizedSearchCV allow specifying multiple metrics for the scoring parameter.
Multimetric scoring can either be specified as a list of strings of predefined scores names or a dict mapping the scorer
name to the scorer function and/or the predefined scorer name(s). See Using multiple metric evaluation for more
details.
When specifying multiple metrics, the refit parameter must be set to the metric (string) for which the
best_params_ will be found and used to build the best_estimator_ on the whole dataset. If the search
should not be refit, set refit=False. Leaving refit to the default value None will result in an error when using
multiple metrics.
See Demonstration of multi-metric evaluation on cross_val_score and GridSearchCV for an example usage.
Composite estimators and parameter spaces
Pipeline: chaining estimators describes building composite estimators whose parameter space can be searched with
these tools.
Model selection: development and evaluation
Model selection by evaluating various parameter settings can be seen as a way to use the labeled data to “train” the
parameters of the grid.

When evaluating the resulting model it is important to do it on held-out samples that were not seen during the grid
search process: it is recommended to split the data into a development set (to be fed to the GridSearchCV instance)
and an evaluation set to compute performance metrics.
This can be done by using the train_test_split utility function.
Parallelism
GridSearchCV and RandomizedSearchCV evaluate each parameter setting independently. Computations can
be run in parallel if your OS supports it, by using the keyword n_jobs=-1. See function signature for more details.
Robustness to failure
Some parameter settings may result in a failure to fit one or more folds of the data. By default, this will cause
the entire search to fail, even if some parameter settings could be fully evaluated. Setting error_score=0 (or
=np.NaN) will make the procedure robust to such failure, issuing a warning and setting the score for that fold to 0 (or
NaN), but completing the search.
Alternatives to brute force parameter search
Model specific cross-validation
Some models can fit data for a range of values of some parameter almost as efficiently as fitting the estimator for
a single value of the parameter. This feature can be leveraged to perform a more efficient cross-validation used for
model selection of this parameter.
The most common parameter amenable to this strategy is the parameter encoding the strength of the regularizer. In
this case we say that we compute the regularization path of the estimator.
Here is the list of such models:
linear_model.ElasticNetCV ([l1_ratio, eps, . . . ]) Elastic Net model with iterative fitting along a regulariza-
tion path.
linear_model.LarsCV ([fit_intercept, . . . ]) Cross-validated Least Angle Regression model.
linear_model.LassoCV ([eps, n_alphas, . . . ]) Lasso linear model with iterative fitting along a regulariza-
tion path.
linear_model.LassoLarsCV ([fit_intercept, . . . ]) Cross-validated Lasso, using the LARS algorithm.
linear_model.LogisticRegressionCV ([Cs, Logistic Regression CV (aka logit, MaxEnt) classifier.
. . . ])
linear_model.MultiTaskElasticNetCV ([. . . ])
Multi-task L1/L2 ElasticNet with built-in cross-validation.
linear_model.MultiTaskLassoCV ([eps, . . . ])
Multi-task Lasso model trained with L1/L2 mixed-norm as
regularizer.
linear_model.OrthogonalMatchingPursuitCV ([.Cross-validated
. . ]) Orthogonal Matching Pursuit model
(OMP).
linear_model.RidgeCV ([alphas, . . . ]) Ridge regression with built-in cross-validation.
linear_model.RidgeClassifierCV ([alphas, Ridge classifier with built-in cross-validation.
. . . ])

sklearn.linear_model.ElasticNetCV
class sklearn.linear_model.ElasticNetCV(l1_ratio=0.5, eps=0.001, n_alphas=100, al-

phas=None, fit_intercept=True, normalize=False,
precompute=’auto’, max_iter=1000, tol=0.0001,
cv=’warn’, copy_X=True, verbose=0, n_jobs=None,
positive=False, random_state=None, selec-
tion=’cyclic’)
Elastic Net model with iterative fitting along a regularization path.
See glossary entry for cross-validation estimator.
Read more in the User Guide.
Parameters
l1_ratio [float or array of floats, optional] float between 0 and 1 passed to ElasticNet (scal-
ing between l1 and l2 penalties). For l1_ratio = 0 the penalty is an L2 penalty. For
l1_ratio = 1 it is an L1 penalty. For 0 < l1_ratio < 1, the penalty is a com-
bination of L1 and L2 This parameter can be a list, in which case the different values are
tested by cross-validation and the one giving the best prediction score is used. Note that a
good choice of list of values for l1_ratio is often to put more values close to 1 (i.e. Lasso)
and less close to 0 (i.e. Ridge), as in [.1, .5, .7, .9, .95, .99, 1]
eps [float, optional] Length of the path. eps=1e-3 means that alpha_min / alpha_max
= 1e-3.
n_alphas [int, optional] Number of alphas along the regularization path, used for each l1_ratio.
alphas [numpy array, optional] List of alphas where to compute the models. If None alphas are
set automatically
fit_intercept [boolean] whether to calculate the intercept for this model. If set to false, no
intercept will be used in calculations (e.g. data is expected to be already centered).
normalize [boolean, optional, default False] This parameter is ignored when
fit_intercept is set to False. If True, the regressors X will be normalized be-
fore regression by subtracting the mean and dividing by the l2-norm. If you wish to
standardize, please use sklearn.preprocessing.StandardScaler before
calling fit on an estimator with normalize=False.
precompute [True | False | ‘auto’ | array-like] Whether to use a precomputed Gram matrix to
speed up calculations. If set to 'auto' let us decide. The Gram matrix can also be passed
as argument.
max_iter [int, optional] The maximum number of iterations
tol [float, optional] The tolerance for the optimization: if the updates are smaller than tol, the
optimization code checks the dual gap for optimality and continues until it is smaller than
tol.
cv [int, cross-validation generator or an iterable, optional] Determines the cross-validation split-
ting strategy. Possible inputs for cv are:
• None, to use the default 3-fold cross-validation,
• integer, to specify the number of folds.
• CV splitter,
• An iterable yielding (train, test) splits as arrays of indices.

For integer/None inputs, KFold is used.

Refer User Guide for the various cross-validation strategies that can be used here.
Changed in version 0.20: cv default value if None will change from 3-fold to 5-fold in
v0.22.
copy_X [boolean, optional, default True] If True, X will be copied; else, it may be overwritten.
verbose [bool or integer] Amount of verbosity.
n_jobs [int or None, optional (default=None)] Number of CPUs to use during the cross valida-
tion. None means 1 unless in a joblib.parallel_backend context. -1 means using
all processors. See Glossary for more details.
positive [bool, optional] When set to True, forces the coefficients to be positive.
random_state [int, RandomState instance or None, optional, default None] The seed of the
pseudo random number generator that selects a random feature to update. If int, ran-
dom_state is the seed used by the random number generator; If RandomState instance,
random_state is the random number generator; If None, the random number generator is
the RandomState instance used by np.random. Used when selection == ‘random’.
selection [str, default ‘cyclic’] If set to ‘random’, a random coefficient is updated every iteration
rather than looping over features sequentially by default. This (setting to ‘random’) often
leads to significantly faster convergence especially when tol is higher than 1e-4.
Attributes
alpha_ [float] The amount of penalization chosen by cross validation
l1_ratio_ [float] The compromise between l1 and l2 penalization chosen by cross validation
coef_ [array, shape (n_features,) | (n_targets, n_features)] Parameter vector (w in the cost func-
tion formula),
intercept_ [float | array, shape (n_targets, n_features)] Independent term in the decision func-
tion.
mse_path_ [array, shape (n_l1_ratio, n_alpha, n_folds)] Mean square error for the test set on
each fold, varying l1_ratio and alpha.
alphas_ [numpy array, shape (n_alphas,) or (n_l1_ratio, n_alphas)] The grid of alphas used for
fitting, for each l1_ratio.
n_iter_ [int] number of iterations run by the coordinate descent solver to reach the specified
tolerance for the optimal alpha.
See also:
enet_path, ElasticNet
Notes
For an example, see examples/linear_model/plot_lasso_model_selection.py.

To avoid unnecessary memory duplication the X argument of the fit method should be directly passed as a
Fortran-contiguous numpy array.
The parameter l1_ratio corresponds to alpha in the glmnet R package while alpha corresponds to the lambda
parameter in glmnet. More specifically, the optimization objective is:

1 / (2 * n_samples) * ||y - Xw||^2_2

+ alpha * l1_ratio * ||w||_1
+ 0.5 * alpha * (1 - l1_ratio) * ||w||^2_2
If you are interested in controlling the L1 and L2 penalty separately, keep in mind that this is equivalent to:
a * L1 + b * L2
for:
alpha = a + b and l1_ratio = a / (a + b).
Examples
>>> from sklearn.linear_model import ElasticNetCV

>>> X, y = make_regression(n_features=2, random_state=0)

>>> regr = ElasticNetCV(cv=5, random_state=0)
>>> regr.fit(X, y)
ElasticNetCV(alphas=None, copy_X=True, cv=5, eps=0.001, fit_intercept=True,
l1_ratio=0.5, max_iter=1000, n_alphas=100, n_jobs=None,
normalize=False, positive=False, precompute='auto', random_state=0,
selection='cyclic', tol=0.0001, verbose=0)
>>> print(regr.alpha_)
0.1994727942696716
>>> print(regr.intercept_)
0.398...
>>> print(regr.predict([[0, 0]]))
[0.398...]
Methods
fit(X, y) Fit linear model with coordinate descent

get_params([deep]) Get parameters for this estimator.
path(X, y[, l1_ratio, eps, n_alphas, . . . ]) Compute elastic net path with coordinate descent
predict(X) Predict using the linear model
score(X, y[, sample_weight]) Returns the coefficient of determination R^2 of the pre-
diction.
set_params(**params) Set the parameters of this estimator.
__init__(l1_ratio=0.5, eps=0.001, n_alphas=100, alphas=None, fit_intercept=True, normal-

ize=False, precompute=’auto’, max_iter=1000, tol=0.0001, cv=’warn’, copy_X=True, ver-
bose=0, n_jobs=None, positive=False, random_state=None, selection=’cyclic’)
fit(X, y)
Fit linear model with coordinate descent
Fit is on grid of alphas and best alpha estimated by cross-validation.
Parameters
X [{array-like}, shape (n_samples, n_features)] Training data. Pass directly as Fortran-

contiguous data to avoid unnecessary memory duplication. If y is mono-output, X can

be sparse.
y [array-like, shape (n_samples,) or (n_samples, n_targets)] Target values
get_params(deep=True)
Get parameters for this estimator.
Parameters
deep [boolean, optional] If True, will return the parameters for this estimator and contained
subobjects that are estimators.
Returns
params [mapping of string to any] Parameter names mapped to their values.
static path(X, y, l1_ratio=0.5, eps=0.001, n_alphas=100, alphas=None, precompute=’auto’,
Xy=None, copy_X=True, coef_init=None, verbose=False, return_n_iter=False, posi-
tive=False, check_input=True, **params)
Compute elastic net path with coordinate descent
The elastic net optimization function varies for mono and multi-outputs.
For mono-output tasks it is:
1 / (2 * n_samples) * ||y - Xw||^2_2

+ 0.5 * alpha * (1 - l1_ratio) * ||w||^2_2
For multi-output tasks it is:
(1 / (2 * n_samples)) * ||Y - XW||^Fro_2

+ alpha * l1_ratio * ||W||_21
+ 0.5 * alpha * (1 - l1_ratio) * ||W||_Fro^2
Where:
||W||_21 = \sum_i \sqrt{\sum_j w_{ij}^2}
i.e. the sum of norm of each row.

Parameters
contiguous data to avoid unnecessary memory duplication. If y is mono-output then X
can be sparse.
y [ndarray, shape (n_samples,) or (n_samples, n_outputs)] Target values
l1_ratio [float, optional] float between 0 and 1 passed to elastic net (scaling between l1 and
l2 penalties). l1_ratio=1 corresponds to the Lasso
eps [float] Length of the path. eps=1e-3 means that alpha_min / alpha_max =
1e-3
n_alphas [int, optional] Number of alphas along the regularization path
alphas [ndarray, optional] List of alphas where to compute the models. If None alphas are
set automatically

precompute [True | False | ‘auto’ | array-like] Whether to use a precomputed Gram matrix
to speed up calculations. If set to 'auto' let us decide. The Gram matrix can also be
passed as argument.
Xy [array-like, optional] Xy = np.dot(X.T, y) that can be precomputed. It is useful only
when the Gram matrix is precomputed.
copy_X [boolean, optional, default True] If True, X will be copied; else, it may be over-
written.
coef_init [array, shape (n_features, ) | None] The initial values of the coefficients.
return_n_iter [bool] whether to return the number of iterations or not.
positive [bool, default False] If set to True, forces coefficients to be positive. (Only allowed
when y.ndim == 1).
check_input [bool, default True] Skip input validation checks, including the Gram matrix
when provided assuming there are handled by the caller when check_input=False.
**params [kwargs] keyword arguments passed to the coordinate descent solver.
Returns
alphas [array, shape (n_alphas,)] The alphas along the path where models are computed.
coefs [array, shape (n_features, n_alphas) or (n_outputs, n_features, n_alphas)] Coefficients
along the path.
dual_gaps [array, shape (n_alphas,)] The dual gaps at the end of the optimization for each
alpha.
n_iters [array-like, shape (n_alphas,)] The number of iterations taken by the coordinate
descent optimizer to reach the specified tolerance for each alpha. (Is returned when
return_n_iter is set to True).
See also:
MultiTaskElasticNet, MultiTaskElasticNetCV , ElasticNet, ElasticNetCV
Notes
For an example, see examples/linear_model/plot_lasso_coordinate_descent_path.py.

predict(X)
Predict using the linear model
Parameters
X [array_like or sparse matrix, shape (n_samples, n_features)] Samples.
Returns
C [array, shape (n_samples,)] Returns predicted values.
score(X, y, sample_weight=None)
Returns the coefficient of determination R^2 of the prediction.
The coefficient R^2 is defined as (1 - u/v), where u is the residual sum of squares ((y_true - y_pred) **
2).sum() and v is the total sum of squares ((y_true - y_true.mean()) ** 2).sum(). The best possible score
is 1.0 and it can be negative (because the model can be arbitrarily worse). A constant model that always
predicts the expected value of y, disregarding the input features, would get a R^2 score of 0.0.

Parameters
X [array-like, shape = (n_samples, n_features)] Test samples. For some estimators this may
be a precomputed kernel matrix instead, shape = (n_samples, n_samples_fitted], where
n_samples_fitted is the number of samples used in the fitting for the estimator.
y [array-like, shape = (n_samples) or (n_samples, n_outputs)] True values for X.
sample_weight [array-like, shape = [n_samples], optional] Sample weights.
Returns
score [float] R^2 of self.predict(X) wrt. y.
set_params(**params)
Set the parameters of this estimator.
The method works on simple estimators as well as on nested objects (such as pipelines). The latter have
parameters of the form <component>__<parameter> so that it’s possible to update each component
of a nested object.
Returns
self
sklearn.linear_model.LarsCV
class sklearn.linear_model.LarsCV(fit_intercept=True, verbose=False, max_iter=500,

normalize=True, precompute=’auto’,
cv=’warn’, max_n_alphas=1000, n_jobs=None,
eps=2.220446049250313e-16, copy_X=True, positive=False)
Cross-validated Least Angle Regression model.
Parameters
verbose [boolean or integer, optional] Sets the verbosity amount
max_iter [integer, optional] Maximum number of iterations to perform.
normalize [boolean, optional, default True] This parameter is ignored when fit_intercept
is set to False. If True, the regressors X will be normalized before regression by sub-
tracting the mean and dividing by the l2-norm. If you wish to standardize, please use
sklearn.preprocessing.StandardScaler before calling fit on an estimator
with normalize=False.
speed up calculations. If set to 'auto' let us decide. The Gram matrix cannot be passed
as argument since we will use only subsets of X.
• CV splitter,


v0.22.
max_n_alphas [integer, optional] The maximum number of points on the path used to compute
the residuals in the cross-validation
eps [float, optional] The machine-precision regularization in the computation of the Cholesky
diagonal factors. Increase this for very ill-conditioned systems.
positive [boolean (default=False)] Restrict coefficients to be >= 0. Be aware that you might
want to remove fit_intercept which is set True by default.
Deprecated since version 0.20: The option is broken and deprecated. It will be removed in
v0.22.
Attributes
coef_ [array, shape (n_features,)] parameter vector (w in the formulation formula)
intercept_ [float] independent term in decision function
coef_path_ [array, shape (n_features, n_alphas)] the varying values of the coefficients along the
path
alpha_ [float] the estimated regularization parameter alpha
alphas_ [array, shape (n_alphas,)] the different values of alpha along the path
cv_alphas_ [array, shape (n_cv_alphas,)] all the values of alpha along the path for the different
folds
mse_path_ [array, shape (n_folds, n_cv_alphas)] the mean square error on left-out for each fold
along the path (alpha values given by cv_alphas)
n_iter_ [array-like or int] the number of iterations run by Lars with the optimal alpha.
See also:
lars_path, LassoLars, LassoLarsCV
Examples
>>> from sklearn.linear_model import LarsCV

>>> X, y = make_regression(n_samples=200, noise=4.0, random_state=0)
>>> reg = LarsCV(cv=5).fit(X, y)
>>> reg.score(X, y)
0.9996...
>>> reg.alpha_
0.0254...

>>> reg.predict(X[:1,])
array([154.0842...])
Methods
fit(X, y) Fit the model using X, y as training data.

diction.
__init__(fit_intercept=True, verbose=False, max_iter=500, normalize=True, precompute=’auto’,

cv=’warn’, max_n_alphas=1000, n_jobs=None, eps=2.220446049250313e-16,
copy_X=True, positive=False)
alpha
DEPRECATED: Attribute alpha is deprecated in 0.19 and will be removed in 0.21. See alpha_ instead
fit(X, y)
Fit the model using X, y as training data.
Parameters
X [array-like, shape (n_samples, n_features)] Training data.
y [array-like, shape (n_samples,)] Target values.
Returns
self [object] returns an instance of self.
Parameters
Returns
predict(X)
Parameters
Returns

Parameters
Returns
of a nested object.
Returns
self
sklearn.linear_model.LassoCV
class sklearn.linear_model.LassoCV(eps=0.001, n_alphas=100, alphas=None, fit_intercept=True,

normalize=False, precompute=’auto’, max_iter=1000,
tol=0.0001, copy_X=True, cv=’warn’, verbose=False,
n_jobs=None, positive=False, random_state=None, selec-
tion=’cyclic’)
Lasso linear model with iterative fitting along a regularization path.
The best model is selected by cross-validation.
The optimization objective for Lasso is:
(1 / (2 * n_samples)) * ||y - Xw||^2_2 + alpha * ||w||_1

Parameters
= 1e-3.
alphas [numpy array, optional] List of alphas where to compute the models. If None alphas
are set automatically
fit_intercept [boolean, default True] whether to calculate the intercept for this model. If set to
false, no intercept will be used in calculations (e.g. data is expected to be already centered).


as argument.
tol.
• CV splitter,
v0.22.
positive [bool, optional] If positive, restrict regression coefficients to be positive
Attributes
coef_ [array, shape (n_features,) | (n_targets, n_features)] parameter vector (w in the cost func-
tion formula)
intercept_ [float | array, shape (n_targets,)] independent term in decision function.
mse_path_ [array, shape (n_alphas, n_folds)] mean square error for the test set on each fold,
varying alpha
alphas_ [numpy array, shape (n_alphas,)] The grid of alphas used for fitting

dual_gap_ [ndarray, shape ()] The dual gap at the end of the optimization for the optimal alpha
(alpha_).
See also:
lars_path, lasso_path, LassoLars, Lasso, LassoLarsCV
Notes
For an example, see examples/linear_model/plot_lasso_model_selection.py.

Examples
>>> from sklearn.linear_model import LassoCV

>>> X, y = make_regression(noise=4, random_state=0)
>>> reg = LassoCV(cv=5, random_state=0).fit(X, y)
>>> reg.score(X, y)
0.9993...
array([-78.4951...])
Methods

path(X, y[, eps, n_alphas, alphas, . . . ]) Compute Lasso path with coordinate descent
diction.
__init__(eps=0.001, n_alphas=100, alphas=None, fit_intercept=True, normalize=False, precom-

pute=’auto’, max_iter=1000, tol=0.0001, copy_X=True, cv=’warn’, verbose=False,
n_jobs=None, positive=False, random_state=None, selection=’cyclic’)
fit(X, y)
Parameters
be sparse.

Parameters
Returns
static path(X, y, eps=0.001, n_alphas=100, alphas=None, precompute=’auto’, Xy=None,
copy_X=True, coef_init=None, verbose=False, return_n_iter=False, positive=False,
**params)
Compute Lasso path with coordinate descent
The Lasso optimization function varies for mono and multi-outputs.
(1 / (2 * n_samples)) * ||Y - XW||^2_Fro + alpha * ||W||_21
Where:

Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Training data. Pass directly
as Fortran-contiguous data to avoid unnecessary memory duplication. If y is mono-output
then X can be sparse.
y [ndarray, shape (n_samples,), or (n_samples, n_outputs)] Target values
eps [float, optional] Length of the path. eps=1e-3 means that alpha_min /
alpha_max = 1e-3
set automatically
passed as argument.
written.


when y.ndim == 1).
Returns
along the path.
alpha.
n_iters [array-like, shape (n_alphas,)] The number of iterations taken by the coordinate de-
scent optimizer to reach the specified tolerance for each alpha.
See also:
lars_path, Lasso, LassoLars, LassoCV , LassoLarsCV , sklearn.decomposition.
sparse_encode
Notes

Note that in certain cases, the Lars solver may be significantly faster to implement this functionality. In
particular, linear interpolation can be used to retrieve model coefficients between the values output by
lars_path
Examples
Comparing lasso_path and lars_path with interpolation:
>>> X = np.array([[1, 2, 3.1], [2.3, 5.4, 4.3]]).T

>>> y = np.array([1, 2, 3.1])
>>> # Use lasso_path to compute a coefficient path
>>> _, coef_path, _ = lasso_path(X, y, alphas=[5., 1., .5])
>>> print(coef_path)
[[0. 0. 0.46874778]
[0.2159048 0.4425765 0.23689075]]
>>> # Now use lars_path and 1D linear interpolation to compute the

>>> # same path
>>> from sklearn.linear_model import lars_path
>>> alphas, active, coef_path_lars = lars_path(X, y, method='lasso')
>>> from scipy import interpolate
>>> coef_path_continuous = interpolate.interp1d(alphas[::-1],
... coef_path_lars[:, ::-1])
>>> print(coef_path_continuous([5., 1., .5]))
[[0. 0. 0.46915237]
[0.2159048 0.4425765 0.23668876]]

predict(X)
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.linear_model.LassoCV
• Cross-validation on diabetes Dataset Exercise

• Feature selection using SelectFromModel and LassoCV
sklearn.linear_model.LassoLarsCV
class sklearn.linear_model.LassoLarsCV(fit_intercept=True, verbose=False, max_iter=500,

normalize=True, precompute=’auto’,
cv=’warn’, max_n_alphas=1000, n_jobs=None,
eps=2.220446049250313e-16, copy_X=True, posi-
tive=False)
Cross-validated Lasso, using the LARS algorithm.


Parameters
precompute [True | False | ‘auto’] Whether to use a precomputed Gram matrix to speed up
calculations. If set to 'auto' let us decide. The Gram matrix cannot be passed as argument
since we will use only subsets of X.
• CV splitter,
v0.22.
max_n_alphas [integer, optional] The maximum number of points on the path used to compute
the residuals in the cross-validation
want to remove fit_intercept which is set True by default. Under the positive restriction the
model coefficients do not converge to the ordinary-least-squares solution for small values
of alpha. Only coefficients up to the smallest alpha value (alphas_[alphas_ > 0.].
min() when fit_path=True) reached by the stepwise Lars-Lasso algorithm are typically in
congruence with the solution of the coordinate descent Lasso estimator. As a consequence
using LassoLarsCV only makes sense for problems where a sparse solution is expected
and/or reached.

Attributes
intercept_ [float] independent term in decision function.
coef_path_ [array, shape (n_features, n_alphas)] the varying values of the coefficients along the
path
alpha_ [float] the estimated regularization parameter alpha
alphas_ [array, shape (n_alphas,)] the different values of alpha along the path
cv_alphas_ [array, shape (n_cv_alphas,)] all the values of alpha along the path for the different
folds
mse_path_ [array, shape (n_folds, n_cv_alphas)] the mean square error on left-out for each fold
along the path (alpha values given by cv_alphas)
n_iter_ [array-like or int] the number of iterations run by Lars with the optimal alpha.
See also:
lars_path, LassoLars, LarsCV , LassoCV
Notes
The object solves the same problem as the LassoCV object. However, unlike the LassoCV, it find the relevant
alphas values by itself. In general, because of this property, it will be more stable. However, it is more fragile to
heavily multicollinear datasets.
It is more efficient than the LassoCV if only a small number of features are selected compared to the total
number, for instance if there are very few samples compared to the number of features.
Examples
>>> from sklearn.linear_model import LassoLarsCV

>>> X, y = make_regression(noise=4.0, random_state=0)
>>> reg = LassoLarsCV(cv=5).fit(X, y)
>>> reg.score(X, y)
0.9992...
>>> reg.alpha_
0.0484...
array([-77.8723...])
Methods

diction.

__init__(fit_intercept=True, verbose=False, max_iter=500, normalize=True, precompute=’auto’,

cv=’warn’, max_n_alphas=1000, n_jobs=None, eps=2.220446049250313e-16,
copy_X=True, positive=False)
alpha
DEPRECATED: Attribute alpha is deprecated in 0.19 and will be removed in 0.21. See alpha_ instead
fit(X, y)
Parameters
Returns
Parameters
Returns
predict(X)
Parameters
Returns
Parameters
Returns

of a nested object.
Returns
self
Examples using sklearn.linear_model.LassoLarsCV
sklearn.linear_model.LogisticRegressionCV
class sklearn.linear_model.LogisticRegressionCV(Cs=10, fit_intercept=True, cv=’warn’,

dual=False, penalty=’l2’, scor-
ing=None, solver=’lbfgs’, tol=0.0001,
max_iter=100, class_weight=None,
n_jobs=None, verbose=0, refit=True, in-
tercept_scaling=1.0, multi_class=’warn’,
random_state=None)
Logistic Regression CV (aka logit, MaxEnt) classifier.
This class implements logistic regression using liblinear, newton-cg, sag of lbfgs optimizer. The newton-cg, sag
and lbfgs solvers support only L2 regularization with primal formulation. The liblinear solver supports both L1
and L2 regularization, with a dual formulation only for the L2 penalty.
For the grid of Cs values (that are set by default to be ten values in a logarithmic scale between 1e-4 and
1e4), the best hyperparameter is selected by the cross-validator StratifiedKFold, but it can be changed using
the cv parameter. In the case of newton-cg and lbfgs solvers, we warm start along the path i.e guess the initial
coefficients of the present fit to be the coefficients got after convergence in the previous fit, so it is supposed to
be faster for high-dimensional dense data.
For a multiclass problem, the hyperparameters for each class are computed using the best scores got by doing a
one-vs-rest in parallel across all folds and classes. Hence this is not the true multinomial loss.
Parameters
Cs [list of floats | int] Each of the values in Cs describes the inverse of regularization strength.
If Cs is as an int, then a grid of Cs values are chosen in a logarithmic scale between 1e-4
and 1e4. Like in support vector machines, smaller values specify stronger regularization.
fit_intercept [bool, default: True] Specifies if a constant (a.k.a. bias or intercept) should be
added to the decision function.
cv [integer or cross-validation generator, default: None] The default cross-validation generator
used is Stratified K-Folds. If an integer is provided, then it is the number of folds used.
See the module sklearn.model_selection module for the list of possible cross-
validation objects.
v0.22.

dual [bool] Dual or primal formulation. Dual formulation is only implemented for l2 penalty
with liblinear solver. Prefer dual=False when n_samples > n_features.
penalty [str, ‘l1’ or ‘l2’] Used to specify the norm used in the penalization. The ‘newton-cg’,
‘sag’ and ‘lbfgs’ solvers support only l2 penalties.
scoring [string, callable, or None] A string (see model evaluation documentation) or a scorer
callable object / function with signature scorer(estimator, X, y). For a list of
scoring functions that can be used, look at sklearn.metrics. The default scoring op-
tion used is ‘accuracy’.
solver [str, {‘newton-cg’, ‘lbfgs’, ‘liblinear’, ‘sag’, ‘saga’}, default: ‘lbfgs’.] Algorithm to use
in the optimization problem.
• For small datasets, ‘liblinear’ is a good choice, whereas ‘sag’ and ‘saga’ are faster for
large ones.
• For multiclass problems, only ‘newton-cg’, ‘sag’, ‘saga’ and ‘lbfgs’ handle multinomial
loss; ‘liblinear’ is limited to one-versus-rest schemes.
• ‘newton-cg’, ‘lbfgs’ and ‘sag’ only handle L2 penalty, whereas ‘liblinear’ and ‘saga’
handle L1 penalty.
• ‘liblinear’ might be slower in LogisticRegressionCV because it does not handle warm-
starting.
Note that ‘sag’ and ‘saga’ fast convergence is only guaranteed on features with approxi-
mately the same scale. You can preprocess the data with a scaler from sklearn.preprocessing.
New in version 0.17: Stochastic Average Gradient descent solver.
New in version 0.19: SAGA solver.
tol [float, optional] Tolerance for stopping criteria.
max_iter [int, optional] Maximum number of iterations of the optimization algorithm.
class_weight [dict or ‘balanced’, optional] Weights associated with classes in the form
{class_label: weight}. If not given, all classes are supposed to have weight one.
The “balanced” mode uses the values of y to automatically adjust weights inversely pro-
portional to class frequencies in the input data as n_samples / (n_classes * np.
bincount(y)).
Note that these weights will be multiplied with sample_weight (passed through the fit
method) if sample_weight is specified.
New in version 0.17: class_weight == ‘balanced’
n_jobs [int or None, optional (default=None)] Number of CPU cores used during the cross-
validation loop. None means 1 unless in a joblib.parallel_backend context. -1
means using all processors. See Glossary for more details.
verbose [int] For the ‘liblinear’, ‘sag’ and ‘lbfgs’ solvers set verbose to any positive number for
verbosity.
refit [bool] If set to True, the scores are averaged across all folds, and the coefs and the C
that corresponds to the best score is taken, and a final refit is done using these parameters.
Otherwise the coefs, intercepts and C that correspond to the best scores across folds are
averaged.
intercept_scaling [float, default 1.] Useful only when the solver ‘liblinear’ is used and
self.fit_intercept is set to True. In this case, x becomes [x, self.intercept_scaling],

i.e. a “synthetic” feature with constant value equal to intercept_scaling is ap-

pended to the instance vector. The intercept becomes intercept_scaling *
synthetic_feature_weight.
Note! the synthetic feature weight is subject to l1/l2 regularization as all other features. To
lessen the effect of regularization on synthetic feature weight (and therefore on the intercept)
intercept_scaling has to be increased.
multi_class [str, {‘ovr’, ‘multinomial’, ‘auto’}, default: ‘ovr’] If the option chosen is ‘ovr’,
then a binary problem is fit for each label. For ‘multinomial’ the loss minimised is the
multinomial loss fit across the entire probability distribution, even when the data is binary.
‘multinomial’ is unavailable when solver=’liblinear’. ‘auto’ selects ‘ovr’ if the data is bi-
nary, or if solver=’liblinear’, and otherwise selects ‘multinomial’.
New in version 0.18: Stochastic Average Gradient descent solver for ‘multinomial’ case.
Changed in version 0.20: Default will change from ‘ovr’ to ‘auto’ in 0.22.
random_state [int, RandomState instance or None, optional, default None] If int, random_state
is the seed used by the random number generator; If RandomState instance, random_state is
the random number generator; If None, the random number generator is the RandomState
instance used by np.random.
Attributes
classes_ [array, shape (n_classes, )] A list of class labels known to the classifier.
coef_ [array, shape (1, n_features) or (n_classes, n_features)] Coefficient of the features in the
decision function.
coef_ is of shape (1, n_features) when the given problem is binary.
intercept_ [array, shape (1,) or (n_classes,)] Intercept (a.k.a. bias) added to the decision func-
tion.
If fit_intercept is set to False, the intercept is set to zero. intercept_ is of shape(1,) when the
problem is binary.
Cs_ [array] Array of C i.e. inverse of regularization parameter values used for cross-validation.
coefs_paths_ [array, shape (n_folds, len(Cs_), n_features) or (n_folds,
len(Cs_), n_features + 1)] dict with classes as the keys, and the path of coef-
ficients obtained during cross-validating across each fold and then across each Cs after
doing an OvR for the corresponding class as values. If the ‘multi_class’ option is set
to ‘multinomial’, then the coefs_paths are the coefficients corresponding to each class.
Each dict value has shape (n_folds, len(Cs_), n_features) or (n_folds,
len(Cs_), n_features + 1) depending on whether the intercept is fit or not.
scores_ [dict] dict with classes as the keys, and the values as the grid of scores obtained dur-
ing cross-validating each fold, after doing an OvR for the corresponding class. If the
‘multi_class’ option given is ‘multinomial’ then the same scores are repeated across all
classes, since this is the multinomial class. Each dict value has shape (n_folds, len(Cs))
C_ [array, shape (n_classes,) or (n_classes - 1,)] Array of C that maps to the best scores across
every class. If refit is set to False, then for each class, the best C is the average of the C’s that
correspond to the best scores for each fold. C_ is of shape(n_classes,) when the problem is
binary.
n_iter_ [array, shape (n_classes, n_folds, n_cs) or (1, n_folds, n_cs)] Actual number of itera-
tions for all classes, folds and Cs. In the binary or multinomial cases, the first dimension is
equal to 1.

See also:
LogisticRegression
Examples

>>> from sklearn.linear_model import LogisticRegressionCV
>>> X, y = load_iris(return_X_y=True)
>>> clf = LogisticRegressionCV(cv=5, random_state=0,
... multi_class='multinomial').fit(X, y)
>>> clf.predict(X[:2, :])
array([0, 0])
>>> clf.predict_proba(X[:2, :]).shape
(2, 3)
>>> clf.score(X, y)
0.98...
Methods
decision_function(X) Predict confidence scores for samples.

densify() Convert coefficient matrix to dense array format.
fit(X, y[, sample_weight]) Fit the model according to the given training data.
predict(X) Predict class labels for samples in X.
predict_log_proba(X) Log of probability estimates.
predict_proba(X) Probability estimates.
score(X, y[, sample_weight]) Returns the score using the scoring option on the given
test data and labels.
sparsify() Convert coefficient matrix to sparse format.
__init__(Cs=10, fit_intercept=True, cv=’warn’, dual=False, penalty=’l2’, scoring=None,

solver=’lbfgs’, tol=0.0001, max_iter=100, class_weight=None, n_jobs=None, verbose=0,
refit=True, intercept_scaling=1.0, multi_class=’warn’, random_state=None)
decision_function(X)
Predict confidence scores for samples.
The confidence score for a sample is the signed distance of that sample to the hyperplane.
Parameters
Returns
array, shape=(n_samples,) if n_classes == 2 else (n_samples, n_classes) Confidence
scores per (sample, class) combination. In the binary case, confidence score for
self.classes_[1] where >0 means this class would be predicted.
densify()
Convert coefficient matrix to dense array format.
Converts the coef_ member (back) to a numpy.ndarray. This is the default format of coef_ and is

required for fitting, so calling this method is only required on models that have previously been sparsified;
otherwise, it is a no-op.
Returns
self [estimator]
fit(X, y, sample_weight=None)
Fit the model according to the given training data.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Training vector, where
n_samples is the number of samples and n_features is the number of features.
y [array-like, shape (n_samples,)] Target vector relative to X.
sample_weight [array-like, shape (n_samples,) optional] Array of weights that are assigned
to individual samples. If not provided, then each sample is given unit weight.
Returns
self [object]
Parameters
Returns
predict(X)
Predict class labels for samples in X.
Parameters
Returns
C [array, shape [n_samples]] Predicted class label per sample.
predict_log_proba(X)
Log of probability estimates.
The returned estimates for all classes are ordered by the label of classes.
Parameters
X [array-like, shape = [n_samples, n_features]]
Returns
T [array-like, shape = [n_samples, n_classes]] Returns the log-probability of the sample for
each class in the model, where classes are ordered as they are in self.classes_.
predict_proba(X)
Probability estimates.
For a multi_class problem, if multi_class is set to be “multinomial” the softmax function is used to find
the predicted probability of each class. Else use a one-vs-rest approach, i.e calculate the probability of

each class assuming it to be positive using the logistic function. and normalize these values across all the
classes.
Parameters
Returns
T [array-like, shape = [n_samples, n_classes]] Returns the probability of the sample for each
class in the model, where classes are ordered as they are in self.classes_.
Returns the score using the scoring option on the given test data and labels.
Parameters
X [array-like, shape = (n_samples, n_features)] Test samples.
y [array-like, shape = (n_samples,)] True labels for X.
Returns
score [float] Score of self.predict(X) wrt. y.
of a nested object.
Returns
self
sparsify()
Convert coefficient matrix to sparse format.
Converts the coef_ member to a scipy.sparse matrix, which for L1-regularized models can be much more
memory- and storage-efficient than the usual numpy.ndarray representation.
The intercept_ member is not converted.
Returns
self [estimator]
Notes
For non-sparse models, i.e. when there are not many zeros in coef_, this may actually increase memory
usage, so use this method with care. A rule of thumb is that the number of zero elements, which can be
computed with (coef_ == 0).sum(), must be more than 50% for this to provide significant benefits.
After calling this method, further fitting with the partial_fit method (if any) will not work until you call
densify.

sklearn.linear_model.MultiTaskElasticNetCV
class sklearn.linear_model.MultiTaskElasticNetCV(l1_ratio=0.5, eps=0.001, n_alphas=100,

alphas=None, fit_intercept=True,
normalize=False, max_iter=1000,
tol=0.0001, cv=’warn’, copy_X=True,
verbose=0, n_jobs=None, ran-
dom_state=None, selection=’cyclic’)
Multi-task L1/L2 ElasticNet with built-in cross-validation.
The optimization objective for MultiTaskElasticNet is:
(1 / (2 * n_samples)) * ||Y - XW||^Fro_2

Where:

Parameters
l1_ratio [float or array of floats] The ElasticNet mixing parameter, with 0 < l1_ratio <= 1. For
l1_ratio = 1 the penalty is an L1/L2 penalty. For l1_ratio = 0 it is an L2 penalty. For 0 <
l1_ratio < 1, the penalty is a combination of L1/L2 and L2. This parameter can be a
list, in which case the different values are tested by cross-validation and the one giving the
best prediction score is used. Note that a good choice of list of values for l1_ratio is often
to put more values close to 1 (i.e. Lasso) and less close to 0 (i.e. Ridge), as in [.1, .5,
.7, .9, .95, .99, 1]
= 1e-3.
alphas [array-like, optional] List of alphas where to compute the models. If not provided, set
automatically.
tol.


• CV splitter,
v0.22.
n_jobs [int or None, optional (default=None)] Number of CPUs to use during the cross vali-
dation. Note that this is used only if multiple values for l1_ratio are given. None means 1
unless in a joblib.parallel_backend context. -1 means using all processors. See
Glossary for more details.
Attributes
intercept_ [array, shape (n_tasks,)] Independent term in decision function.
coef_ [array, shape (n_tasks, n_features)] Parameter vector (W in the cost function formula).
Note that coef_ stores the transpose of W, W.T.
mse_path_ [array, shape (n_alphas, n_folds) or (n_l1_ratio, n_alphas, n_folds)] mean square
error for the test set on each fold, varying alpha
alphas_ [numpy array, shape (n_alphas,) or (n_l1_ratio, n_alphas)] The grid of alphas used for
fitting, for each l1_ratio
l1_ratio_ [float] best l1_ratio obtained by cross-validation.
See also:
MultiTaskElasticNet, ElasticNetCV , MultiTaskLassoCV
Notes
The algorithm used to fit the model is coordinate descent.


Examples

>>> clf = linear_model.MultiTaskElasticNetCV(cv=3)
>>> clf.fit([[0,0], [1, 1], [2, 2]],
... [[0, 0], [1, 1], [2, 2]])
...
MultiTaskElasticNetCV(alphas=None, copy_X=True, cv=3, eps=0.001,
fit_intercept=True, l1_ratio=0.5, max_iter=1000, n_alphas=100,
n_jobs=None, normalize=False, random_state=None, selection='cyclic',
tol=0.0001, verbose=0)
>>> print(clf.coef_)
[[0.52875032 0.46958558]
[0.52875032 0.46958558]]
>>> print(clf.intercept_)
[0.00166409 0.00166409]
Methods

diction.
__init__(l1_ratio=0.5, eps=0.001, n_alphas=100, alphas=None, fit_intercept=True, nor-

malize=False, max_iter=1000, tol=0.0001, cv=’warn’, copy_X=True, verbose=0,
n_jobs=None, random_state=None, selection=’cyclic’)
fit(X, y)
Parameters
be sparse.
Parameters
Returns


1 / (2 * n_samples) * ||y - Xw||^2_2

+ 0.5 * alpha * (1 - l1_ratio) * ||w||^2_2
(1 / (2 * n_samples)) * ||Y - XW||^Fro_2

Where:

Parameters
can be sparse.
1e-3
set automatically
passed as argument.
written.
when y.ndim == 1).

Returns
along the path.
alpha.
See also:
Notes

predict(X)
Parameters
Returns
Parameters
Returns

of a nested object.
Returns
self
sklearn.linear_model.MultiTaskLassoCV
class sklearn.linear_model.MultiTaskLassoCV(eps=0.001, n_alphas=100, alphas=None,

fit_intercept=True, normalize=False,
max_iter=1000, tol=0.0001, copy_X=True,
cv=’warn’, verbose=False, n_jobs=None,
random_state=None, selection=’cyclic’)
Multi-task Lasso model trained with L1/L2 mixed-norm as regularizer.
The optimization objective for MultiTaskLasso is:
(1 / (2 * n_samples)) * ||Y - XW||^Fro_2 + alpha * ||W||_21
Where:

Parameters
= 1e-3.
alphas [array-like, optional] List of alphas where to compute the models. If not provided, set
automatically.
max_iter [int, optional] The maximum number of iterations.
tol.


• CV splitter,
v0.22.
n_jobs [int or None, optional (default=None)] Number of CPUs to use during the cross vali-
dation. Note that this is used only if multiple values for l1_ratio are given. None means 1
the RandomState instance used by np.random. Used when selection == ‘random’
Attributes
mse_path_ [array, shape (n_alphas, n_folds)] mean square error for the test set on each fold,
varying alpha
alphas_ [numpy array, shape (n_alphas,)] The grid of alphas used for fitting.
See also:
MultiTaskElasticNet, ElasticNetCV , MultiTaskElasticNetCV
Notes


Examples
>>> from sklearn.linear_model import MultiTaskLassoCV

>>> X, y = make_regression(n_targets=2, noise=4, random_state=0)
>>> reg = MultiTaskLassoCV(cv=5, random_state=0).fit(X, y)
>>> reg.score(X, y)
0.9994...
>>> reg.alpha_
0.5713...
array([[153.7971..., 94.9015...]])
Methods

path(X, y[, eps, n_alphas, alphas, . . . ]) Compute Lasso path with coordinate descent
diction.
__init__(eps=0.001, n_alphas=100, alphas=None, fit_intercept=True, normalize=False,

max_iter=1000, tol=0.0001, copy_X=True, cv=’warn’, verbose=False, n_jobs=None,
random_state=None, selection=’cyclic’)
fit(X, y)
Parameters
be sparse.
Parameters
Returns
static path(X, y, eps=0.001, n_alphas=100, alphas=None, precompute=’auto’, Xy=None,
**params)

Where:

Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Training data. Pass directly
as Fortran-contiguous data to avoid unnecessary memory duplication. If y is mono-output
then X can be sparse.
eps [float, optional] Length of the path. eps=1e-3 means that alpha_min /
alpha_max = 1e-3
set automatically
passed as argument.
written.
when y.ndim == 1).
Returns
along the path.
alpha.
n_iters [array-like, shape (n_alphas,)] The number of iterations taken by the coordinate de-
scent optimizer to reach the specified tolerance for each alpha.

See also:
sparse_encode
Notes

Note that in certain cases, the Lars solver may be significantly faster to implement this functionality. In
particular, linear interpolation can be used to retrieve model coefficients between the values output by
lars_path
Examples
>>> X = np.array([[1, 2, 3.1], [2.3, 5.4, 4.3]]).T

>>> y = np.array([1, 2, 3.1])
[[0. 0. 0.46874778]
[0.2159048 0.4425765 0.23689075]]

>>> # same path
[[0. 0. 0.46915237]
[0.2159048 0.4425765 0.23668876]]
predict(X)
Parameters
Returns

Parameters
Returns
of a nested object.
Returns
self
sklearn.linear_model.OrthogonalMatchingPursuitCV
class sklearn.linear_model.OrthogonalMatchingPursuitCV(copy=True, fit_intercept=True,

normalize=True,
max_iter=None, cv=’warn’,
n_jobs=None, verbose=False)
Cross-validated Orthogonal Matching Pursuit model (OMP).
Parameters
copy [bool, optional] Whether the design matrix X must be copied by the algorithm. A false
value is only helpful if X is already Fortran-ordered, otherwise a copy is made anyway.
fit_intercept [boolean, optional] whether to calculate the intercept for this model. If set to false,
no intercept will be used in calculations (e.g. data is expected to be already centered).
max_iter [integer, optional] Maximum numbers of iterations to perform, therefore maximum
features to include. 10% of n_features but at least 5 if available.
• CV splitter,


v0.22.
Attributes
intercept_ [float or array, shape (n_targets,)] Independent term in decision function.
coef_ [array, shape (n_features,) or (n_targets, n_features)] Parameter vector (w in the problem
formulation).
n_nonzero_coefs_ [int] Estimated number of non-zero coefficients giving the best mean
squared error over the cross-validation folds.
n_iter_ [int or array-like] Number of active features across every target for the model refit with
the best hyperparameters got by cross-validating across all folds.
See also:
orthogonal_mp, orthogonal_mp_gram, lars_path, Lars, LassoLars,
OrthogonalMatchingPursuit, LarsCV , LassoLarsCV , decomposition.sparse_encode
Examples
>>> from sklearn.linear_model import OrthogonalMatchingPursuitCV

>>> X, y = make_regression(n_features=100, n_informative=10,
... noise=4, random_state=0)
>>> reg = OrthogonalMatchingPursuitCV(cv=5).fit(X, y)
>>> reg.score(X, y)
0.9991...
>>> reg.n_nonzero_coefs_
10
array([-78.3854...])
Methods

diction.
__init__(copy=True, fit_intercept=True, normalize=True, max_iter=None, cv=’warn’, n_jobs=None,

verbose=False)

fit(X, y)
Parameters
X [array-like, shape [n_samples, n_features]] Training data.
y [array-like, shape [n_samples]] Target values. Will be cast to X’s dtype if necessary
Returns
Parameters
Returns
predict(X)
Parameters
Returns
Parameters
Returns
of a nested object.
Returns

self
Examples using sklearn.linear_model.OrthogonalMatchingPursuitCV
sklearn.linear_model.RidgeCV
class sklearn.linear_model.RidgeCV(alphas=(0.1, 1.0, 10.0), fit_intercept=True, normal-

ize=False, scoring=None, cv=None, gcv_mode=None,
store_cv_values=False)
Ridge regression with built-in cross-validation.
By default, it performs Generalized Cross-Validation, which is a form of efficient Leave-One-Out cross-
validation.
Parameters
alphas [numpy array of shape [n_alphas]] Array of alpha values to try. Regularization strength;
must be a positive float. Regularization improves the conditioning of the problem and re-
duces the variance of the estimates. Larger values specify stronger regularization. Alpha
corresponds to C^-1 in other linear models such as LogisticRegression or LinearSVC.
fit_intercept [boolean] Whether to calculate the intercept for this model. If set to false, no
scoring [string, callable or None, optional, default: None] A string (see model evaluation doc-
umentation) or a scorer callable object / function with signature scorer(estimator,
X, y).
• None, to use the efficient Leave-One-Out cross-validation
• CV splitter,
For integer/None inputs, if y is binary or multiclass, sklearn.model_selection.
StratifiedKFold is used, else, sklearn.model_selection.KFold is used.
gcv_mode [{None, ‘auto’, ‘svd’, eigen’}, optional] Flag indicating which strategy to use when
performing Generalized Cross-Validation. Options are:

'auto' : use svd if n_samples > n_features or when X is a sparse

matrix, otherwise use eigen
'svd' : force computation via singular value decomposition of X
(does not work for sparse matrices)
'eigen' : force computation via eigendecomposition of X^T X
The ‘auto’ mode is the default and is intended to pick the cheaper option of the two depend-
ing upon the shape and format of the training data.
store_cv_values [boolean, default=False] Flag indicating if the cross-validation values corre-
sponding to each alpha should be stored in the cv_values_ attribute (see below). This
flag is only compatible with cv=None (i.e. using Generalized Cross-Validation).
Attributes
cv_values_ [array, shape = [n_samples, n_alphas] or shape = [n_samples, n_targets, n_alphas],
optional] Cross-validation values for each alpha (if store_cv_values=True and
cv=None). After fit() has been called, this attribute will contain the mean squared
errors (by default) or the values of the {loss,score}_func function (if provided in the
constructor).
coef_ [array, shape = [n_features] or [n_targets, n_features]] Weight vector(s).
intercept_ [float | array, shape = (n_targets,)] Independent term in decision function. Set to 0.0
if fit_intercept = False.
alpha_ [float] Estimated regularization parameter.
See also:
Ridge Ridge regression

RidgeClassifier Ridge classifier
RidgeClassifierCV Ridge classifier with built-in cross validation
Examples
>>> from sklearn.datasets import load_diabetes

>>> from sklearn.linear_model import RidgeCV
>>> X, y = load_diabetes(return_X_y=True)
>>> clf = RidgeCV(alphas=[1e-3, 1e-2, 1e-1, 1]).fit(X, y)
>>> clf.score(X, y)
0.5166...
Methods
fit(X, y[, sample_weight]) Fit Ridge regression model

diction.

__init__(alphas=(0.1, 1.0, 10.0), fit_intercept=True, normalize=False, scoring=None, cv=None,

gcv_mode=None, store_cv_values=False)
Fit Ridge regression model
Parameters
X [array-like, shape = [n_samples, n_features]] Training data
y [array-like, shape = [n_samples] or [n_samples, n_targets]] Target values. Will be cast to
X’s dtype if necessary
sample_weight [float or array-like of shape [n_samples]] Sample weight
Returns
self [object]
Parameters
Returns
predict(X)
Parameters
Returns
Parameters
Returns

of a nested object.
Returns
self
Examples using sklearn.linear_model.RidgeCV

• Effect of transforming the targets in regression model
sklearn.linear_model.RidgeClassifierCV
class sklearn.linear_model.RidgeClassifierCV(alphas=(0.1, 1.0, 10.0), fit_intercept=True,

normalize=False, scoring=None, cv=None,
class_weight=None, store_cv_values=False)
Ridge classifier with built-in cross-validation.
By default, it performs Generalized Cross-Validation, which is a form of efficient Leave-One-Out cross-
validation. Currently, only the n_features > n_samples case is handled efficiently.
Parameters
alphas [numpy array of shape [n_alphas]] Array of alpha values to try. Regularization strength;
must be a positive float. Regularization improves the conditioning of the problem and re-
duces the variance of the estimates. Larger values specify stronger regularization. Alpha
corresponds to C^-1 in other linear models such as LogisticRegression or LinearSVC.
X, y).
• None, to use the efficient Leave-One-Out cross-validation
• CV splitter,

bincount(y))
store_cv_values [boolean, default=False] Flag indicating if the cross-validation values corre-
sponding to each alpha should be stored in the cv_values_ attribute (see below). This
flag is only compatible with cv=None (i.e. using Generalized Cross-Validation).
Attributes
cv_values_ [array, shape = [n_samples, n_targets, n_alphas], optional] Cross-validation values
for each alpha (if store_cv_values=True and cv=None). After fit() has been
called, this attribute will contain the mean squared errors (by default) or the values of the
{loss,score}_func function (if provided in the constructor).
coef_ [array, shape = [n_features] or [n_targets, n_features]] Weight vector(s).
alpha_ [float] Estimated regularization parameter
See also:

RidgeCV Ridge regression with built-in cross validation
Notes
For multi-class classification, n_class classifiers are trained in a one-versus-all approach. Concretely, this is
implemented by taking advantage of the multi-variate response support in Ridge.
Examples
>>> from sklearn.datasets import load_breast_cancer

>>> from sklearn.linear_model import RidgeClassifierCV
>>> X, y = load_breast_cancer(return_X_y=True)
>>> clf = RidgeClassifierCV(alphas=[1e-3, 1e-2, 1e-1, 1]).fit(X, y)
>>> clf.score(X, y)
0.9630...
Methods

fit(X, y[, sample_weight]) Fit the ridge classifier.
Continued on next page

Table 3.11 – continued from previous page

score(X, y[, sample_weight]) Returns the mean accuracy on the given test data and
labels.
__init__(alphas=(0.1, 1.0, 10.0), fit_intercept=True, normalize=False, scoring=None, cv=None,

class_weight=None, store_cv_values=False)
Parameters
Returns
Fit the ridge classifier.
Parameters
X [array-like, shape (n_samples, n_features)] Training vectors, where n_samples is the num-
ber of samples and n_features is the number of features.
y [array-like, shape (n_samples,)] Target values. Will be cast to X’s dtype if necessary
sample_weight [float or numpy array of shape (n_samples,)] Sample weight.
Returns
self [object]
Parameters
Returns
predict(X)
Parameters
Returns
Returns the mean accuracy on the given test data and labels.

In multi-label classification, this is the subset accuracy which is a harsh metric since you require for each
sample that each label set be correctly predicted.
Parameters
y [array-like, shape = (n_samples) or (n_samples, n_outputs)] True labels for X.
Returns
score [float] Mean accuracy of self.predict(X) wrt. y.
of a nested object.
Returns
self
Information Criterion
Some models can offer an information-theoretic closed-form formula of the optimal estimate of the regularization
parameter by computing a single regularization path (instead of several when using cross-validation).
Here is the list of models benefiting from the Akaike Information Criterion (AIC) or the Bayesian Information Criterion
(BIC) for automated model selection:
linear_model.LassoLarsIC([criterion, . . . ]) Lasso model fit with Lars using BIC or AIC for model se-
lection
sklearn.linear_model.LassoLarsIC
class sklearn.linear_model.LassoLarsIC(criterion=’aic’, fit_intercept=True, verbose=False,

normalize=True, precompute=’auto’, max_iter=500,
eps=2.220446049250313e-16, copy_X=True, posi-
tive=False)
Lasso model fit with Lars using BIC or AIC for model selection
AIC is the Akaike information criterion and BIC is the Bayes Information criterion. Such criteria are useful
to select the value of the regularization parameter by making a trade-off between the goodness of fit and the
complexity of the model. A good model should explain well the data while being simple.
Parameters
criterion [‘bic’ | ‘aic’] The type of criterion to use.

as argument.
max_iter [integer, optional] Maximum number of iterations to perform. Can be used for early
stopping.
diagonal factors. Increase this for very ill-conditioned systems. Unlike the tol parameter in
some iterative optimization-based algorithms, this parameter does not control the tolerance
of the optimization.
model coefficients do not converge to the ordinary-least-squares solution for small values
congruence with the solution of the coordinate descent Lasso estimator. As a consequence
using LassoLarsIC only makes sense for problems where a sparse solution is expected and/or
reached.
Attributes
intercept_ [float] independent term in decision function.
alpha_ [float] the alpha parameter chosen by the information criterion
n_iter_ [int] number of iterations run by lars_path to find the grid of alphas.
criterion_ [array, shape (n_alphas,)] The value of the information criteria (‘aic’, ‘bic’) across
all alphas. The alpha which has the smallest information criterion is chosen. This value is
larger by a factor of n_samples compared to Eqns. 2.15 and 2.16 in (Zou et al, 2007).
See also:
lars_path, LassoLars, LassoLarsCV
Notes
The estimation of the number of degrees of freedom is given by:

“On the degrees of freedom of the lasso” Hui Zou, Trevor Hastie, and Robert Tibshirani Ann. Statist. Volume
35, Number 5 (2007), 2173-2192.
https://en.wikipedia.org/wiki/Akaike_information_criterion https://en.wikipedia.org/wiki/Bayesian_
information_criterion

Examples

>>> reg = linear_model.LassoLarsIC(criterion='bic')
>>> reg.fit([[-1, 1], [0, 0], [1, 1]], [-1.1111, 0, -1.1111])
...
LassoLarsIC(copy_X=True, criterion='bic', eps=..., fit_intercept=True,
max_iter=500, normalize=True, positive=False, precompute='auto',
verbose=False)
>>> print(reg.coef_)
[ 0. -1.11...]
Methods
fit(X, y[, copy_X]) Fit the model using X, y as training data.

diction.
__init__(criterion=’aic’, fit_intercept=True, verbose=False, normalize=True, precompute=’auto’,

max_iter=500, eps=2.220446049250313e-16, copy_X=True, positive=False)
fit(X, y, copy_X=True)
Parameters
X [array-like, shape (n_samples, n_features)] training data.
y [array-like, shape (n_samples,)] target values. Will be cast to X’s dtype if necessary
written.
Returns
Parameters
Returns
predict(X)
Parameters
Returns


Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.linear_model.LassoLarsIC
Out of Bag Estimates
When using ensemble methods base upon bagging, i.e. generating new training sets using sampling with replacement,
part of the training set remains unused. For each classifier in the ensemble, a different part of the training set is left
out.
This left out portion can be used to estimate the generalization error without having to rely on a separate validation
set. This estimate comes “for free” as no additional data is needed and can be used for model selection.
This is currently implemented in the following classes:
ensemble.RandomForestClassifier([. . . ]) A random forest classifier.

ensemble.RandomForestRegressor([. . . ]) A random forest regressor.
ensemble.ExtraTreesClassifier([. . . ]) An extra-trees classifier.
ensemble.ExtraTreesRegressor([n_estimators, An extra-trees regressor.
. . . ])
ensemble.GradientBoostingClassifier([loss, Gradient Boosting for classification.
. . . ])


ensemble.GradientBoostingRegressor([loss, Gradient Boosting for regression.
. . . ])
sklearn.ensemble.RandomForestClassifier
class sklearn.ensemble.RandomForestClassifier(n_estimators=’warn’, crite-

rion=’gini’, max_depth=None,
min_samples_split=2, min_samples_leaf=1,
min_weight_fraction_leaf=0.0,
max_features=’auto’,
max_leaf_nodes=None,
min_impurity_decrease=0.0,
min_impurity_split=None, bootstrap=True,
oob_score=False, n_jobs=None,
random_state=None, verbose=0,
warm_start=False, class_weight=None)
A random forest classifier.
A random forest is a meta estimator that fits a number of decision tree classifiers on various sub-samples of the
dataset and uses averaging to improve the predictive accuracy and control over-fitting. The sub-sample size is
always the same as the original input sample size but the samples are drawn with replacement if bootstrap=True
(default).
Parameters
n_estimators [integer, optional (default=10)] The number of trees in the forest.
Changed in version 0.20: The default value of n_estimators will change from 10 in
version 0.20 to 100 in version 0.22.
criterion [string, optional (default=”gini”)] The function to measure the quality of a split. Sup-
ported criteria are “gini” for the Gini impurity and “entropy” for the information gain. Note:
this parameter is tree-specific.
max_depth [integer or None, optional (default=None)] The maximum depth of the tree. If
None, then nodes are expanded until all leaves are pure or until all leaves contain less than
min_samples_split samples.
min_samples_split [int, float, optional (default=2)] The minimum number of samples required
to split an internal node:
• If int, then consider min_samples_split as the minimum number.
• If float, then min_samples_split is a fraction and ceil(min_samples_split * n_samples) are
the minimum number of samples for each split.
Changed in version 0.18: Added float values for fractions.
min_samples_leaf [int, float, optional (default=1)] The minimum number of samples required
to be at a leaf node. A split point at any depth will only be considered if it leaves at least
min_samples_leaf training samples in each of the left and right branches. This may
have the effect of smoothing the model, especially in regression.
• If int, then consider min_samples_leaf as the minimum number.
• If float, then min_samples_leaf is a fraction and ceil(min_samples_leaf * n_samples) are
the minimum number of samples for each node.


min_weight_fraction_leaf [float, optional (default=0.)] The minimum weighted fraction of the
sum total of weights (of all the input samples) required to be at a leaf node. Samples have
equal weight when sample_weight is not provided.
max_features [int, float, string or None, optional (default=”auto”)] The number of features to
consider when looking for the best split:
• If int, then consider max_features features at each split.
• If float, then max_features is a fraction and int(max_features * n_features) features are
considered at each split.
• If “auto”, then max_features=sqrt(n_features).
• If “sqrt”, then max_features=sqrt(n_features) (same as “auto”).
• If “log2”, then max_features=log2(n_features).
• If None, then max_features=n_features.
Note: the search for a split does not stop until at least one valid partition of the node samples
is found, even if it requires to effectively inspect more than max_features features.
max_leaf_nodes [int or None, optional (default=None)] Grow trees with max_leaf_nodes
in best-first fashion. Best nodes are defined as relative reduction in impurity. If None then
unlimited number of leaf nodes.
min_impurity_decrease [float, optional (default=0.)] A node will be split if this split induces
a decrease of the impurity greater than or equal to this value.
The weighted impurity decrease equation is the following:
N_t / N * (impurity - N_t_R / N_t * right_impurity

- N_t_L / N_t * left_impurity)
where N is the total number of samples, N_t is the number of samples at the current node,
N_t_L is the number of samples in the left child, and N_t_R is the number of samples in
the right child.
N, N_t, N_t_R and N_t_L all refer to the weighted sum, if sample_weight is passed.
New in version 0.19.
min_impurity_split [float, (default=1e-7)] Threshold for early stopping in tree growth. A node
will split if its impurity is above the threshold, otherwise it is a leaf.
Deprecated since version 0.19: min_impurity_split has been deprecated in favor of
min_impurity_decrease in 0.19. The default value of min_impurity_split
will change from 1e-7 to 0 in 0.23 and it will be removed in 0.25. Use
min_impurity_decrease instead.
bootstrap [boolean, optional (default=True)] Whether bootstrap samples are used when build-
ing trees. If False, the whole datset is used to build each tree.
oob_score [bool (default=False)] Whether to use out-of-bag samples to estimate the general-
ization accuracy.
n_jobs [int or None, optional (default=None)] The number of jobs to run in parallel for both
fit and predict. None means 1 unless in a joblib.parallel_backend context. -1

random_state [int, RandomState instance or None, optional (default=None)] If int, ran-

the RandomState instance used by np.random.
verbose [int, optional (default=0)] Controls the verbosity when fitting and predicting.
warm_start [bool, optional (default=False)] When set to True, reuse the solution of the pre-
vious call to fit and add more estimators to the ensemble, otherwise, just fit a whole new
forest. See the Glossary.
class_weight [dict, list of dicts, “balanced”, “balanced_subsample” or None, optional (de-
fault=None)] Weights associated with classes in the form {class_label: weight}.
If not given, all classes are supposed to have weight one. For multi-output problems, a list
of dicts can be provided in the same order as the columns of y.
Note that for multioutput (including multilabel) weights should be defined for each class of
every column in its own dict. For example, for four-class multilabel classification weights
should be [{0: 1, 1: 1}, {0: 1, 1: 5}, {0: 1, 1: 1}, {0: 1, 1: 1}] instead of [{1:1}, {2:5},
{3:1}, {4:1}].
bincount(y))
The “balanced_subsample” mode is the same as “balanced” except that weights are com-
puted based on the bootstrap sample for every tree grown.
For multi-output, the weights of each column of y will be multiplied.
Attributes
estimators_ [list of DecisionTreeClassifier] The collection of fitted sub-estimators.
classes_ [array of shape = [n_classes] or a list of such arrays] The classes labels (single output
problem), or a list of arrays of class labels (multi-output problem).
n_classes_ [int or list] The number of classes (single output problem), or a list containing the
number of classes for each output (multi-output problem).
n_features_ [int] The number of features when fit is performed.
n_outputs_ [int] The number of outputs when fit is performed.
feature_importances_ [array of shape = [n_features]] Return the feature importances
(the higher, the more important the feature).
oob_score_ [float] Score of the training dataset obtained using an out-of-bag estimate.
oob_decision_function_ [array of shape = [n_samples, n_classes]] Decision function com-
puted with out-of-bag estimate on the training set. If n_estimators is small it might
be possible that a data point was never left out during the bootstrap. In this case,
oob_decision_function_ might contain NaN.
See also:
DecisionTreeClassifier, ExtraTreesClassifier

Notes
The default values for the parameters controlling the size of the trees (e.g. max_depth,
min_samples_leaf, etc.) lead to fully grown and unpruned trees which can potentially be very large on
some data sets. To reduce memory consumption, the complexity and size of the trees should be controlled by
setting those parameter values.
The features are always randomly permuted at each split. Therefore, the best found split may vary, even with
the same training data, max_features=n_features and bootstrap=False, if the improvement of the
criterion is identical for several splits enumerated during the search of the best split. To obtain a deterministic
behaviour during fitting, random_state has to be fixed.
References
[1]
Examples

>>> X, y = make_classification(n_samples=1000, n_features=4,

... n_informative=2, n_redundant=0,
... random_state=0, shuffle=False)
>>> clf = RandomForestClassifier(n_estimators=100, max_depth=2,
... random_state=0)
>>> clf.fit(X, y)
RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',
max_depth=2, max_features='auto', max_leaf_nodes=None,
min_impurity_decrease=0.0, min_impurity_split=None,
min_samples_leaf=1, min_samples_split=2,
min_weight_fraction_leaf=0.0, n_estimators=100, n_jobs=None,
oob_score=False, random_state=0, verbose=0, warm_start=False)
>>> print(clf.feature_importances_)
[0.14205973 0.76664038 0.0282433 0.06305659]
>>> print(clf.predict([[0, 0, 0, 0]]))
[1]
Methods
apply(X) Apply trees in the forest to X, return leaf indices.

decision_path(X) Return the decision path in the forest
fit(X, y[, sample_weight]) Build a forest of trees from the training set (X, y).
predict(X) Predict class for X.
predict_log_proba(X) Predict class log-probabilities for X.
predict_proba(X) Predict class probabilities for X.
labels.

__init__(n_estimators=’warn’, criterion=’gini’, max_depth=None, min_samples_split=2,

min_samples_leaf=1, min_weight_fraction_leaf=0.0, max_features=’auto’,
max_leaf_nodes=None, min_impurity_decrease=0.0, min_impurity_split=None,
bootstrap=True, oob_score=False, n_jobs=None, random_state=None, verbose=0,
apply(X)
Apply trees in the forest to X, return leaf indices.
Parameters
X [array-like or sparse matrix, shape = [n_samples, n_features]] The input samples. Inter-
nally, its dtype will be converted to dtype=np.float32. If a sparse matrix is provided,
it will be converted into a sparse csr_matrix.
Returns
X_leaves [array_like, shape = [n_samples, n_estimators]] For each datapoint x in X and for
each tree in the forest, return the index of the leaf x ends up in.
decision_path(X)
Return the decision path in the forest
Parameters
Returns
indicator [sparse csr array, shape = [n_samples, n_nodes]] Return a node indicator matrix
where non zero elements indicates that the samples goes through the nodes.
n_nodes_ptr [array of size (n_estimators + 1, )] The columns from indica-
tor[n_nodes_ptr[i]:n_nodes_ptr[i+1]] gives the indicator value for the i-th estimator.
feature_importances_
Return the feature importances (the higher, the more important the feature).
Returns
feature_importances_ [array, shape = [n_features]]
Build a forest of trees from the training set (X, y).
Parameters
X [array-like or sparse matrix of shape = [n_samples, n_features]] The training input sam-
ples. Internally, its dtype will be converted to dtype=np.float32. If a sparse matrix
is provided, it will be converted into a sparse csc_matrix.
y [array-like, shape = [n_samples] or [n_samples, n_outputs]] The target values (class labels
in classification, real numbers in regression).
sample_weight [array-like, shape = [n_samples] or None] Sample weights. If None, then
samples are equally weighted. Splits that would create child nodes with net zero or neg-
ative weight are ignored while searching for a split in each node. In the case of classifi-
cation, splits are also ignored if they would result in any single class carrying a negative
weight in either child node.

Returns
self [object]
Parameters
Returns
predict(X)
Predict class for X.
The predicted class of an input sample is a vote by the trees in the forest, weighted by their probability
estimates. That is, the predicted class is the one with highest mean probability estimate across the trees.
Parameters
X [array-like or sparse matrix of shape = [n_samples, n_features]] The input samples. Inter-
Returns
y [array of shape = [n_samples] or [n_samples, n_outputs]] The predicted classes.
Predict class log-probabilities for X.
The predicted class log-probabilities of an input sample is computed as the log of the mean predicted class
probabilities of the trees in the forest.
Parameters
Returns
p [array of shape = [n_samples, n_classes], or a list of n_outputs] such arrays if n_outputs
> 1. The class probabilities of the input samples. The order of the classes corresponds to
that in the attribute classes_.
predict_proba(X)
Predict class probabilities for X.
The predicted class probabilities of an input sample are computed as the mean predicted class probabilities
of the trees in the forest. The class probability of a single tree is the fraction of samples of the same class
in a leaf.
Parameters
Returns


Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.ensemble.RandomForestClassifier
• Comparison of Calibration of Classifiers

• Probability Calibration for 3-class classification
• Classifier comparison
• Plot class probabilities calculated by the VotingClassifier
• Feature transformations with ensembles of trees
• Comparing randomized search and grid search for hyperparameter estimation

sklearn.ensemble.RandomForestRegressor
class sklearn.ensemble.RandomForestRegressor(n_estimators=’warn’, crite-

rion=’mse’, max_depth=None,
max_features=’auto’, max_leaf_nodes=None,
min_impurity_split=None, bootstrap=True,
warm_start=False)
A random forest regressor.
A random forest is a meta estimator that fits a number of classifying decision trees on various sub-samples of the
dataset and uses averaging to improve the predictive accuracy and control over-fitting. The sub-sample size is
always the same as the original input sample size but the samples are drawn with replacement if bootstrap=True
(default).
Parameters
criterion [string, optional (default=”mse”)] The function to measure the quality of a split. Sup-
ported criteria are “mse” for the mean squared error, which is equal to variance reduction as
feature selection criterion, and “mae” for the mean absolute error.
New in version 0.18: Mean Absolute Error (MAE) criterion.


• If “auto”, then max_features=n_features.
• If “sqrt”, then max_features=sqrt(n_features).

the right child.
bootstrap [boolean, optional (default=True)] Whether bootstrap samples are used when build-
oob_score [bool, optional (default=False)] whether to use out-of-bag samples to estimate the
R^2 on unseen data.
fit and predict. None‘ means 1 unless in a joblib.parallel_backend context. -1

Attributes
estimators_ [list of DecisionTreeRegressor] The collection of fitted sub-estimators.
oob_prediction_ [array of shape = [n_samples]] Prediction computed with out-of-bag estimate
on the training set.
See also:
DecisionTreeRegressor, ExtraTreesRegressor
Notes
the same training data, max_features=n_features and bootstrap=False, if the improvement of the
criterion is identical for several splits enumerated during the search of the best split. To obtain a deterministic
behaviour during fitting, random_state has to be fixed.
The default value max_features="auto" uses n_features rather than n_features / 3. The latter
was originally suggested in [1], whereas the former was more recently justified empirically in [2].
References
[1], [2]
Examples
>>> from sklearn.ensemble import RandomForestRegressor

>>> X, y = make_regression(n_features=4, n_informative=2,

... random_state=0, shuffle=False)
>>> regr = RandomForestRegressor(max_depth=2, random_state=0,
... n_estimators=100)

>>> regr.fit(X, y)
RandomForestRegressor(bootstrap=True, criterion='mse', max_depth=2,
max_features='auto', max_leaf_nodes=None,
oob_score=False, random_state=0, verbose=0, warm_start=False)
>>> print(regr.feature_importances_)
[0.18146984 0.81473937 0.00145312 0.00233767]
>>> print(regr.predict([[0, 0, 0, 0]]))
[-8.32987858]
Methods

predict(X) Predict regression target for X.
diction.
__init__(n_estimators=’warn’, criterion=’mse’, max_depth=None, min_samples_split=2,

max_leaf_nodes=None, min_impurity_decrease=0.0, min_impurity_split=None,
bootstrap=True, oob_score=False, n_jobs=None, random_state=None, verbose=0,
warm_start=False)
apply(X)
Parameters
Returns
decision_path(X)
Parameters
Returns

Returns
Parameters
Returns
self [object]
Parameters
Returns
predict(X)
Predict regression target for X.
The predicted regression target of an input sample is computed as the mean predicted regression targets of
the trees in the forest.
Parameters
Returns
y [array of shape = [n_samples] or [n_samples, n_outputs]] The predicted values.

Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.ensemble.RandomForestRegressor
• Imputing missing values before building an estimator

• Prediction Latency
• Comparing random forests and the multi-output meta estimator
sklearn.ensemble.ExtraTreesClassifier
class sklearn.ensemble.ExtraTreesClassifier(n_estimators=’warn’, crite-

rion=’gini’, max_depth=None,
min_impurity_split=None, bootstrap=False,
An extra-trees classifier.
This class implements a meta estimator that fits a number of randomized decision trees (a.k.a. extra-trees) on
various sub-samples of the dataset and uses averaging to improve the predictive accuracy and control over-fitting.

Parameters
ported criteria are “gini” for the Gini impurity and “entropy” for the information gain.


the right child.
bootstrap [boolean, optional (default=False)] Whether bootstrap samples are used when build-
oob_score [bool, optional (default=False)] Whether to use out-of-bag samples to estimate the
generalization accuracy.
class_weight [dict, list of dicts, “balanced”, “balanced_subsample” or None, optional (de-
fault=None)] Weights associated with classes in the form {class_label: weight}.
If not given, all classes are supposed to have weight one. For multi-output problems, a list
of dicts can be provided in the same order as the columns of y.
{3:1}, {4:1}].
bincount(y))
The “balanced_subsample” mode is the same as “balanced” except that weights are com-
puted based on the bootstrap sample for every tree grown.

Attributes
n_classes_ [int or list] The number of classes (single output problem), or a list containing the
number of classes for each output (multi-output problem).
See also:
sklearn.tree.ExtraTreeClassifier Base classifier for this ensemble.

RandomForestClassifier Ensemble Classifier based on trees with optimal splits.
Notes
References
[1]
Methods



labels.
__init__(n_estimators=’warn’, criterion=’gini’, max_depth=None, min_samples_split=2,

max_leaf_nodes=None, min_impurity_decrease=0.0, min_impurity_split=None, boot-
strap=False, oob_score=False, n_jobs=None, random_state=None, verbose=0,
apply(X)
Parameters
Returns
decision_path(X)
Parameters
Returns
Returns
Parameters


Returns
self [object]
Parameters
Returns
predict(X)
The predicted class of an input sample is a vote by the trees in the forest, weighted by their probability
estimates. That is, the predicted class is the one with highest mean probability estimate across the trees.
Parameters
Returns
y [array of shape = [n_samples] or [n_samples, n_outputs]] The predicted classes.
probabilities of the trees in the forest.
Parameters
Returns
predict_proba(X)
The predicted class probabilities of an input sample are computed as the mean predicted class probabilities
of the trees in the forest. The class probability of a single tree is the fraction of samples of the same class
in a leaf.
Parameters

Returns
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.ensemble.ExtraTreesClassifier


sklearn.ensemble.ExtraTreesRegressor
class sklearn.ensemble.ExtraTreesRegressor(n_estimators=’warn’, crite-

rion=’mse’, max_depth=None,
min_impurity_split=None, bootstrap=False,
warm_start=False)
An extra-trees regressor.
This class implements a meta estimator that fits a number of randomized decision trees (a.k.a. extra-trees) on
various sub-samples of the dataset and uses averaging to improve the predictive accuracy and control over-fitting.
Parameters
feature selection criterion, and “mae” for the mean absolute error.


the right child.
bootstrap [boolean, optional (default=False)] Whether bootstrap samples are used when build-
R^2 on unseen data.

Attributes
estimators_ [list of DecisionTreeRegressor] The collection of fitted sub-estimators.
n_features_ [int] The number of features.
n_outputs_ [int] The number of outputs.
on the training set.
See also:
sklearn.tree.ExtraTreeRegressor Base estimator for this ensemble.

RandomForestRegressor Ensemble regressor using trees with optimal splits.
Notes
References
[1]
Methods

diction.
__init__(n_estimators=’warn’, criterion=’mse’, max_depth=None, min_samples_split=2,

max_leaf_nodes=None, min_impurity_decrease=0.0, min_impurity_split=None, boot-
strap=False, oob_score=False, n_jobs=None, random_state=None, verbose=0,
warm_start=False)
apply(X)


Parameters
Returns
decision_path(X)
Parameters
Returns
Returns
Parameters
Returns
self [object]

Parameters
Returns
predict(X)
the trees in the forest.
Parameters
Returns
y [array of shape = [n_samples] or [n_samples, n_outputs]] The predicted values.
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.ensemble.ExtraTreesRegressor

sklearn.ensemble.GradientBoostingClassifier
class sklearn.ensemble.GradientBoostingClassifier(loss=’deviance’, learning_rate=0.1,

n_estimators=100, subsam-
ple=1.0, criterion=’friedman_mse’,
min_samples_split=2,
min_samples_leaf=1,
max_depth=3,
min_impurity_split=None,
init=None, random_state=None,
max_features=None, ver-
bose=0, max_leaf_nodes=None,
warm_start=False, pre-
sort=’auto’, validation_fraction=0.1,
n_iter_no_change=None, tol=0.0001)
Gradient Boosting for classification.
GB builds an additive model in a forward stage-wise fashion; it allows for the optimization of arbitrary differen-
tiable loss functions. In each stage n_classes_ regression trees are fit on the negative gradient of the binomial
or multinomial deviance loss function. Binary classification is a special case where only a single regression tree
is induced.
Parameters
loss [{‘deviance’, ‘exponential’}, optional (default=’deviance’)] loss function to be optimized.
‘deviance’ refers to deviance (= logistic regression) for classification with probabilistic out-
puts. For loss ‘exponential’ gradient boosting recovers the AdaBoost algorithm.
learning_rate [float, optional (default=0.1)] learning rate shrinks the contribution of each tree
by learning_rate. There is a trade-off between learning_rate and n_estimators.
n_estimators [int (default=100)] The number of boosting stages to perform. Gradient boosting
is fairly robust to over-fitting so a large number usually results in better performance.
subsample [float, optional (default=1.0)] The fraction of samples to be used for fitting the in-
dividual base learners. If smaller than 1.0 this results in Stochastic Gradient Boosting. sub-
sample interacts with the parameter n_estimators. Choosing subsample < 1.0 leads to a
reduction of variance and an increase in bias.
criterion [string, optional (default=”friedman_mse”)] The function to measure the quality of
a split. Supported criteria are “friedman_mse” for the mean squared error with improve-
ment score by Friedman, “mse” for mean squared error, and “mae” for the mean absolute
error. The default value of “friedman_mse” is generally the best as it can provide a better
approximation in some cases.

max_depth [integer, optional (default=3)] maximum depth of the individual regression estima-
tors. The maximum depth limits the number of nodes in the tree. Tune this parameter for
best performance; the best value depends on the interaction of the input variables.

the right child.
init [estimator, optional] An estimator object that is used to compute the initial predictions.
init has to provide fit and predict. If None it uses loss.init_estimator.
max_features [int, float, string or None, optional (default=None)] The number of features to


Choosing max_features < n_features leads to a reduction of variance and an increase in bias.
verbose [int, default: 0] Enable verbose output. If 1 then it prints progress and performance
once in a while (the more trees the lower the frequency). If greater than 1 then it prints
progress and performance for every tree.
warm_start [bool, default: False] When set to True, reuse the solution of the previous call to
fit and add more estimators to the ensemble, otherwise, just erase the previous solution. See
the Glossary.
presort [bool or ‘auto’, optional (default=’auto’)] Whether to presort the data to speed up the
finding of best splits in fitting. Auto mode by default will use presorting on dense data and
default to normal sorting on sparse data. Setting presort to true on sparse data will raise an
error.
New in version 0.17: presort parameter.
validation_fraction [float, optional, default 0.1] The proportion of training data to set
aside as validation set for early stopping. Must be between 0 and 1. Only used if
n_iter_no_change is set to an integer.
n_iter_no_change [int, default None] n_iter_no_change is used to decide if early stop-
ping will be used to terminate training when validation score is not improving. By de-
fault it is set to None to disable early stopping. If set to a number, it will set aside
validation_fraction size of the training data as validation and terminate training
when validation score is not improving in all of the previous n_iter_no_change num-
bers of iterations.
tol [float, optional, default 1e-4] Tolerance for the early stopping. When the loss is not improv-
ing by at least tol for n_iter_no_change iterations (if set to a number), the training
stops.
Attributes
n_estimators_ [int] The number of estimators as selected by early stopping (if
n_iter_no_change is specified). Otherwise it is set to n_estimators.
feature_importances_ [array, shape (n_features,)] Return the feature importances (the
higher, the more important the feature).
oob_improvement_ [array, shape (n_estimators,)] The improvement in loss (= deviance) on
the out-of-bag samples relative to the previous iteration. oob_improvement_[0] is the
improvement in loss of the first stage over the init estimator.

train_score_ [array, shape (n_estimators,)] The i-th score train_score_[i] is the de-
viance (= loss) of the model at iteration i on the in-bag sample. If subsample == 1
this is the deviance on the training data.
loss_ [LossFunction] The concrete LossFunction object.
init_ [estimator] The estimator that provides the initial predictions. Set via the init argument
or loss.init_estimator.
estimators_ [ndarray of DecisionTreeRegressor,shape (n_estimators, loss_.K)] The collec-
tion of fitted sub-estimators. loss_.K is 1 for binary classification, otherwise n_classes.
See also:
sklearn.tree.DecisionTreeClassifier, RandomForestClassifier,
AdaBoostClassifier
Notes
the same training data and max_features=n_features, if the improvement of the criterion is identical for
several splits enumerated during the search of the best split. To obtain a deterministic behaviour during fitting,
random_state has to be fixed.
References
J. Friedman, Greedy Function Approximation: A Gradient Boosting Machine, The Annals of Statistics, Vol. 29,
No. 5, 2001.
10. Friedman, Stochastic Gradient Boosting, 1999
T. Hastie, R. Tibshirani and J. Friedman. Elements of Statistical Learning Ed. 2, Springer, 2009.
Methods
apply(X) Apply trees in the ensemble to X, return leaf indices.

decision_function(X) Compute the decision function of X.
fit(X, y[, sample_weight, monitor]) Fit the gradient boosting model.
labels.
staged_decision_function(X) Compute decision function of X for each iteration.
staged_predict(X) Predict class at each stage for X.
staged_predict_proba(X) Predict class probabilities at each stage for X.

__init__(loss=’deviance’, learning_rate=0.1, n_estimators=100, subsample=1.0,

criterion=’friedman_mse’, min_samples_split=2, min_samples_leaf=1,
min_weight_fraction_leaf=0.0, max_depth=3, min_impurity_decrease=0.0,
min_impurity_split=None, init=None, random_state=None, max_features=None,
verbose=0, max_leaf_nodes=None, warm_start=False, presort=’auto’, valida-
tion_fraction=0.1, n_iter_no_change=None, tol=0.0001)
apply(X)
Apply trees in the ensemble to X, return leaf indices.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] The input samples. Inter-
it will be converted to a sparse csr_matrix.
Returns
X_leaves [array-like, shape (n_samples, n_estimators, n_classes)] For each datapoint x in X
and for each tree in the ensemble, return the index of the leaf x ends up in each estimator.
In the case of binary classification n_classes is 1.
Compute the decision function of X.
Parameters
nally, it will be converted to dtype=np.float32 and if a sparse matrix is provided to
a sparse csr_matrix.
Returns
score [array, shape (n_samples, n_classes) or (n_samples,)] The decision function of the
input samples. The order of the classes corresponds to that in the attribute classes_. Re-
gression and binary classification produce an array of shape [n_samples].
Returns
feature_importances_ [array, shape (n_features,)]
fit(X, y, sample_weight=None, monitor=None)

Fit the gradient boosting model.
Parameters
y [array-like, shape (n_samples,)] Target values (strings or integers in classification, real
numbers in regression) For classification, labels must correspond to classes.
sample_weight [array-like, shape (n_samples,) or None] Sample weights. If None, then

monitor [callable, optional] The monitor is called after each iteration with the current iter-
ation, a reference to the estimator and the local variables of _fit_stages as keyword
arguments callable(i, self, locals()). If the callable returns True the fit-
ting procedure is stopped. The monitor can be used for various things such as computing
held-out estimates, early stopping, model introspect, and snapshoting.
Returns
self [object]
Parameters
Returns
n_features
DEPRECATED: Attribute n_features was deprecated in version 0.19 and will be removed in 0.21.
predict(X)
Parameters
Returns
y [array, shape (n_samples,)] The predicted values.
Parameters
Returns
p [array, shape (n_samples, n_classes)] The class log-probabilities of the input samples. The
order of the classes corresponds to that in the attribute classes_.
Raises
AttributeError If the loss does not support probabilities.
predict_proba(X)
Parameters

Returns
p [array, shape (n_samples, n_classes)] The class probabilities of the input samples. The
Raises
AttributeError If the loss does not support probabilities.
Parameters
Returns
of a nested object.
Returns
self
staged_decision_function(X)
Compute decision function of X for each iteration.
This method allows monitoring (i.e. determine error on testing set) after each stage.
Parameters
Returns
score [generator of array, shape (n_samples, k)] The decision function of the input samples.
The order of the classes corresponds to that in the attribute classes_. Regression and binary
classification are special cases with k == 1, otherwise k==n_classes.
staged_predict(X)
Predict class at each stage for X.
Parameters

Returns
y [generator of array of shape (n_samples,)] The predicted value of the input samples.
staged_predict_proba(X)
Predict class probabilities at each stage for X.
Parameters
Returns
Examples using sklearn.ensemble.GradientBoostingClassifier

• Early stopping of Gradient Boosting
• Feature discretization
sklearn.ensemble.GradientBoostingRegressor
class sklearn.ensemble.GradientBoostingRegressor(loss=’ls’, learning_rate=0.1,

n_estimators=100, subsam-
ple=1.0, criterion=’friedman_mse’,
min_samples_split=2,
min_samples_leaf=1,
max_depth=3,
min_impurity_split=None,
init=None, random_state=None,
max_features=None, alpha=0.9,
verbose=0, max_leaf_nodes=None,
warm_start=False, pre-
sort=’auto’, validation_fraction=0.1,
n_iter_no_change=None, tol=0.0001)
Gradient Boosting for regression.
GB builds an additive model in a forward stage-wise fashion; it allows for the optimization of arbitrary differ-
entiable loss functions. In each stage a regression tree is fit on the negative gradient of the given loss function.
Parameters

loss [{‘ls’, ‘lad’, ‘huber’, ‘quantile’}, optional (default=’ls’)] loss function to be optimized. ‘ls’
refers to least squares regression. ‘lad’ (least absolute deviation) is a highly robust loss
function solely based on order information of the input variables. ‘huber’ is a combination
of the two. ‘quantile’ allows quantile regression (use alpha to specify the quantile).
learning_rate [float, optional (default=0.1)] learning rate shrinks the contribution of each tree
by learning_rate. There is a trade-off between learning_rate and n_estimators.
n_estimators [int (default=100)] The number of boosting stages to perform. Gradient boosting
is fairly robust to over-fitting so a large number usually results in better performance.
subsample [float, optional (default=1.0)] The fraction of samples to be used for fitting the in-
dividual base learners. If smaller than 1.0 this results in Stochastic Gradient Boosting. sub-
sample interacts with the parameter n_estimators. Choosing subsample < 1.0 leads to a
reduction of variance and an increase in bias.
criterion [string, optional (default=”friedman_mse”)] The function to measure the quality of
a split. Supported criteria are “friedman_mse” for the mean squared error with improve-
ment score by Friedman, “mse” for mean squared error, and “mae” for the mean absolute
error. The default value of “friedman_mse” is generally the best as it can provide a better
approximation in some cases.
max_depth [integer, optional (default=3)] maximum depth of the individual regression estima-
tors. The maximum depth limits the number of nodes in the tree. Tune this parameter for
best performance; the best value depends on the interaction of the input variables.


the right child.
init [estimator, optional (default=None)] An estimator object that is used to compute the ini-
tial predictions. init has to provide fit and predict. If None it uses loss.
init_estimator.
Choosing max_features < n_features leads to a reduction of variance and an increase in bias.
alpha [float (default=0.9)] The alpha-quantile of the huber loss function and the quantile loss
function. Only if loss='huber' or loss='quantile'.
verbose [int, default: 0] Enable verbose output. If 1 then it prints progress and performance
once in a while (the more trees the lower the frequency). If greater than 1 then it prints
progress and performance for every tree.
fit and add more estimators to the ensemble, otherwise, just erase the previous solution. See
the Glossary.
presort [bool or ‘auto’, optional (default=’auto’)] Whether to presort the data to speed up the
finding of best splits in fitting. Auto mode by default will use presorting on dense data and

default to normal sorting on sparse data. Setting presort to true on sparse data will raise an
error.
New in version 0.17: optional parameter presort.
validation_fraction [float, optional, default 0.1] The proportion of training data to set
aside as validation set for early stopping. Must be between 0 and 1. Only used if
n_iter_no_change is set to an integer.
n_iter_no_change [int, default None] n_iter_no_change is used to decide if early stop-
ping will be used to terminate training when validation score is not improving. By de-
fault it is set to None to disable early stopping. If set to a number, it will set aside
validation_fraction size of the training data as validation and terminate training
when validation score is not improving in all of the previous n_iter_no_change num-
bers of iterations.
tol [float, optional, default 1e-4] Tolerance for the early stopping. When the loss is not improv-
ing by at least tol for n_iter_no_change iterations (if set to a number), the training
stops.
Attributes
feature_importances_ [array, shape (n_features,)] Return the feature importances (the
higher, the more important the feature).
oob_improvement_ [array, shape (n_estimators,)] The improvement in loss (= deviance) on
the out-of-bag samples relative to the previous iteration. oob_improvement_[0] is the
improvement in loss of the first stage over the init estimator.
train_score_ [array, shape (n_estimators,)] The i-th score train_score_[i] is the de-
viance (= loss) of the model at iteration i on the in-bag sample. If subsample == 1
this is the deviance on the training data.
loss_ [LossFunction] The concrete LossFunction object.
init_ [estimator] The estimator that provides the initial predictions. Set via the init argument
or loss.init_estimator.
estimators_ [array of DecisionTreeRegressor, shape (n_estimators, 1)] The collection of fitted
sub-estimators.
See also:
DecisionTreeRegressor, RandomForestRegressor
Notes

References
J. Friedman, Greedy Function Approximation: A Gradient Boosting Machine, The Annals of Statistics, Vol. 29,
No. 5, 2001.
10. Friedman, Stochastic Gradient Boosting, 1999
T. Hastie, R. Tibshirani and J. Friedman. Elements of Statistical Learning Ed. 2, Springer, 2009.
Methods
apply(X) Apply trees in the ensemble to X, return leaf indices.

fit(X, y[, sample_weight, monitor]) Fit the gradient boosting model.
diction.
staged_predict(X) Predict regression target at each stage for X.
__init__(loss=’ls’, learning_rate=0.1, n_estimators=100, subsample=1.0, cri-

terion=’friedman_mse’, min_samples_split=2, min_samples_leaf=1,
min_weight_fraction_leaf=0.0, max_depth=3, min_impurity_decrease=0.0,
min_impurity_split=None, init=None, random_state=None, max_features=None, al-
pha=0.9, verbose=0, max_leaf_nodes=None, warm_start=False, presort=’auto’, valida-
tion_fraction=0.1, n_iter_no_change=None, tol=0.0001)
apply(X)
Apply trees in the ensemble to X, return leaf indices.
Parameters
it will be converted to a sparse csr_matrix.
Returns
X_leaves [array-like, shape (n_samples, n_estimators)] For each datapoint x in X and for
each tree in the ensemble, return the index of the leaf x ends up in each estimator.
Returns
feature_importances_ [array, shape (n_features,)]
fit(X, y, sample_weight=None, monitor=None)

Fit the gradient boosting model.
Parameters

y [array-like, shape (n_samples,)] Target values (strings or integers in classification, real
numbers in regression) For classification, labels must correspond to classes.
sample_weight [array-like, shape (n_samples,) or None] Sample weights. If None, then
monitor [callable, optional] The monitor is called after each iteration with the current iter-
ation, a reference to the estimator and the local variables of _fit_stages as keyword
arguments callable(i, self, locals()). If the callable returns True the fit-
ting procedure is stopped. The monitor can be used for various things such as computing
held-out estimates, early stopping, model introspect, and snapshoting.
Returns
self [object]
Parameters
Returns
n_features
DEPRECATED: Attribute n_features was deprecated in version 0.19 and will be removed in 0.21.
predict(X)
Parameters
Returns
y [array, shape (n_samples,)] The predicted values.
Parameters


Returns
of a nested object.
Returns
self
staged_predict(X)
Predict regression target at each stage for X.
Parameters
Returns
Examples using sklearn.ensemble.GradientBoostingRegressor
• Model Complexity Influence

• Prediction Intervals for Gradient Boosting Regression
3.3.3 Model evaluation: quantifying the quality of predictions
There are 3 different APIs for evaluating the quality of a model’s predictions:
• Estimator score method: Estimators have a score method providing a default evaluation criterion for the
problem they are designed to solve. This is not discussed on this page, but in each estimator’s documentation.
• Scoring parameter: Model-evaluation tools using cross-validation (such as model_selection.
cross_val_score and model_selection.GridSearchCV ) rely on an internal scoring strategy. This
is discussed in the section The scoring parameter: defining model evaluation rules.
• Metric functions: The metrics module implements functions assessing prediction error for specific purposes.
These metrics are detailed in sections on Classification metrics, Multilabel ranking metrics, Regression metrics
and Clustering metrics.
Finally, Dummy estimators are useful to get a baseline value of those metrics for random predictions.
See also:

For “pairwise” metrics, between samples and not estimators or predictions, see the Pairwise metrics, Affinities and
Kernels section.
The scoring parameter: defining model evaluation rules
Model selection and evaluation using tools, such as model_selection.GridSearchCV and

model_selection.cross_val_score, take a scoring parameter that controls what metric they ap-
ply to the estimators evaluated.
Common cases: predefined values
For the most common use cases, you can designate a scorer object with the scoring parameter; the table below
shows all possible values. All scorer objects follow the convention that higher return values are better than
lower return values. Thus metrics which measure the distance between the model and the data, like metrics.
mean_squared_error, are available as neg_mean_squared_error which return the negated value of the metric.
Scoring Function Comment

Classification
‘accuracy’ metrics.accuracy_score
‘balanced_accuracy’ metrics.balanced_accuracy_score for binary targets
‘average_precision’ metrics.average_precision_score
‘brier_score_loss’ metrics.brier_score_loss
‘f1’ metrics.f1_score for binary targets
‘f1_micro’ metrics.f1_score micro-averaged
‘f1_macro’ metrics.f1_score macro-averaged
‘f1_weighted’ metrics.f1_score weighted average
‘f1_samples’ metrics.f1_score by multilabel sample
‘neg_log_loss’ metrics.log_loss requires predict_proba support
‘precision’ etc. metrics.precision_score suffixes apply as with ‘f1’
‘recall’ etc. metrics.recall_score suffixes apply as with ‘f1’
‘roc_auc’ metrics.roc_auc_score
Clustering
‘adjusted_mutual_info_score’ metrics.adjusted_mutual_info_score
‘adjusted_rand_score’ metrics.adjusted_rand_score
‘completeness_score’ metrics.completeness_score
‘fowlkes_mallows_score’ metrics.fowlkes_mallows_score
‘homogeneity_score’ metrics.homogeneity_score
‘mutual_info_score’ metrics.mutual_info_score
‘normalized_mutual_info_score’ metrics.normalized_mutual_info_score
‘v_measure_score’ metrics.v_measure_score
Regression
‘explained_variance’ metrics.explained_variance_score
‘neg_mean_absolute_error’ metrics.mean_absolute_error
‘neg_mean_squared_error’ metrics.mean_squared_error
‘neg_mean_squared_log_error’ metrics.mean_squared_log_error
‘neg_median_absolute_error’ metrics.median_absolute_error
‘r2’ metrics.r2_score
Usage examples:

>>> from sklearn import svm, datasets

>>> clf = svm.SVC(gamma='scale', random_state=0)
>>> cross_val_score(clf, X, y, scoring='recall_macro',
... cv=5)
array([0.96..., 0.96..., 0.96..., 0.93..., 1. ])
>>> model = svm.SVC()
>>> cross_val_score(model, X, y, cv=5, scoring='wrong_choice')
Traceback (most recent call last):
ValueError: 'wrong_choice' is not a valid scoring value. Use sorted(sklearn.metrics.
˓→SCORERS.keys()) to get valid options.
Note: The values listed by the ValueError exception correspond to the functions measuring prediction accuracy
described in the following sections. The scorer objects for those functions are stored in the dictionary sklearn.
metrics.SCORERS.
Defining your scoring strategy from metric functions
The module sklearn.metrics also exposes a set of simple functions measuring a prediction error given ground
truth and prediction:
• functions ending with _score return a value to maximize, the higher the better.
• functions ending with _error or _loss return a value to minimize, the lower the better. When converting into
a scorer object using make_scorer, set the greater_is_better parameter to False (True by default; see
the parameter description below).
Metrics available for various machine learning tasks are detailed in sections below.
Many metrics are not given names to be used as scoring values, sometimes because they require additional param-
eters, such as fbeta_score. In such cases, you need to generate an appropriate scoring object. The simplest way
to generate a callable object for scoring is by using make_scorer. That function converts metrics into callables that
can be used for model evaluation.
One typical use case is to wrap an existing metric function from the library with non-default values for its parameters,
such as the beta parameter for the fbeta_score function:
>>> from sklearn.metrics import fbeta_score, make_scorer

>>> ftwo_scorer = make_scorer(fbeta_score, beta=2)
>>> grid = GridSearchCV(LinearSVC(), param_grid={'C': [1, 10]},
... scoring=ftwo_scorer, cv=5)
The second use case is to build a completely custom scorer object from a simple python function using
make_scorer, which can take several parameters:
• the python function you want to use (my_custom_loss_func in the example below)
• whether the python function returns a score (greater_is_better=True, the default) or a loss
(greater_is_better=False). If a loss, the output of the python function is negated by the scorer ob-
ject, conforming to the cross validation convention that scorers return higher values for better models.

• for classification metrics only: whether the python function you provided requires continuous decision certain-
ties (needs_threshold=True). The default value is False.
• any additional parameters, such as beta or labels in f1_score.
Here is an example of building custom scorers, and of using the greater_is_better parameter:

>>> def my_custom_loss_func(y_true, y_pred):
... diff = np.abs(y_true - y_pred).max()
... return np.log1p(diff)
...
>>> # score will negate the return value of my_custom_loss_func,
>>> # which will be np.log(2), 0.693, given the values for X
>>> # and y defined below.
>>> score = make_scorer(my_custom_loss_func, greater_is_better=False)
>>> X = [[1], [1]]
>>> y = [0, 1]
>>> from sklearn.dummy import DummyClassifier
>>> clf = DummyClassifier(strategy='most_frequent', random_state=0)
>>> my_custom_loss_func(clf.predict(X), y)
0.69...
>>> score(clf, X, y)
-0.69...
Implementing your own scoring object
You can generate even more flexible model scorers by constructing your own scoring object from scratch, without using
the make_scorer factory. For a callable to be a scorer, it needs to meet the protocol specified by the following two
rules:
• It can be called with parameters (estimator, X, y), where estimator is the model that should be
evaluated, X is validation data, and y is the ground truth target for X (in the supervised case) or None (in the
unsupervised case).
• It returns a floating point number that quantifies the estimator prediction quality on X, with reference to y.
Again, by convention higher numbers are better, so if your scorer returns loss, that value should be negated.
Note: Using custom scorers in functions where n_jobs > 1

While defining the custom scoring function alongside the calling function should work out of the box with the default
joblib backend (loky), importing it from another module will be a more robust approach and work independently of
the joblib backend.
For example, to use, n_jobs greater than 1 in the example below, custom_scoring_function function is
saved in a user-created module (custom_scorer_module.py) and imported:
>>> from custom_scorer_module import custom_scoring_function

>>> cross_val_score(model,
... X_train,
... y_train,
... scoring=make_scorer(custom_scoring_function, greater_is_better=False),
... cv=5,
... n_jobs=-1)

Using multiple metric evaluation
Scikit-learn also permits evaluation of multiple metrics in GridSearchCV, RandomizedSearchCV and

cross_validate.
There are two ways to specify multiple scoring metrics for the scoring parameter:
• As an iterable of string metrics::
>>> scoring = ['accuracy', 'precision']
• As a dict mapping the scorer name to the scoring function::
>>> from sklearn.metrics import accuracy_score

>>> from sklearn.metrics import make_scorer
>>> scoring = {'accuracy': make_scorer(accuracy_score),
... 'prec': 'precision'}
Note that the dict values can either be scorer functions or one of the predefined metric strings.
Currently only those scorer functions that return a single score can be passed inside the dict. Scorer functions that
return multiple values are not permitted and will require a wrapper to return a single metric:

>>> from sklearn.metrics import confusion_matrix
>>> # A sample toy binary classification dataset
>>> X, y = datasets.make_classification(n_classes=2, random_state=0)
>>> svm = LinearSVC(random_state=0)
>>> def tn(y_true, y_pred): return confusion_matrix(y_true, y_pred)[0, 0]
>>> def fp(y_true, y_pred): return confusion_matrix(y_true, y_pred)[0, 1]
>>> def fn(y_true, y_pred): return confusion_matrix(y_true, y_pred)[1, 0]
>>> def tp(y_true, y_pred): return confusion_matrix(y_true, y_pred)[1, 1]
>>> scoring = {'tp': make_scorer(tp), 'tn': make_scorer(tn),
... 'fp': make_scorer(fp), 'fn': make_scorer(fn)}
>>> cv_results = cross_validate(svm.fit(X, y), X, y,
... scoring=scoring, cv=5)
>>> # Getting the test set true positive scores
>>> print(cv_results['test_tp'])
[10 9 8 7 8]
>>> # Getting the test set false negative scores
>>> print(cv_results['test_fn'])
[0 1 2 3 2]
Classification metrics
The sklearn.metrics module implements several loss, score, and utility functions to measure classification per-
formance. Some metrics might require probability estimates of the positive class, confidence values, or binary deci-
sions values. Most implementations allow each sample to provide a weighted contribution to the overall score, through
the sample_weight parameter.
Some of these are restricted to the binary classification case:
precision_recall_curve(y_true, probas_pred) Compute precision-recall pairs for different probability

thresholds
roc_curve(y_true, y_score[, pos_label, . . . ]) Compute Receiver operating characteristic (ROC)
balanced_accuracy_score(y_true, y_pred[, . . . ]) Compute the balanced accuracy

Others also work in the multiclass case:
cohen_kappa_score(y1, y2[, labels, weights, . . . ]) Cohen’s kappa: a statistic that measures inter-annotator
agreement.
confusion_matrix(y_true, y_pred[, labels, . . . ]) Compute confusion matrix to evaluate the accuracy of a
classification
hinge_loss(y_true, pred_decision[, labels, . . . ]) Average hinge loss (non-regularized)
matthews_corrcoef(y_true, y_pred[, . . . ]) Compute the Matthews correlation coefficient (MCC)
Some also work in the multilabel case:
accuracy_score(y_true, y_pred[, normalize, . . . ]) Accuracy classification score.

classification_report(y_true, y_pred[, . . . ]) Build a text report showing the main classification metrics
f1_score(y_true, y_pred[, labels, . . . ]) Compute the F1 score, also known as balanced F-score or
F-measure
fbeta_score(y_true, y_pred, beta[, labels, . . . ]) Compute the F-beta score
hamming_loss(y_true, y_pred[, labels, . . . ]) Compute the average Hamming loss.
jaccard_similarity_score(y_true, y_pred[, . . . ]) Jaccard similarity coefficient score
log_loss(y_true, y_pred[, eps, normalize, . . . ]) Log loss, aka logistic loss or cross-entropy loss.
precision_recall_fscore_support(y_true, Compute precision, recall, F-measure and support for each
y_pred) class
precision_score(y_true, y_pred[, labels, . . . ]) Compute the precision
recall_score(y_true, y_pred[, labels, . . . ]) Compute the recall
zero_one_loss(y_true, y_pred[, normalize, . . . ]) Zero-one classification loss.
And some work with binary and multilabel (but not multiclass) problems:
average_precision_score(y_true, y_score[, . . . ]) Compute average precision (AP) from prediction scores

roc_auc_score(y_true, y_score[, average, . . . ]) Compute Area Under the Receiver Operating Characteris-
tic Curve (ROC AUC) from prediction scores.
In the following sub-sections, we will describe each of those functions, preceded by some notes on common API and
metric definition.
From binary to multiclass and multilabel
Some metrics are essentially defined for binary classification tasks (e.g. f1_score, roc_auc_score). In these
cases, by default only the positive label is evaluated, assuming by default that the positive class is labelled 1 (though
this may be configurable through the pos_label parameter). In extending a binary metric to multiclass or multilabel
problems, the data is treated as a collection of binary problems, one for each class. There are then a number of ways
to average binary metric calculations across the set of classes, each of which may be useful in some scenario. Where
available, you should select among these using the average parameter.
• "macro" simply calculates the mean of the binary metrics, giving equal weight to each class. In problems
where infrequent classes are nonetheless important, macro-averaging may be a means of highlighting their
performance. On the other hand, the assumption that all classes are equally important is often untrue, such that
macro-averaging will over-emphasize the typically low performance on an infrequent class.
• "weighted" accounts for class imbalance by computing the average of binary metrics in which each class’s
score is weighted by its presence in the true data sample.

• "micro" gives each sample-class pair an equal contribution to the overall metric (except as a result of sample-
weight). Rather than summing the metric per class, this sums the dividends and divisors that make up the
per-class metrics to calculate an overall quotient. Micro-averaging may be preferred in multilabel settings,
including multiclass classification where a majority class is to be ignored.
• "samples" applies only to multilabel problems. It does not calculate a per-class measure, instead calculat-
ing the metric over the true and predicted classes for each sample in the evaluation data, and returning their
(sample_weight-weighted) average.
• Selecting average=None will return an array with the score for each class.
While multiclass data is provided to the metric, like binary targets, as an array of class labels, multilabel data is
specified as an indicator matrix, in which cell [i, j] has value 1 if sample i has label j and value 0 otherwise.
Accuracy score
The accuracy_score function computes the accuracy, either the fraction (default) or the count (normalize=False)
of correct predictions.
In multilabel classification, the function returns the subset accuracy. If the entire set of predicted labels for a sample
strictly match with the true set of labels, then the subset accuracy is 1.0; otherwise it is 0.0.
If 𝑦ˆ𝑖 is the predicted value of the 𝑖-th sample and 𝑦𝑖 is the corresponding true value, then the fraction of correct
predictions over 𝑛samples is defined as
𝑛samples −1
1 ∑︁
accuracy(𝑦, 𝑦ˆ) = 1(ˆ
𝑦𝑖 = 𝑦𝑖 )
𝑛samples 𝑖=0
where 1(𝑥) is the indicator function.

>>> y_pred = [0, 2, 1, 3]
>>> y_true = [0, 1, 2, 3]
>>> accuracy_score(y_true, y_pred)
0.5
>>> accuracy_score(y_true, y_pred, normalize=False)
2
In the multilabel case with binary label indicators:
>>> accuracy_score(np.array([[0, 1], [1, 1]]), np.ones((2, 2)))

0.5
Example:
• See Test with permutations the significance of a classification score for an example of accuracy score usage
using permutations of the dataset.
Balanced accuracy score
The balanced_accuracy_score function computes the balanced accuracy, which avoids inflated performance
estimates on imbalanced datasets. It is the macro-average of recall scores per class or, equivalently, raw accuracy

where each sample is weighted according to the inverse prevalence of its true class. Thus for balanced datasets, the
score is equal to accuracy.
In the binary case, balanced accuracy is equal to the arithmetic mean of sensitivity (true positive rate) and specificity
(true negative rate), or the area under the ROC curve with binary predictions rather than scores.
If the classifier performs equally well on either class, this term reduces to the conventional accuracy (i.e., the number
of correct predictions divided by the total number of predictions).
In contrast, if the conventional accuracy is above chance only because the classifier takes advantage of an imbalanced
1
test set, then the balanced accuracy, as appropriate, will drop to 𝑛_𝑐𝑙𝑎𝑠𝑠𝑒𝑠 .
1
The score ranges from 0 to 1, or when adjusted=True is used, it rescaled to the range 1−𝑛_𝑐𝑙𝑎𝑠𝑠𝑒𝑠 to 1, inclusive,
with performance at random scoring 0.
If 𝑦𝑖 is the true value of the 𝑖-th sample, and 𝑤𝑖 is the corresponding sample weight, then we adjust the sample weight
to:
𝑤𝑖
𝑤
ˆ𝑖 = ∑︀
𝑗 1(𝑦𝑗 = 𝑦𝑖 )𝑤𝑗
where 1(𝑥) is the indicator function. Given predicted 𝑦ˆ𝑖 for sample 𝑖, balanced accuracy is defined as:
1 ∑︁
balanced-accuracy(𝑦, 𝑦ˆ, 𝑤) = ∑︀ 1(ˆ
𝑦𝑖 = 𝑦𝑖 )𝑤
ˆ𝑖
𝑤
ˆ𝑖 𝑖
With adjusted=True, balanced accuracy reports the relative increase from balanced-accuracy(𝑦, 0, 𝑤) =
1
𝑛_𝑐𝑙𝑎𝑠𝑠𝑒𝑠 . In the binary case, this is also known as *Youden’s J statistic*, or informedness.
Note: The multiclass definition here seems the most reasonable extension of the metric used in binary classification,
though there is no certain consensus in the literature:
• Our definition: [Mosley2013], [Kelleher2015] and [Guyon2015], where [Guyon2015] adopt the adjusted ver-
sion to ensure that random predictions have a score of 0 and perfect predictions have a score of 1..
• Class balanced accuracy as described in [Mosley2013]: the minimum between the precision and the recall for
each class is computed. Those values are then averaged over the total number of classes to get the balanced
accuracy.
• Balanced Accuracy as described in [Urbanowicz2015]: the average of sensitivity and specificity is computed
for each class and then averaged over total number of classes.
References:
Cohen’s kappa
The function cohen_kappa_score computes Cohen’s kappa statistic. This measure is intended to compare label-
ings by different human annotators, not a classifier versus a ground truth.
The kappa score (see docstring) is a number between -1 and 1. Scores above .8 are generally considered good agree-
ment; zero or lower means no agreement (practically random labels).
Kappa scores can be computed for binary or multiclass problems, but not for multilabel problems (except by manually
computing a per-label score) and not for more than two annotators.

>>> from sklearn.metrics import cohen_kappa_score

>>> y_true = [2, 0, 2, 2, 0, 1]
>>> y_pred = [0, 0, 2, 2, 0, 2]
>>> cohen_kappa_score(y_true, y_pred)
0.4285714285714286
Confusion matrix
The confusion_matrix function evaluates classification accuracy by computing the confusion matrix with each
row corresponding to the true class <https://en.wikipedia.org/wiki/Confusion_matrix>‘_. (Wikipedia and other refer-
ences may use different convention for axes.)
By definition, entry 𝑖, 𝑗 in a confusion matrix is the number of observations actually in group 𝑖, but predicted to be in
group 𝑗. Here is an example:

>>> y_true = [2, 0, 2, 2, 0, 1]
>>> y_pred = [0, 0, 2, 2, 0, 2]
>>> confusion_matrix(y_true, y_pred)
array([[2, 0, 0],
[0, 0, 1],
[1, 0, 2]])
Here is a visual representation of such a confusion matrix (this figure comes from the Confusion matrix example):
For binary problems, we can

get counts of true negatives, false positives, false negatives and true positives as follows:
>>> y_true = [0, 0, 0, 1, 1, 1, 1, 1]

>>> y_pred = [0, 1, 0, 1, 0, 1, 0, 1]
>>> tn, fp, fn, tp = confusion_matrix(y_true, y_pred).ravel()
>>> tn, fp, fn, tp
(2, 1, 2, 3)

Example:
• See Confusion matrix for an example of using a confusion matrix to evaluate classifier output quality.
• See Recognizing hand-written digits for an example of using a confusion matrix to classify hand-written
digits.
• See Classification of text documents using sparse features for an example of using a confusion matrix to
classify text documents.
Classification report
The classification_report function builds a text report showing the main classification metrics. Here is a
small example with custom target_names and inferred labels:
>>> from sklearn.metrics import classification_report

>>> y_true = [0, 1, 2, 2, 0]
>>> y_pred = [0, 0, 2, 1, 0]
>>> target_names = ['class 0', 'class 1', 'class 2']
>>> print(classification_report(y_true, y_pred, target_names=target_names))
class 0 0.67 1.00 0.80 2

class 1 0.00 0.00 0.00 1
class 2 1.00 0.50 0.67 2
micro avg 0.60 0.60 0.60 5

macro avg 0.56 0.50 0.49 5
weighted avg 0.67 0.60 0.59 5
Example:
• See Recognizing hand-written digits for an example of classification report usage for hand-written digits.
• See Classification of text documents using sparse features for an example of classification report usage for
text documents.
• See Parameter estimation using grid search with cross-validation for an example of classification report usage
for grid search with nested cross-validation.
Hamming loss
The hamming_loss computes the average Hamming loss or Hamming distance between two sets of samples.
If 𝑦ˆ𝑗 is the predicted value for the 𝑗-th label of a given sample, 𝑦𝑗 is the corresponding true value, and 𝑛labels is the
number of classes or labels, then the Hamming loss 𝐿𝐻𝑎𝑚𝑚𝑖𝑛𝑔 between two samples is defined as:
1 ∑︁−1
𝑛labels
𝐿𝐻𝑎𝑚𝑚𝑖𝑛𝑔 (𝑦, 𝑦ˆ) = 𝑦𝑗 ̸= 𝑦𝑗 )
1(ˆ
𝑛labels 𝑗=0

>>> from sklearn.metrics import hamming_loss

>>> y_pred = [1, 2, 3, 4]
>>> y_true = [2, 2, 3, 4]
>>> hamming_loss(y_true, y_pred)
0.25
>>> hamming_loss(np.array([[0, 1], [1, 1]]), np.zeros((2, 2)))

0.75
Note: In multiclass classification, the Hamming loss corresponds to the Hamming distance between y_true and
y_pred which is similar to the Zero one loss function. However, while zero-one loss penalizes prediction sets that
do not strictly match true sets, the Hamming loss penalizes individual labels. Thus the Hamming loss, upper bounded
by the zero-one loss, is always between zero and one, inclusive; and predicting a proper subset or superset of the true
labels will give a Hamming loss between zero and one, exclusive.
Jaccard similarity coefficient score
The jaccard_similarity_score function computes the average (default) or sum of Jaccard similarity coeffi-
cients, also called the Jaccard index, between pairs of label sets.
The Jaccard similarity coefficient of the 𝑖-th samples, with a ground truth label set 𝑦𝑖 and predicted label set 𝑦ˆ𝑖 , is
defined as
|𝑦𝑖 ∩ 𝑦ˆ𝑖 |
𝐽(𝑦𝑖 , 𝑦ˆ𝑖 ) = .
|𝑦𝑖 ∪ 𝑦ˆ𝑖 |
In binary and multiclass classification, the Jaccard similarity coefficient score is equal to the classification accuracy.

>>> from sklearn.metrics import jaccard_similarity_score
>>> y_pred = [0, 2, 1, 3]
>>> y_true = [0, 1, 2, 3]
>>> jaccard_similarity_score(y_true, y_pred)
0.5
>>> jaccard_similarity_score(y_true, y_pred, normalize=False)
2
>>> jaccard_similarity_score(np.array([[0, 1], [1, 1]]), np.ones((2, 2)))

0.75
Precision, recall and F-measures
Intuitively, precision is the ability of the classifier not to label as positive a sample that is negative, and recall is the
ability of the classifier to find all the positive samples.
The F-measure (𝐹𝛽 and 𝐹1 measures) can be interpreted as a weighted harmonic mean of the precision and recall. A
𝐹𝛽 measure reaches its best value at 1 and its worst score at 0. With 𝛽 = 1, 𝐹𝛽 and 𝐹1 are equivalent, and the recall
and the precision are equally important.

The precision_recall_curve computes a precision-recall curve from the ground truth label and a score given
by the classifier by varying a decision threshold.
The average_precision_score function computes the average precision (AP) from prediction scores. The
value is between 0 and 1 and higher is better. AP is defined as
∑︁
AP = (𝑅𝑛 − 𝑅𝑛−1 )𝑃𝑛
𝑛
where 𝑃𝑛 and 𝑅𝑛 are the precision and recall at the nth threshold. With random predictions, the AP is the fraction of
positive samples.
References [Manning2008] and [Everingham2010] present alternative variants of AP that interpolate the precision-
recall curve. Currently, average_precision_score does not implement any interpolated variant. References
[Davis2006] and [Flach2015] describe why a linear interpolation of points on the precision-recall curve provides an
overly-optimistic measure of classifier performance. This linear interpolation is used when computing area under the
curve with the trapezoidal rule in auc.
Several functions allow you to analyze the precision, recall and F-measures score:
average_precision_score(y_true, y_score[, . . . ]) Compute average precision (AP) from prediction scores

f1_score(y_true, y_pred[, labels, . . . ]) Compute the F1 score, also known as balanced F-score or
F-measure
fbeta_score(y_true, y_pred, beta[, labels, . . . ]) Compute the F-beta score
precision_recall_curve(y_true, probas_pred) Compute precision-recall pairs for different probability
thresholds
precision_recall_fscore_support(y_true, Compute precision, recall, F-measure and support for each
y_pred) class
precision_score(y_true, y_pred[, labels, . . . ]) Compute the precision
recall_score(y_true, y_pred[, labels, . . . ]) Compute the recall
Note that the precision_recall_curve function is restricted to the binary case. The
average_precision_score function works only in binary classification and multilabel indicator format.
Examples:
• See Classification of text documents using sparse features for an example of f1_score usage to classify
text documents.
• See Parameter estimation using grid search with cross-validation for an example of precision_score
and recall_score usage to estimate parameters using grid search with nested cross-validation.
• See Precision-Recall for an example of precision_recall_curve usage to evaluate classifier output
quality.
References:
Binary classification
In a binary classification task, the terms ‘’positive” and ‘’negative” refer to the classifier’s prediction, and the terms
‘’true” and ‘’false” refer to whether that prediction corresponds to the external judgment (sometimes known as the
‘’observation’‘). Given these definitions, we can formulate the following table:

Actual class (observation)

Predicted class (expectation) tp (true positive) Correct result fp (false positive) Unexpected result
fn (false negative) Missing result tn (true negative) Correct absence of result
In this context, we can define the notions of precision, recall and F-measure:
𝑡𝑝
precision = ,
𝑡𝑝 + 𝑓 𝑝
𝑡𝑝
recall = ,
𝑡𝑝 + 𝑓 𝑛
precision × recall
𝐹𝛽 = (1 + 𝛽 2 ) .
𝛽 2 precision + recall
Here are some small examples in binary classification:

>>> y_pred = [0, 1, 0, 0]
>>> y_true = [0, 1, 0, 1]
>>> metrics.precision_score(y_true, y_pred)
1.0
>>> metrics.recall_score(y_true, y_pred)
0.5
>>> metrics.f1_score(y_true, y_pred)
0.66...
>>> metrics.fbeta_score(y_true, y_pred, beta=0.5)
0.83...
>>> metrics.fbeta_score(y_true, y_pred, beta=1)
0.66...
>>> metrics.fbeta_score(y_true, y_pred, beta=2)
0.55...
>>> metrics.precision_recall_fscore_support(y_true, y_pred, beta=0.5)
(array([0.66..., 1. ]), array([1. , 0.5]), array([0.71..., 0.83...]), array([2,
˓→ 2]))

>>> from sklearn.metrics import precision_recall_curve
>>> from sklearn.metrics import average_precision_score
>>> y_true = np.array([0, 0, 1, 1])
>>> y_scores = np.array([0.1, 0.4, 0.35, 0.8])
>>> precision, recall, threshold = precision_recall_curve(y_true, y_scores)
>>> precision
array([0.66..., 0.5 , 1. , 1. ])
>>> recall
array([1. , 0.5, 0.5, 0. ])
>>> threshold
array([0.35, 0.4 , 0.8 ])
>>> average_precision_score(y_true, y_scores)
0.83...
Multiclass and multilabel classification
In multiclass and multilabel classification task, the notions of precision, recall, and F-measures can be ap-
plied to each label independently. There are a few ways to combine results across labels, specified by the

average argument to the average_precision_score (multilabel only), f1_score, fbeta_score,

precision_recall_fscore_support, precision_score and recall_score functions, as described
above. Note that if all labels are included, “micro”-averaging in a multiclass setting will produce precision, recall and
𝐹 that are all identical to accuracy. Also note that “weighted” averaging may produce an F-score that is not between
precision and recall.
To make this more explicit, consider the following notation:
• 𝑦 the set of predicted (𝑠𝑎𝑚𝑝𝑙𝑒, 𝑙𝑎𝑏𝑒𝑙) pairs
• 𝑦ˆ the set of true (𝑠𝑎𝑚𝑝𝑙𝑒, 𝑙𝑎𝑏𝑒𝑙) pairs
• 𝐿 the set of labels
• 𝑆 the set of samples
• 𝑦𝑠 the subset of 𝑦 with sample 𝑠, i.e. 𝑦𝑠 := {(𝑠′ , 𝑙) ∈ 𝑦|𝑠′ = 𝑠}
• 𝑦𝑙 the subset of 𝑦 with label 𝑙
• similarly, 𝑦ˆ𝑠 and 𝑦ˆ𝑙 are subsets of 𝑦ˆ
|𝐴∩𝐵|
• 𝑃 (𝐴, 𝐵) := |𝐴|
|𝐴∩𝐵|
• 𝑅(𝐴, 𝐵) := |𝐵| (Conventions vary on handling 𝐵 = ∅; this implementation uses 𝑅(𝐴, 𝐵) := 0, and similar
for 𝑃 .)
• 𝐹𝛽 (𝐴, 𝐵) := 1 + 𝛽 2 𝛽𝑃2 𝑃(𝐴,𝐵)×𝑅(𝐴,𝐵)
(︀ )︀
(𝐴,𝐵)+𝑅(𝐴,𝐵)
Then the metrics are defined as:
average Precision Recall F_beta

"micro" 𝑃 (𝑦,
∑︀𝑦ˆ) 𝑅(𝑦,
∑︀𝑦ˆ) 𝐹𝛽 (𝑦,
∑︀ 𝑦ˆ)
1 1 1
"samples" |𝑆| ∑︀𝑠∈𝑆 𝑃 (𝑦𝑠 , 𝑦 ˆ𝑠 ) |𝑆| ∑︀𝑠∈𝑆 𝑅(𝑦𝑠 , 𝑦 ˆ𝑠 ) |𝑆| ∑︀𝑠∈𝑆 𝐹𝛽 (𝑦𝑠 , 𝑦 ˆ𝑠 )
1 1 1
"macro" |𝐿| 𝑙∈𝐿 𝑃 (𝑦𝑙 , 𝑦
ˆ𝑙 ) |𝐿| 𝑙∈𝐿 𝑅(𝑦𝑙 , 𝑦
ˆ𝑙 ) |𝐿| 𝑙∈𝐿 𝐹𝛽 (𝑦𝑙 , 𝑦
ˆ𝑙 )
∑︀ 1 ∑︀ 1 ∑︀ 1
∑︀ ∑︀ ∑︀
"weighted"
𝑙∈𝐿 |^
𝑦𝑙 | 𝑙∈𝐿 𝑦𝑙 | 𝑃 (𝑦𝑙 , 𝑦
|ˆ ˆ𝑙 )
𝑙∈𝐿 |^
𝑦𝑙 | 𝑙∈𝐿 𝑦𝑙 | 𝑅(𝑦𝑙 , 𝑦
|ˆ ˆ𝑙 )
𝑙∈𝐿 |^
𝑦𝑙 | 𝑙∈𝐿 𝑙 | 𝐹𝛽 (𝑦𝑙 , 𝑦
|ˆ
𝑦 ˆ𝑙 )
None ⟨𝑃 (𝑦𝑙 , 𝑦ˆ𝑙 )|𝑙 ∈ 𝐿⟩ ⟨𝑅(𝑦𝑙 , 𝑦ˆ𝑙 )|𝑙 ∈ 𝐿⟩ ⟨𝐹𝛽 (𝑦𝑙 , 𝑦ˆ𝑙 )|𝑙 ∈ 𝐿⟩

>>> y_true = [0, 1, 2, 0, 1, 2]
>>> y_pred = [0, 2, 1, 0, 0, 1]
>>> metrics.precision_score(y_true, y_pred, average='macro')
0.22...
>>> metrics.recall_score(y_true, y_pred, average='micro')
...
0.33...
>>> metrics.f1_score(y_true, y_pred, average='weighted')
0.26...
>>> metrics.fbeta_score(y_true, y_pred, average='macro', beta=0.5)
0.23...
>>> metrics.precision_recall_fscore_support(y_true, y_pred, beta=0.5, average=None)
...
(array([0.66..., 0. , 0. ]), array([1., 0., 0.]), array([0.71..., 0.
˓→ , 0. ]), array([2, 2, 2]...))
For multiclass classification with a “negative class”, it is possible to exclude some labels:
>>> metrics.recall_score(y_true, y_pred, labels=[1, 2], average='micro')
... # excluding 0, no labels were correctly recalled
0.0

Similarly, labels not present in the data sample may be accounted for in macro-averaging.
>>> metrics.precision_score(y_true, y_pred, labels=[0, 1, 2, 3], average='macro')
...
0.166...
Hinge loss
The hinge_loss function computes the average distance between the model and the data using hinge loss, a one-
sided metric that considers only prediction errors. (Hinge loss is used in maximal margin classifiers such as support
vector machines.)
If the labels are encoded with +1 and -1, 𝑦: is the true value, and 𝑤 is the predicted decisions as output by
decision_function, then the hinge loss is defined as:
𝐿Hinge (𝑦, 𝑤) = max {1 − 𝑤𝑦, 0} = |1 − 𝑤𝑦|+
If there are more than two labels, hinge_loss uses a multiclass variant due to Crammer & Singer. Here is the paper
describing it.
If 𝑦𝑤 is the predicted decision for true label and 𝑦𝑡 is the maximum of the predicted decisions for all other labels,
where predicted decisions are output by decision function, then multiclass hinge loss is defined by:
𝐿Hinge (𝑦𝑤 , 𝑦𝑡 ) = max {1 + 𝑦𝑡 − 𝑦𝑤 , 0}
Here a small example demonstrating the use of the hinge_loss function with a svm classifier in a binary class
problem:
>>> from sklearn.metrics import hinge_loss
>>> X = [[0], [1]]
>>> y = [-1, 1]
>>> est = svm.LinearSVC(random_state=0)
>>> est.fit(X, y)
multi_class='ovr', penalty='l2', random_state=0, tol=0.0001,
verbose=0)
>>> pred_decision = est.decision_function([[-2], [3], [0.5]])
>>> pred_decision
array([-2.18..., 2.36..., 0.09...])
>>> hinge_loss([-1, 1, 1], pred_decision)
0.3...
Here is an example demonstrating the use of the hinge_loss function with a svm classifier in a multiclass problem:
>>> X = np.array([[0], [1], [2], [3]])
>>> Y = np.array([0, 1, 2, 3])
>>> labels = np.array([0, 1, 2, 3])
>>> est = svm.LinearSVC()
>>> est.fit(X, Y)
verbose=0)

>>> pred_decision = est.decision_function([[-1], [2], [3]])

>>> y_true = [0, 2, 3]
>>> hinge_loss(y_true, pred_decision, labels)
0.56...
Log loss
Log loss, also called logistic regression loss or cross-entropy loss, is defined on probability estimates. It is commonly
used in (multinomial) logistic regression and neural networks, as well as in some variants of expectation-maximization,
and can be used to evaluate the probability outputs (predict_proba) of a classifier instead of its discrete predic-
tions.
For binary classification with a true label 𝑦 ∈ {0, 1} and a probability estimate 𝑝 = Pr(𝑦 = 1), the log loss per sample
is the negative log-likelihood of the classifier given the true label:
𝐿log (𝑦, 𝑝) = − log Pr(𝑦|𝑝) = −(𝑦 log(𝑝) + (1 − 𝑦) log(1 − 𝑝))
This extends to the multiclass case as follows. Let the true labels for a set of samples be encoded as a 1-of-K binary
indicator matrix 𝑌 , i.e., 𝑦𝑖,𝑘 = 1 if sample 𝑖 has label 𝑘 taken from a set of 𝐾 labels. Let 𝑃 be a matrix of probability
estimates, with 𝑝𝑖,𝑘 = Pr(𝑡𝑖,𝑘 = 1). Then the log loss of the whole set is
𝑁 −1 𝐾−1
1 ∑︁ ∑︁
𝐿log (𝑌, 𝑃 ) = − log Pr(𝑌 |𝑃 ) = − 𝑦𝑖,𝑘 log 𝑝𝑖,𝑘
𝑁 𝑖=0
𝑘=0
To see how this generalizes the binary log loss given above, note that in the binary case, 𝑝𝑖,0 = 1 − 𝑝𝑖,1 and 𝑦𝑖,0 =
1 − 𝑦𝑖,1 , so expanding the inner sum over 𝑦𝑖,𝑘 ∈ {0, 1} gives the binary log loss.
The log_loss function computes log loss given a list of ground-truth labels and a probability matrix, as returned by
an estimator’s predict_proba method.
>>> from sklearn.metrics import log_loss
>>> y_true = [0, 0, 1, 1]
>>> y_pred = [[.9, .1], [.8, .2], [.3, .7], [.01, .99]]
>>> log_loss(y_true, y_pred)
0.1738...
The first [.9, .1] in y_pred denotes 90% probability that the first sample has label 0. The log loss is non-negative.
Matthews correlation coefficient
The matthews_corrcoef function computes the Matthew’s correlation coefficient (MCC) for binary classes.
Quoting Wikipedia:
“The Matthews correlation coefficient is used in machine learning as a measure of the quality of binary
(two-class) classifications. It takes into account true and false positives and negatives and is generally
regarded as a balanced measure which can be used even if the classes are of very different sizes. The
MCC is in essence a correlation coefficient value between -1 and +1. A coefficient of +1 represents
a perfect prediction, 0 an average random prediction and -1 an inverse prediction. The statistic is also
known as the phi coefficient.”
In the binary (two-class) case, 𝑡𝑝, 𝑡𝑛, 𝑓 𝑝 and 𝑓 𝑛 are respectively the number of true positives, true negatives, false
positives and false negatives, the MCC is defined as
𝑡𝑝 × 𝑡𝑛 − 𝑓 𝑝 × 𝑓 𝑛
𝑀 𝐶𝐶 = √︀ .
(𝑡𝑝 + 𝑓 𝑝)(𝑡𝑝 + 𝑓 𝑛)(𝑡𝑛 + 𝑓 𝑝)(𝑡𝑛 + 𝑓 𝑛)

In the multiclass case, the Matthews correlation coefficient can be defined in terms of a confusion_matrix 𝐶 for
𝐾 classes. To simplify the definition consider the following intermediate variables:
∑︀𝐾
• 𝑡𝑘 = 𝑖 𝐶𝑖𝑘 the number of times class 𝑘 truly occurred,
∑︀𝐾
• 𝑝𝑘 = 𝑖 𝐶𝑘𝑖 the number of times class 𝑘 was predicted,
∑︀𝐾
• 𝑐 = 𝑘 𝐶𝑘𝑘 the total number of samples correctly predicted,
∑︀𝐾 ∑︀𝐾
• 𝑠= 𝑖 𝑗 𝐶𝑖𝑗 the total number of samples.
Then the multiclass MCC is defined as:

∑︀𝐾
𝑐 × 𝑠 − 𝑘 𝑝 𝑘 × 𝑡𝑘
𝑀 𝐶𝐶 = √︁ ∑︀𝐾 ∑︀𝐾
(𝑠2 − 𝑘 𝑝2𝑘 ) × (𝑠2 − 𝑘 𝑡2𝑘 )
When there are more than two labels, the value of the MCC will no longer range between -1 and +1. Instead the
minimum value will be somewhere between -1 and 0 depending on the number and distribution of ground true labels.
The maximum value is always +1.
Here is a small example illustrating the usage of the matthews_corrcoef function:
>>> from sklearn.metrics import matthews_corrcoef

>>> y_true = [+1, +1, +1, -1]
>>> y_pred = [+1, -1, +1, +1]
>>> matthews_corrcoef(y_true, y_pred)
-0.33...
Receiver operating characteristic (ROC)
The function roc_curve computes the receiver operating characteristic curve, or ROC curve. Quoting Wikipedia :
“A receiver operating characteristic (ROC), or simply ROC curve, is a graphical plot which illustrates
the performance of a binary classifier system as its discrimination threshold is varied. It is created by
plotting the fraction of true positives out of the positives (TPR = true positive rate) vs. the fraction of false
positives out of the negatives (FPR = false positive rate), at various threshold settings. TPR is also known
as sensitivity, and FPR is one minus the specificity or true negative rate.”
This function requires the true binary value and the target scores, which can either be probability estimates of the
positive class, confidence values, or binary decisions. Here is a small example of how to use the roc_curve function:

>>> from sklearn.metrics import roc_curve
>>> y = np.array([1, 1, 2, 2])
>>> scores = np.array([0.1, 0.4, 0.35, 0.8])
>>> fpr, tpr, thresholds = roc_curve(y, scores, pos_label=2)
>>> fpr
array([0. , 0. , 0.5, 0.5, 1. ])
>>> tpr
array([0. , 0.5, 0.5, 1. , 1. ])
>>> thresholds
array([1.8 , 0.8 , 0.4 , 0.35, 0.1 ])

This figure shows an example of such an ROC curve:

The roc_auc_score function computes the area under the receiver operating characteristic (ROC) curve, which is
also denoted by AUC or AUROC. By computing the area under the roc curve, the curve information is summarized in
one number. For more information see the Wikipedia article on AUC.

>>> from sklearn.metrics import roc_auc_score
>>> y_true = np.array([0, 0, 1, 1])
>>> y_scores = np.array([0.1, 0.4, 0.35, 0.8])
>>> roc_auc_score(y_true, y_scores)
0.75
In multi-label classification, the roc_auc_score function is extended by averaging over the labels as above.
Compared to metrics such as the subset accuracy, the Hamming loss, or the F1 score, ROC doesn’t require optimizing a
threshold for each label. The roc_auc_score function can also be used in multi-class classification, if the predicted
outputs have been binarized.
In applications where a high false positive rate is not tolerable the parameter max_fpr
of roc_auc_score can be used to summarize the ROC curve up to the given limit.

Examples:
• See Receiver Operating Characteristic (ROC) for an example of using ROC to evaluate the quality of the
output of a classifier.
• See Receiver Operating Characteristic (ROC) with cross validation for an example of using ROC to evaluate
classifier output quality, using cross-validation.
• See Species distribution modeling for an example of using ROC to model species distribution.
Zero one loss
The zero_one_loss function computes the sum or the average of the 0-1 classification loss (𝐿0−1 ) over 𝑛samples .
By default, the function normalizes over the sample. To get the sum of the 𝐿0−1 , set normalize to False.
In multilabel classification, the zero_one_loss scores a subset as one if its labels strictly match the predictions,
and as a zero if there are any errors. By default, the function returns the percentage of imperfectly predicted subsets.
To get the count of such subsets instead, set normalize to False
If 𝑦ˆ𝑖 is the predicted value of the 𝑖-th sample and 𝑦𝑖 is the corresponding true value, then the 0-1 loss 𝐿0−1 is defined
as:
𝑦𝑖 ̸= 𝑦𝑖 )
𝐿0−1 (𝑦𝑖 , 𝑦ˆ𝑖 ) = 1(ˆ
>>> from sklearn.metrics import zero_one_loss

>>> y_pred = [1, 2, 3, 4]
>>> y_true = [2, 2, 3, 4]
>>> zero_one_loss(y_true, y_pred)
0.25

>>> zero_one_loss(y_true, y_pred, normalize=False)

1
In the multilabel case with binary label indicators, where the first label set [0,1] has an error:
>>> zero_one_loss(np.array([[0, 1], [1, 1]]), np.ones((2, 2)))

0.5
>>> zero_one_loss(np.array([[0, 1], [1, 1]]), np.ones((2, 2)), normalize=False)

1
Example:
• See Recursive feature elimination with cross-validation for an example of zero one loss usage to perform
recursive feature elimination with cross-validation.
Brier score loss
The brier_score_loss function computes the Brier score for binary classes. Quoting Wikipedia:
“The Brier score is a proper score function that measures the accuracy of probabilistic predictions. It is
applicable to tasks in which predictions must assign probabilities to a set of mutually exclusive discrete
outcomes.”
This function returns a score of the mean square difference between the actual outcome and the predicted probability
of the possible outcome. The actual outcome has to be 1 or 0 (true or false), while the predicted probability of the
actual outcome can be a value between 0 and 1.
The brier score loss is also between 0 to 1 and the lower the score (the mean square difference is smaller), the more
accurate the prediction is. It can be thought of as a measure of the “calibration” of a set of probabilistic predictions.
𝑁
1 ∑︁
𝐵𝑆 = (𝑓𝑡 − 𝑜𝑡 )2
𝑁 𝑡=1
where : 𝑁 is the total number of predictions, 𝑓𝑡 is the predicted probability of the actual outcome 𝑜𝑡 .
Here is a small example of usage of this function::

>>> from sklearn.metrics import brier_score_loss
>>> y_true = np.array([0, 1, 1, 0])
>>> y_true_categorical = np.array(["spam", "ham", "ham", "spam"])
>>> y_prob = np.array([0.1, 0.9, 0.8, 0.4])
>>> y_pred = np.array([0, 1, 1, 0])
>>> brier_score_loss(y_true, y_prob)
0.055
>>> brier_score_loss(y_true, 1 - y_prob, pos_label=0)
0.055
>>> brier_score_loss(y_true_categorical, y_prob, pos_label="ham")
0.055
>>> brier_score_loss(y_true, y_prob > 0.5)
0.0

Example:
• See Probability calibration of classifiers for an example of Brier score loss usage to perform probability
calibration of classifiers.
References:
• G. Brier, Verification of forecasts expressed in terms of probability, Monthly weather review 78.1 (1950)
Multilabel ranking metrics
In multilabel learning, each sample can have any number of ground truth labels associated with it. The goal is to give
high scores and better rank to the ground truth labels.
Coverage error
The coverage_error function computes the average number of labels that have to be included in the final predic-
tion such that all true labels are predicted. This is useful if you want to know how many top-scored-labels you have
to predict in average without missing any true one. The best value of this metrics is thus the average number of true
labels.
Note: Our implementation’s score is 1 greater than the one given in Tsoumakas et al., 2010. This extends it to handle
the degenerate case in which an instance has 0 true labels.
𝑛samples ×𝑛labels
Formally, given a binary indicator matrix of the ground truth labels 𝑦 ∈ {0, 1} and the score associated
with each label 𝑓ˆ ∈ R𝑛samples ×𝑛labels , the coverage is defined as
𝑛samples −1
1 ∑︁
𝑐𝑜𝑣𝑒𝑟𝑎𝑔𝑒(𝑦, 𝑓ˆ) = max rank𝑖𝑗
𝑛samples 𝑖=0
𝑗:𝑦𝑖𝑗 =1
⃒{︁ }︁⃒
with rank𝑖𝑗 = ⃒ 𝑘 : 𝑓ˆ𝑖𝑘 ≥ 𝑓ˆ𝑖𝑗 ⃒. Given the rank definition, ties in y_scores are broken by giving the maximal rank
⃒ ⃒
that would have been assigned to all tied values.
Here is a small example of usage of this function:

>>> from sklearn.metrics import coverage_error
>>> y_true = np.array([[1, 0, 0], [0, 0, 1]])
>>> y_score = np.array([[0.75, 0.5, 1], [1, 0.2, 0.1]])
>>> coverage_error(y_true, y_score)
2.5
Label ranking average precision
The label_ranking_average_precision_score function implements label ranking average precision

(LRAP). This metric is linked to the average_precision_score function, but is based on the notion of la-
bel ranking instead of precision and recall.

Label ranking average precision (LRAP) averages over the samples the answer to the following question: for each
ground truth label, what fraction of higher-ranked labels were true labels? This performance measure will be higher if
you are able to give better rank to the labels associated with each sample. The obtained score is always strictly greater
than 0, and the best value is 1. If there is exactly one relevant label per sample, label ranking average precision is
equivalent to the mean reciprocal rank.
with each label 𝑓ˆ ∈ R𝑛samples ×𝑛labels , the average precision is defined as
𝑛samples −1
1 ∑︁ 1 ∑︁ |ℒ𝑖𝑗 |
𝐿𝑅𝐴𝑃 (𝑦, 𝑓ˆ) =
𝑛samples 𝑖=0
||𝑦𝑖 ||0 𝑗:𝑦 =1 rank𝑖𝑗
𝑖𝑗
{︁ }︁ ⃒{︁ }︁⃒
where ℒ𝑖𝑗 = 𝑘 : 𝑦𝑖𝑘 = 1, 𝑓ˆ𝑖𝑘 ≥ 𝑓ˆ𝑖𝑗 , rank𝑖𝑗 = ⃒ 𝑘 : 𝑓ˆ𝑖𝑘 ≥ 𝑓ˆ𝑖𝑗 ⃒, | · | computes the cardinality of the set (i.e., the
⃒ ⃒
number of elements in the set), and || · ||0 is the ℓ0 “norm” (which computes the number of nonzero elements in a
vector).

>>> from sklearn.metrics import label_ranking_average_precision_score
>>> y_true = np.array([[1, 0, 0], [0, 0, 1]])
>>> y_score = np.array([[0.75, 0.5, 1], [1, 0.2, 0.1]])
>>> label_ranking_average_precision_score(y_true, y_score)
0.416...
Ranking loss
The label_ranking_loss function computes the ranking loss which averages over the samples the number of
label pairs that are incorrectly ordered, i.e. true labels have a lower score than false labels, weighted by the inverse of
the number of ordered pairs of false and true labels. The lowest achievable ranking loss is zero.
with each label 𝑓ˆ ∈ R𝑛samples ×𝑛labels , the ranking loss is defined as
𝑛samples −1
1 ∑︁ 1 ⃒{︁ }︁⃒
𝑟𝑎𝑛𝑘𝑖𝑛𝑔_𝑙𝑜𝑠𝑠(𝑦, 𝑓ˆ) = ⃒ (𝑘, 𝑙) : 𝑓ˆ𝑖𝑘 ≤ 𝑓ˆ𝑖𝑙 , 𝑦𝑖𝑘 = 1, 𝑦𝑖𝑙 = 0 ⃒
⃒ ⃒
𝑛samples 𝑖=0
||𝑦𝑖 ||0 (𝑛labels − ||𝑦𝑖 ||0 )
where | · | computes the cardinality of the set (i.e., the number of elements in the set) and || · ||0 is the ℓ0 “norm” (which
computes the number of nonzero elements in a vector).

>>> from sklearn.metrics import label_ranking_loss
>>> y_true = np.array([[1, 0, 0], [0, 0, 1]])
>>> y_score = np.array([[0.75, 0.5, 1], [1, 0.2, 0.1]])
>>> label_ranking_loss(y_true, y_score)
0.75...
>>> # With the following prediction, we have perfect and minimal loss
>>> y_score = np.array([[1.0, 0.1, 0.2], [0.1, 0.2, 0.9]])
>>> label_ranking_loss(y_true, y_score)
0.0

References:
• Tsoumakas, G., Katakis, I., & Vlahavas, I. (2010). Mining multi-label data. In Data mining and knowledge
discovery handbook (pp. 667-685). Springer US.
Regression metrics
The sklearn.metrics module implements several loss, score, and utility functions to measure regression
performance. Some of those have been enhanced to handle the multioutput case: mean_squared_error,
mean_absolute_error, explained_variance_score and r2_score.
These functions have an multioutput keyword argument which specifies the way the scores or losses for each
individual target should be averaged. The default is 'uniform_average', which specifies a uniformly weighted
mean over outputs. If an ndarray of shape (n_outputs,) is passed, then its entries are interpreted as weights
and an according weighted average is returned. If multioutput is 'raw_values' is specified, then all unaltered
individual scores or losses will be returned in an array of shape (n_outputs,).
The r2_score and explained_variance_score accept an additional value 'variance_weighted' for
the multioutput parameter. This option leads to a weighting of each individual score by the variance of the
corresponding target variable. This setting quantifies the globally captured unscaled variance. If the target vari-
ables are of different scale, then this score puts more importance on well explaining the higher variance variables.
multioutput='variance_weighted' is the default value for r2_score for backward compatibility. This
will be changed to uniform_average in the future.
Explained variance score
The explained_variance_score computes the explained variance regression score.

If 𝑦ˆ is the estimated target output, 𝑦 the corresponding (correct) target output, and 𝑉 𝑎𝑟 is Variance, the square of the
standard deviation, then the explained variance is estimated as follow:
𝑉 𝑎𝑟{𝑦 − 𝑦ˆ}
𝑒𝑥𝑝𝑙𝑎𝑖𝑛𝑒𝑑_𝑣𝑎𝑟𝑖𝑎𝑛𝑐𝑒(𝑦, 𝑦ˆ) = 1 −
𝑉 𝑎𝑟{𝑦}
The best possible score is 1.0, lower values are worse.

Here is a small example of usage of the explained_variance_score function:
>>> from sklearn.metrics import explained_variance_score

>>> y_true = [3, -0.5, 2, 7]
>>> y_pred = [2.5, 0.0, 2, 8]
>>> explained_variance_score(y_true, y_pred)
0.957...
>>> y_true = [[0.5, 1], [-1, 1], [7, -6]]
>>> y_pred = [[0, 2], [-1, 2], [8, -5]]
>>> explained_variance_score(y_true, y_pred, multioutput='raw_values')
...
array([0.967..., 1. ])
>>> explained_variance_score(y_true, y_pred, multioutput=[0.3, 0.7])
...
0.990...

Mean absolute error
The mean_absolute_error function computes mean absolute error, a risk metric corresponding to the expected
value of the absolute error loss or 𝑙1-norm loss.
If 𝑦ˆ𝑖 is the predicted value of the 𝑖-th sample, and 𝑦𝑖 is the corresponding true value, then the mean absolute error
(MAE) estimated over 𝑛samples is defined as
𝑛samples −1
1 ∑︁
MAE(𝑦, 𝑦ˆ) = |𝑦𝑖 − 𝑦ˆ𝑖 | .
𝑛samples 𝑖=0
Here is a small example of usage of the mean_absolute_error function:
>>> from sklearn.metrics import mean_absolute_error

>>> y_true = [3, -0.5, 2, 7]
>>> y_pred = [2.5, 0.0, 2, 8]
>>> mean_absolute_error(y_true, y_pred)
0.5
>>> y_true = [[0.5, 1], [-1, 1], [7, -6]]
>>> y_pred = [[0, 2], [-1, 2], [8, -5]]
0.75
>>> mean_absolute_error(y_true, y_pred, multioutput='raw_values')
array([0.5, 1. ])
>>> mean_absolute_error(y_true, y_pred, multioutput=[0.3, 0.7])
...
0.85...
Mean squared error
The mean_squared_error function computes mean square error, a risk metric corresponding to the expected
value of the squared (quadratic) error or loss.
If 𝑦ˆ𝑖 is the predicted value of the 𝑖-th sample, and 𝑦𝑖 is the corresponding true value, then the mean squared error
(MSE) estimated over 𝑛samples is defined as
𝑛samples −1
1 ∑︁
MSE(𝑦, 𝑦ˆ) = (𝑦𝑖 − 𝑦ˆ𝑖 )2 .
𝑛samples 𝑖=0
Here is a small example of usage of the mean_squared_error function:

>>> y_true = [3, -0.5, 2, 7]
>>> y_pred = [2.5, 0.0, 2, 8]
>>> mean_squared_error(y_true, y_pred)
0.375
>>> y_true = [[0.5, 1], [-1, 1], [7, -6]]
>>> y_pred = [[0, 2], [-1, 2], [8, -5]]
0.7083...
Examples:

• See Gradient Boosting regression for an example of mean squared error usage to evaluate gradient boosting
regression.
Mean squared logarithmic error
The mean_squared_log_error function computes a risk metric corresponding to the expected value of the
squared logarithmic (quadratic) error or loss.
If 𝑦ˆ𝑖 is the predicted value of the 𝑖-th sample, and 𝑦𝑖 is the corresponding true value, then the mean squared logarithmic
error (MSLE) estimated over 𝑛samples is defined as
𝑛samples −1
1 ∑︁
MSLE(𝑦, 𝑦ˆ) = (log𝑒 (1 + 𝑦𝑖 ) − log𝑒 (1 + 𝑦ˆ𝑖 ))2 .
𝑛samples 𝑖=0
Where log𝑒 (𝑥) means the natural logarithm of 𝑥. This metric is best to use when targets having exponential growth,
such as population counts, average sales of a commodity over a span of years etc. Note that this metric penalizes an
under-predicted estimate greater than an over-predicted estimate.
Here is a small example of usage of the mean_squared_log_error function:
>>> from sklearn.metrics import mean_squared_log_error
>>> y_true = [3, 5, 2.5, 7]
>>> y_pred = [2.5, 5, 4, 8]
>>> mean_squared_log_error(y_true, y_pred)
0.039...
>>> y_true = [[0.5, 1], [1, 2], [7, 6]]
>>> y_pred = [[0.5, 2], [1, 2.5], [8, 8]]
0.044...
Median absolute error
The median_absolute_error is particularly interesting because it is robust to outliers. The loss is calculated by
taking the median of all absolute differences between the target and the prediction.
If 𝑦ˆ𝑖 is the predicted value of the 𝑖-th sample and 𝑦𝑖 is the corresponding true value, then the median absolute error
(MedAE) estimated over 𝑛samples is defined as
MedAE(𝑦, 𝑦ˆ) = median(| 𝑦1 − 𝑦ˆ1 |, . . . , | 𝑦𝑛 − 𝑦ˆ𝑛 |).
The median_absolute_error does not support multioutput.
Here is a small example of usage of the median_absolute_error function:
>>> from sklearn.metrics import median_absolute_error
>>> y_true = [3, -0.5, 2, 7]
>>> y_pred = [2.5, 0.0, 2, 8]
>>> median_absolute_error(y_true, y_pred)
0.5
R2 score, the coefficient of determination
The r2_score function computes R2 , the coefficient of determination. It provides a measure of how well future
samples are likely to be predicted by the model. Best possible score is 1.0 and it can be negative (because the model

can be arbitrarily worse). A constant model that always predicts the expected value of y, disregarding the input features,
would get a R^2 score of 0.0.
If 𝑦ˆ𝑖 is the predicted value of the 𝑖-th sample and 𝑦𝑖 is the corresponding true value, then the score R2 estimated over
𝑛samples is defined as
∑︀𝑛samples −1
(𝑦𝑖 − 𝑦ˆ𝑖 )2
𝑅2 (𝑦, 𝑦ˆ) = 1 − ∑︀𝑖=0
𝑛samples −1
𝑖=0 (𝑦𝑖 − 𝑦¯)2
1
∑︀𝑛samples −1
where 𝑦¯ = 𝑛samples 𝑖=0 𝑦𝑖 .
Here is a small example of usage of the r2_score function:
>>> from sklearn.metrics import r2_score

>>> y_true = [3, -0.5, 2, 7]
>>> y_pred = [2.5, 0.0, 2, 8]
>>> r2_score(y_true, y_pred)
0.948...
>>> y_true = [[0.5, 1], [-1, 1], [7, -6]]
>>> y_pred = [[0, 2], [-1, 2], [8, -5]]
>>> r2_score(y_true, y_pred, multioutput='variance_weighted')
...
0.938...
>>> y_true = [[0.5, 1], [-1, 1], [7, -6]]
>>> y_pred = [[0, 2], [-1, 2], [8, -5]]
>>> r2_score(y_true, y_pred, multioutput='uniform_average')
...
0.936...
>>> r2_score(y_true, y_pred, multioutput='raw_values')
...
array([0.965..., 0.908...])
>>> r2_score(y_true, y_pred, multioutput=[0.3, 0.7])
...
0.925...
Example:
• See Lasso and Elastic Net for Sparse Signals for an example of R2 score usage to evaluate Lasso and Elastic
Net on sparse signals.
Clustering metrics
The sklearn.metrics module implements several loss, score, and utility functions. For more information see the
Clustering performance evaluation section for instance clustering, and Biclustering evaluation for biclustering.
Dummy estimators
When doing supervised learning, a simple sanity check consists of comparing one’s estimator against simple rules of
thumb. DummyClassifier implements several such simple strategies for classification:
• stratified generates random predictions by respecting the training set class distribution.
• most_frequent always predicts the most frequent label in the training set.

• prior always predicts the class that maximizes the class prior (like most_frequent) and
predict_proba returns the class prior.
• uniform generates predictions uniformly at random.
• constant always predicts a constant label that is provided by the user. A major motivation of this
method is F1-scoring, when the positive class is in the minority.
Note that with all these strategies, the predict method completely ignores the input data!
To illustrate DummyClassifier, first let’s create an imbalanced dataset:

>>> y[y != 1] = -1
>>> X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)
Next, let’s compare the accuracy of SVC and most_frequent:
>>> from sklearn.dummy import DummyClassifier

>>> clf = SVC(kernel='linear', C=1).fit(X_train, y_train)
0.63...
>>> clf = DummyClassifier(strategy='most_frequent', random_state=0)
>>> clf.fit(X_train, y_train)
DummyClassifier(constant=None, random_state=0, strategy='most_frequent')
0.57...
We see that SVC doesn’t do much better than a dummy classifier. Now, let’s change the kernel:
>>> clf = SVC(gamma='scale', kernel='rbf', C=1).fit(X_train, y_train)

0.94...
We see that the accuracy was boosted to almost 100%. A cross validation strategy is recommended for a better
estimate of the accuracy, if it is not too CPU costly. For more information see the Cross-validation: evaluating
estimator performance section. Moreover if you want to optimize over the parameter space, it is highly recommended
to use an appropriate methodology; see the Tuning the hyper-parameters of an estimator section for details.
More generally, when the accuracy of a classifier is too close to random, it probably means that something went wrong:
features are not helpful, a hyperparameter is not correctly tuned, the classifier is suffering from class imbalance, etc. . .
DummyRegressor also implements four simple rules of thumb for regression:
• mean always predicts the mean of the training targets.
• median always predicts the median of the training targets.
• quantile always predicts a user provided quantile of the training targets.
• constant always predicts a constant value that is provided by the user.
In all these strategies, the predict method completely ignores the input data.

3.3.4 Model persistence
After training a scikit-learn model, it is desirable to have a way to persist the model for future use without having to
retrain. The following section gives you an example of how to persist a model with pickle. We’ll also review a few
security and maintainability issues when working with pickle serialization.
An alternative to pickling is to export the model to another format using one of the model export tools listed under
Related Projects. Unlike pickling, once exported you cannot recover the full Scikit-learn estimator object, but you can
deploy the model for prediction, usually by using tools supporting open model interchange formats such as ‘ONNX‘_
or ‘PMML‘_.
Persistence example
It is possible to save a model in scikit-learn by using Python’s built-in persistence model, namely pickle:

>>> clf.fit(X, y)
>>> import pickle

>>> s = pickle.dumps(clf)
>>> clf2 = pickle.loads(s)
>>> clf2.predict(X[0:1])
array([0])
>>> y[0]
0
In the specific case of scikit-learn, it may be better to use joblib’s replacement of pickle (dump & load), which is
more efficient on objects that carry large numpy arrays internally as is often the case for fitted scikit-learn estimators,
but can only pickle to the disk and not to a string:
>>> from joblib import dump, load

>>> dump(clf, 'filename.joblib')
Later you can load back the pickled model (possibly in another Python process) with:
>>> clf = load('filename.joblib')
Note: dump and load functions also accept file-like object instead of filenames. More information on data persis-
tence with Joblib is available here.
Security & maintainability limitations
pickle (and joblib by extension), has some issues regarding maintainability and security. Because of this,
• Never unpickle untrusted data as it could lead to malicious code being executed upon loading.

• While models saved using one version of scikit-learn might load in other versions, this is entirely unsupported
and inadvisable. It should also be kept in mind that operations performed on such data could give different and
unexpected results.
In order to rebuild a similar model with future versions of scikit-learn, additional metadata should be saved along the
pickled model:
• The training data, e.g. a reference to an immutable snapshot
• The python source code used to generate the model
• The versions of scikit-learn and its dependencies
• The cross validation score obtained on the training data
This should make it possible to check that the cross-validation score is in the same range as before.
Since a model internal representation may be different on two different architectures, dumping a model on one archi-
tecture and loading it on another architecture is not supported.
If you want to know more about these issues and explore other possible serialization methods, please refer to this talk
by Alex Gaynor.
3.3.5 Validation curves: plotting scores to evaluate models
Every estimator has its advantages and drawbacks. Its generalization error can be decomposed in terms of bias,
variance and noise. The bias of an estimator is its average error for different training sets. The variance of an
estimator indicates how sensitive it is to varying training sets. Noise is a property of the data.
In the following plot, we see a function 𝑓 (𝑥) = cos( 32 𝜋𝑥) and some noisy samples from that function. We use three
different estimators to fit the function: linear regression with polynomial features of degree 1, 4 and 15. We see that
the first estimator can at best provide only a poor fit to the samples and the true function because it is too simple
(high bias), the second estimator approximates it almost perfectly and the last estimator approximates the training data
perfectly but does not fit the true function very well, i.e. it is very sensitive to varying training data (high variance).
Bias and variance are inherent properties of estimators and we usually have to select learning algorithms and hyper-
parameters so that both bias and variance are as low as possible (see Bias-variance dilemma). Another way to reduce
the variance of a model is to use more training data. However, you should only collect more training data if the true
function is too complex to be approximated by an estimator with a lower variance.
In the simple one-dimensional problem that we have seen in the example it is easy to see whether the estimator suffers
from bias or variance. However, in high-dimensional spaces, models can become very difficult to visualize. For this
reason, it is often helpful to use the tools described below.

Examples:
• Underfitting vs. Overfitting

• Plotting Validation Curves
• Plotting Learning Curves
Validation curve
To validate a model we need a scoring function (see Model evaluation: quantifying the quality of predictions), for
example accuracy for classifiers. The proper way of choosing multiple hyperparameters of an estimator are of course
grid search or similar methods (see Tuning the hyper-parameters of an estimator) that select the hyperparameter with
the maximum score on a validation set or multiple validation sets. Note that if we optimized the hyperparameters
based on a validation score the validation score is biased and not a good estimate of the generalization any longer. To
get a proper estimate of the generalization we have to compute the score on another test set.
However, it is sometimes helpful to plot the influence of a single hyperparameter on the training score and the valida-
tion score to find out whether the estimator is overfitting or underfitting for some hyperparameter values.
The function validation_curve can help in this case:

>>> from sklearn.model_selection import validation_curve
>>> from sklearn.linear_model import Ridge
>>> indices = np.arange(y.shape[0])
>>> np.random.shuffle(indices)
>>> X, y = X[indices], y[indices]
>>> train_scores, valid_scores = validation_curve(Ridge(), X, y, "alpha",

... np.logspace(-7, 3, 3),
... cv=5)
>>> train_scores
array([[0.93..., 0.94..., 0.92..., 0.91..., 0.92...],
[0.93..., 0.94..., 0.92..., 0.91..., 0.92...],
[0.51..., 0.52..., 0.49..., 0.47..., 0.49...]])
>>> valid_scores
array([[0.90..., 0.84..., 0.94..., 0.96..., 0.93...],
[0.90..., 0.84..., 0.94..., 0.96..., 0.93...],
[0.46..., 0.25..., 0.50..., 0.49..., 0.52...]])
If the training score and the validation score are both low, the estimator will be underfitting. If the training score is
high and the validation score is low, the estimator is overfitting and otherwise it is working very well. A low training
score and a high validation score is usually not possible. All three cases can be found in the plot below where we vary
the parameter 𝛾 of an SVM on the digits dataset.
Learning curve
A learning curve shows the validation and training score of an estimator for varying numbers of training samples. It
is a tool to find out how much we benefit from adding more training data and whether the estimator suffers more from

a variance error or a bias error. If both the validation score and the training score converge to a value that is too low
with increasing size of the training set, we will not benefit much from more training data. In the following plot you
can see an example: naive Bayes roughly converges to a low score.
We will probably have to use an estimator or a parametrization of the current estimator that can learn more complex
concepts (i.e. has a lower bias). If the training score is much greater than the validation score for the maximum number
of training samples, adding more training samples will most likely increase generalization. In the following plot you
can see that the SVM could benefit from more training examples.
We can use the function learning_curve to generate the values that are required to plot such a learning curve
(number of samples that have been used, the average scores on the training sets and the average scores on the validation
sets):
>>> from sklearn.model_selection import learning_curve

>>> train_sizes, train_scores, valid_scores = learning_curve(

... SVC(kernel='linear'), X, y, train_sizes=[50, 80, 110], cv=5)
>>> train_sizes
array([ 50, 80, 110])
>>> train_scores
array([[0.98..., 0.98 , 0.98..., 0.98..., 0.98...],
[0.98..., 1. , 0.98..., 0.98..., 0.98...],

[0.98..., 1. , 0.98..., 0.98..., 0.99...]])

>>> valid_scores
array([[1. , 0.93..., 1. , 1. , 0.96...],
[1. , 0.96..., 1. , 1. , 0.96...],
[1. , 0.96..., 1. , 1. , 0.96...]])
3.4 Dataset transformations
scikit-learn provides a library of transformers, which may clean (see Preprocessing data), reduce (see Unsupervised
dimensionality reduction), expand (see Kernel Approximation) or generate (see Feature extraction) feature representa-
tions.
Like other estimators, these are represented by classes with a fit method, which learns model parameters (e.g.
mean and standard deviation for normalization) from a training set, and a transform method which applies this
transformation model to unseen data. fit_transform may be more convenient and efficient for modelling and
transforming the training data simultaneously.
Combining such transformers, either in parallel or series is covered in Pipelines and composite estimators. Pair-
wise metrics, Affinities and Kernels covers transforming feature spaces into affinity matrices, while Transforming the
prediction target (y) considers transformations of the target space (e.g. categorical labels) for use in scikit-learn.
3.4.1 Pipelines and composite estimators
Transformers are usually combined with classifiers, regressors or other estimators to build a composite estimator. The
most common tool is a Pipeline. Pipeline is often used in combination with FeatureUnion which concatenates the
output of transformers into a composite feature space. TransformedTargetRegressor deals with transforming the target
(i.e. log-transform y). In contrast, Pipelines only transform the observed data (X).
Pipeline: chaining estimators
Pipeline can be used to chain multiple estimators into one. This is useful as there is often a fixed sequence of steps
in processing the data, for example feature selection, normalization and classification. Pipeline serves multiple
purposes here:
Convenience and encapsulation You only have to call fit and predict once on your data to fit a whole sequence
of estimators.
3.4. Dataset transformations 535

Joint parameter selection You can grid search over parameters of all estimators in the pipeline at once.
Safety Pipelines help avoid leaking statistics from your test data into the trained model in cross-validation, by ensuring
that the same samples are used to train the transformers and predictors.
All estimators in a pipeline, except the last one, must be transformers (i.e. must have a transform method). The
last estimator may be any type (transformer, classifier, etc.).
Usage
The Pipeline is built using a list of (key, value) pairs, where the key is a string containing the name you
want to give this step and value is an estimator object:

>>> from sklearn.decomposition import PCA
>>> estimators = [('reduce_dim', PCA()), ('clf', SVC())]
>>> pipe = Pipeline(estimators)
>>> pipe
Pipeline(memory=None,
steps=[('reduce_dim', PCA(copy=True,...)),
('clf', SVC(C=1.0,...))])
The utility function make_pipeline is a shorthand for constructing pipelines; it takes a variable number of estima-
tors and returns a pipeline, filling in the names automatically:
>>> from sklearn.pipeline import make_pipeline

>>> from sklearn.preprocessing import Binarizer
>>> make_pipeline(Binarizer(), MultinomialNB())
steps=[('binarizer', Binarizer(copy=True, threshold=0.0)),
('multinomialnb', MultinomialNB(alpha=1.0,
class_prior=None,
fit_prior=True))])
The estimators of a pipeline are stored as a list in the steps attribute:
>>> pipe.steps[0]
('reduce_dim', PCA(copy=True, iterated_power='auto', n_components=None, random_
˓→state=None,
svd_solver='auto', tol=0.0, whiten=False))
and as a dict in named_steps:
>>> pipe.named_steps['reduce_dim']
PCA(copy=True, iterated_power='auto', n_components=None, random_state=None,
Parameters of the estimators in the pipeline can be accessed using the <estimator>__<parameter> syntax:
>>> pipe.set_params(clf__C=10)
steps=[('reduce_dim', PCA(copy=True, iterated_power='auto',...)),
('clf', SVC(C=10, cache_size=200, class_weight=None,...))])
Attributes of named_steps map to keys, enabling tab completion in interactive environments:

>>> pipe.named_steps.reduce_dim is pipe.named_steps['reduce_dim']

True
This is particularly important for doing grid searches:

>>> param_grid = dict(reduce_dim__n_components=[2, 5, 10],
... clf__C=[0.1, 10, 100])
>>> grid_search = GridSearchCV(pipe, param_grid=param_grid)
Individual steps may also be replaced as parameters, and non-final steps may be ignored by setting them to None:

>>> param_grid = dict(reduce_dim=[None, PCA(5), PCA(10)],
... clf=[SVC(), LogisticRegression()],
... clf__C=[0.1, 10, 100])
>>> grid_search = GridSearchCV(pipe, param_grid=param_grid)
Examples:
• Pipeline Anova SVM

• Sample pipeline for text feature extraction and evaluation
• Pipelining: chaining a PCA and a logistic regression
• Explicit feature map approximation for RBF kernels
• SVM-Anova: SVM with univariate feature selection
• Selecting dimensionality reduction with Pipeline and GridSearchCV
See also:
• Tuning the hyper-parameters of an estimator
Notes
Calling fit on the pipeline is the same as calling fit on each estimator in turn, transform the input and pass it
on to the next step. The pipeline has all the methods that the last estimator in the pipeline has, i.e. if the last estimator
is a classifier, the Pipeline can be used as a classifier. If the last estimator is a transformer, again, so is the pipeline.
Caching transformers: avoid repeated computation
Fitting transformers may be computationally expensive. With its memory parameter set, Pipeline will cache each
transformer after calling fit. This feature is used to avoid computing the fit transformers within a pipeline if the
parameters and input data are identical. A typical example is the case of a grid search in which the transformers can
be fitted only once and reused for each configuration.
The parameter memory is needed in order to cache the transformers. memory can be either a string containing the
directory where to cache the transformers or a joblib.Memory object:

>>> from tempfile import mkdtemp

>>> from shutil import rmtree
>>> estimators = [('reduce_dim', PCA()), ('clf', SVC())]
>>> cachedir = mkdtemp()
>>> pipe = Pipeline(estimators, memory=cachedir)
>>> pipe
Pipeline(...,
steps=[('reduce_dim', PCA(copy=True,...)),
('clf', SVC(C=1.0,...))])
>>> # Clear the cache directory when you don't need it anymore
>>> rmtree(cachedir)
Warning: Side effect of caching transformers

Using a Pipeline without cache enabled, it is possible to inspect the original instance such as:
>>> from sklearn.datasets import load_digits
>>> digits = load_digits()
>>> pca1 = PCA()
>>> svm1 = SVC(gamma='scale')
>>> pipe = Pipeline([('reduce_dim', pca1), ('clf', svm1)])
>>> pipe.fit(digits.data, digits.target)
...
steps=[('reduce_dim', PCA(...)), ('clf', SVC(...))])
>>> # The pca instance can be inspected directly
>>> print(pca1.components_)
[[-1.77484909e-19 ... 4.07058917e-18]]
Enabling caching triggers a clone of the transformers before fitting. Therefore, the transformer instance given to
the pipeline cannot be inspected directly. In following example, accessing the PCA instance pca2 will raise an
AttributeError since pca2 will be an unfitted transformer. Instead, use the attribute named_steps to
inspect estimators within the pipeline:
>>> cachedir = mkdtemp()
>>> pca2 = PCA()
>>> svm2 = SVC(gamma='scale')
>>> cached_pipe = Pipeline([('reduce_dim', pca2), ('clf', svm2)],
... memory=cachedir)
>>> cached_pipe.fit(digits.data, digits.target)
...
Pipeline(memory=...,
steps=[('reduce_dim', PCA(...)), ('clf', SVC(...))])
>>> print(cached_pipe.named_steps['reduce_dim'].components_)
...
[[-1.77484909e-19 ... 4.07058917e-18]]
>>> # Remove the cache directory
>>> rmtree(cachedir)
Examples:

Transforming target in regression
TransformedTargetRegressor transforms the targets y before fitting a regression model. The predictions are
mapped back to the original space via an inverse transform. It takes as an argument the regressor that will be used for
prediction, and the transformer that will be applied to the target variable:

>>> from sklearn.datasets import load_boston
>>> from sklearn.compose import TransformedTargetRegressor
>>> from sklearn.preprocessing import QuantileTransformer
>>> boston = load_boston()
>>> X = boston.data
>>> y = boston.target
>>> transformer = QuantileTransformer(output_distribution='normal')
>>> regressor = LinearRegression()
>>> regr = TransformedTargetRegressor(regressor=regressor,
... transformer=transformer)
>>> regr.fit(X_train, y_train)
TransformedTargetRegressor(...)
>>> print('R2 score: {0:.2f}'.format(regr.score(X_test, y_test)))
R2 score: 0.67
>>> raw_target_regr = LinearRegression().fit(X_train, y_train)
>>> print('R2 score: {0:.2f}'.format(raw_target_regr.score(X_test, y_test)))
R2 score: 0.64
For simple transformations, instead of a Transformer object, a pair of functions can be passed, defining the transfor-
mation and its inverse mapping:
>>> from __future__ import division

>>> def func(x):
... return np.log(x)
>>> def inverse_func(x):
... return np.exp(x)
Subsequently, the object is created as:

... func=func,
... inverse_func=inverse_func)
R2 score: 0.65
By default, the provided functions are checked at each fit to be the inverse of each other. However, it is possible to
bypass this checking by setting check_inverse to False:
>>> def inverse_func(x):

... return x
... func=func,
... inverse_func=inverse_func,
... check_inverse=False)


R2 score: -4.50
Note: The transformation can be triggered by setting either transformer or the pair of functions func and
inverse_func. However, setting both options will raise an error.
Examples:
FeatureUnion: composite feature spaces
FeatureUnion combines several transformer objects into a new transformer that combines their output. A
FeatureUnion takes a list of transformer objects. During fitting, each of these is fit to the data independently.
The transformers are applied in parallel, and the feature matrices they output are concatenated side-by-side into a
larger matrix.
When you want to apply different transformations to each field of the data, see the related class sklearn.compose.
ColumnTransformer (see user guide).
FeatureUnion serves the same purposes as Pipeline - convenience and joint parameter estimation and valida-
tion.
FeatureUnion and Pipeline can be combined to create complex models.
(A FeatureUnion has no way of checking whether two transformers might produce identical features. It only
produces a union when the feature sets are disjoint, and making sure they are the caller’s responsibility.)
Usage
A FeatureUnion is built using a list of (key, value) pairs, where the key is the name you want to give to a
given transformation (an arbitrary string; it only serves as an identifier) and value is an estimator object:
>>> from sklearn.pipeline import FeatureUnion

>>> from sklearn.decomposition import KernelPCA
>>> estimators = [('linear_pca', PCA()), ('kernel_pca', KernelPCA())]
>>> combined = FeatureUnion(estimators)
>>> combined
FeatureUnion(n_jobs=None,
transformer_list=[('linear_pca', PCA(copy=True,...)),
('kernel_pca', KernelPCA(alpha=1.0,...))],
transformer_weights=None)
Like pipelines, feature unions have a shorthand constructor called make_union that does not require explicit naming
of the components.
Like Pipeline, individual steps may be replaced using set_params, and ignored by setting to 'drop':
>>> combined.set_params(kernel_pca='drop')
...

transformer_list=[('linear_pca', PCA(copy=True,...)),
('kernel_pca', 'drop')],
Examples:
• Concatenating multiple feature extraction methods
ColumnTransformer for heterogeneous data
Warning: The compose.ColumnTransformer class is experimental and the API is subject to change.
Many datasets contain features of different types, say text, floats, and dates, where each type of feature requires
separate preprocessing or feature extraction steps. Often it is easiest to preprocess data before applying scikit-learn
methods, for example using pandas. Processing your data before passing it to scikit-learn might be problematic for
one of the following reasons:
1. Incorporating statistics from test data into the preprocessors makes cross-validation scores unreliable (known as
data leakage), for example in the case of scalers or imputing missing values.
2. You may want to include the parameters of the preprocessors in a parameter search.
The ColumnTransformer helps performing different transformations for different columns of the data, within a
Pipeline that is safe from data leakage and that can be parametrized. ColumnTransformer works on arrays,
sparse matrices, and pandas DataFrames.
To each column, a different transformation can be applied, such as preprocessing or a specific feature extraction
method:
>>> import pandas as pd
>>> X = pd.DataFrame(
... {'city': ['London', 'London', 'Paris', 'Sallisaw'],
... 'title': ["His Last Bow", "How Watson Learned the Trick",
... "A Moveable Feast", "The Grapes of Wrath"],
... 'expert_rating': [5, 3, 4, 5],
... 'user_rating': [4, 5, 4, 3]})
For this data, we might want to encode the 'city' column as a categorical variable using preprocessing.
OneHotEncoder but apply a feature_extraction.text.CountVectorizer to the 'title' column.
As we might use multiple feature extraction methods on the same column, we give each transformer a unique
name, say 'city_category' and 'title_bow'. By default, the remaining rating columns are ignored
(remainder='drop'):
>>> from sklearn.compose import ColumnTransformer
>>> from sklearn.preprocessing import OneHotEncoder
>>> column_trans = ColumnTransformer(
... [('city_category', OneHotEncoder(dtype='int'),['city']),
... ('title_bow', CountVectorizer(), 'title')],
... remainder='drop')
>>> column_trans.fit(X)
ColumnTransformer(n_jobs=None, remainder='drop', sparse_threshold=0.3,

transformer_weights=None,
transformers=...)
>>> column_trans.get_feature_names()
...
['city_category__x0_London', 'city_category__x0_Paris', 'city_category__x0_Sallisaw',
'title_bow__bow', 'title_bow__feast', 'title_bow__grapes', 'title_bow__his',
'title_bow__how', 'title_bow__last', 'title_bow__learned', 'title_bow__moveable',
'title_bow__of', 'title_bow__the', 'title_bow__trick', 'title_bow__watson',
'title_bow__wrath']
>>> column_trans.transform(X).toarray()
...
array([[1, 0, 0, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 1, 1, 1, 0],
[0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1]]...)
In the above example, the CountVectorizer expects a 1D array as input and therefore the columns were specified
as a string ('title'). However, preprocessing.OneHotEncoder as most of other transformers expects 2D
data, therefore in that case you need to specify the column as a list of strings (['city']).
Apart from a scalar or a single item list, the column selection can be specified as a list of multiple items, an integer
array, a slice, or a boolean mask. Strings can reference columns if the input is a DataFrame, integers are always
interpreted as the positional columns.
We can keep the remaining rating columns by setting remainder='passthrough'. The values are appended to
the end of the transformation:

... [('city_category', OneHotEncoder(dtype='int'),['city']),
... remainder='passthrough')
>>> column_trans.fit_transform(X)
...
array([[1, 0, 0, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 5, 4],
[1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 1, 1, 1, 0, 3, 5],
[0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 4, 4],
[0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 5, 3]]...)
The remainder parameter can be set to an estimator to transform the remaining rating columns. The transformed
values are appended to the end of the transformation:
>>> from sklearn.preprocessing import MinMaxScaler

... [('city_category', OneHotEncoder(), ['city']),
... remainder=MinMaxScaler())
>>> column_trans.fit_transform(X)[:, -2:]

...
array([[1. , 0.5],
[0. , 1. ],
[0.5, 0.5],
[1. , 0. ]])

The make_column_transformer function is available to more easily create a ColumnTransformer object.

Specifically, the names will be given automatically. The equivalent for the above example would be:
>>> from sklearn.compose import make_column_transformer
>>> column_trans = make_column_transformer(
... (OneHotEncoder(), ['city']),
... (CountVectorizer(), 'title'),
... remainder=MinMaxScaler())
>>> column_trans
ColumnTransformer(n_jobs=None, remainder=MinMaxScaler(copy=True, ...),
sparse_threshold=0.3,
transformers=[('onehotencoder', ...)
Examples:
• Column Transformer with Heterogeneous Data Sources

• Column Transformer with Mixed Types
3.4.2 Feature extraction
The sklearn.feature_extraction module can be used to extract features in a format supported by machine
learning algorithms from datasets consisting of formats such as text and image.
Note: Feature extraction is very different from Feature selection: the former consists in transforming arbitrary data,
such as text or images, into numerical features usable for machine learning. The latter is a machine learning technique
applied on these features.
Loading features from dicts
The class DictVectorizer can be used to convert feature arrays represented as lists of standard Python dict
objects to the NumPy/SciPy representation used by scikit-learn estimators.
While not particularly fast to process, Python’s dict has the advantages of being convenient to use, being sparse
(absent features need not be stored) and storing feature names in addition to values.
DictVectorizer implements what is called one-of-K or “one-hot” coding for categorical (aka nominal, discrete)
features. Categorical features are “attribute-value” pairs where the value is restricted to a list of discrete of possibilities
without ordering (e.g. topic identifiers, types of objects, tags, names. . . ).
In the following, “city” is a categorical attribute while “temperature” is a traditional numerical feature:
>>> measurements = [
... {'city': 'Dubai', 'temperature': 33.},
... {'city': 'London', 'temperature': 12.},
... {'city': 'San Francisco', 'temperature': 18.},
... ]
>>> from sklearn.feature_extraction import DictVectorizer

>>> vec = DictVectorizer()
>>> vec.fit_transform(measurements).toarray()

array([[ 1., 0., 0., 33.],

[ 0., 1., 0., 12.],
[ 0., 0., 1., 18.]])
>>> vec.get_feature_names()
['city=Dubai', 'city=London', 'city=San Francisco', 'temperature']
DictVectorizer is also a useful representation transformation for training sequence classifiers in Natural Lan-
guage Processing models that typically work by extracting feature windows around a particular word of interest.
For example, suppose that we have a first algorithm that extracts Part of Speech (PoS) tags that we want to use as
complementary tags for training a sequence classifier (e.g. a chunker). The following dict could be such a window of
features extracted around the word ‘sat’ in the sentence ‘The cat sat on the mat.’:
>>> pos_window = [
... {
... 'word-2': 'the',
... 'pos-2': 'DT',
... 'word-1': 'cat',
... 'pos-1': 'NN',
... 'word+1': 'on',
... 'pos+1': 'PP',
... },
... # in a real application one would extract many such dictionaries
... ]
This description can be vectorized into a sparse two-dimensional matrix suitable for feeding into a classifier (maybe
after being piped into a text.TfidfTransformer for normalization):
>>> vec = DictVectorizer()

>>> pos_vectorized = vec.fit_transform(pos_window)
>>> pos_vectorized
<1x6 sparse matrix of type '<... 'numpy.float64'>'
with 6 stored elements in Compressed Sparse ... format>
>>> pos_vectorized.toarray()
array([[1., 1., 1., 1., 1., 1.]])
>>> vec.get_feature_names()
['pos+1=PP', 'pos-1=NN', 'pos-2=DT', 'word+1=on', 'word-1=cat', 'word-2=the']
As you can imagine, if one extracts such a context around each individual word of a corpus of documents the resulting
matrix will be very wide (many one-hot-features) with most of them being valued to zero most of the time. So as to
make the resulting data structure able to fit in memory the DictVectorizer class uses a scipy.sparse matrix
by default instead of a numpy.ndarray.
Feature hashing
The class FeatureHasher is a high-speed, low-memory vectorizer that uses a technique known as feature hashing,
or the “hashing trick”. Instead of building a hash table of the features encountered in training, as the vectorizers
do, instances of FeatureHasher apply a hash function to the features to determine their column index in sample
matrices directly. The result is increased speed and reduced memory usage, at the expense of inspectability; the hasher
does not remember what the input features looked like and has no inverse_transform method.
Since the hash function might cause collisions between (unrelated) features, a signed hash function is used and the
sign of the hash value determines the sign of the value stored in the output matrix for a feature. This way, collisions
are likely to cancel out rather than accumulate error, and the expected mean of any output feature’s value is zero.
This mechanism is enabled by default with alternate_sign=True and is particularly useful for small hash table

sizes (n_features < 10000). For large hash table sizes, it can be disabled, to allow the output to be passed
to estimators like sklearn.naive_bayes.MultinomialNB or sklearn.feature_selection.chi2
feature selectors that expect non-negative inputs.
FeatureHasher accepts either mappings (like Python’s dict and its variants in the collections module),
(feature, value) pairs, or strings, depending on the constructor parameter input_type. Mapping are treated
as lists of (feature, value) pairs, while single strings have an implicit value of 1, so ['feat1', 'feat2',
'feat3'] is interpreted as [('feat1', 1), ('feat2', 1), ('feat3', 1)]. If a single feature occurs
multiple times in a sample, the associated values will be summed (so ('feat', 2) and ('feat', 3.5) become
('feat', 5.5)). The output from FeatureHasher is always a scipy.sparse matrix in the CSR format.
Feature hashing can be employed in document classification, but unlike text.CountVectorizer,
FeatureHasher does not do word splitting or any other preprocessing except Unicode-to-UTF-8 encoding; see
Vectorizing a large text corpus with the hashing trick, below, for a combined tokenizer/hasher.
As an example, consider a word-level natural language processing task that needs features extracted from (token,
part_of_speech) pairs. One could use a Python generator function to extract features:
def token_features(token, part_of_speech):

if token.isdigit():
yield "numeric"
else:
yield "token={}".format(token.lower())
yield "token,pos={},{}".format(token, part_of_speech)
if token[0].isupper():
yield "uppercase_initial"
if token.isupper():
yield "all_uppercase"
yield "pos={}".format(part_of_speech)
Then, the raw_X to be fed to FeatureHasher.transform can be constructed using:
raw_X = (token_features(tok, pos_tagger(tok)) for tok in corpus)
and fed to a hasher with:
hasher = FeatureHasher(input_type='string')
X = hasher.transform(raw_X)
to get a scipy.sparse matrix X.

Note the use of a generator comprehension, which introduces laziness into the feature extraction: tokens are only
processed on demand from the hasher.
FeatureHasher uses the signed 32-bit variant of MurmurHash3. As a result (and because of limitations in scipy.
sparse), the maximum number of features supported is currently 231 − 1.
The original formulation of the hashing trick by Weinberger et al. used two separate hash functions ℎ and 𝜉 to deter-
mine the column index and sign of a feature, respectively. The present implementation works under the assumption
that the sign bit of MurmurHash3 is independent of its other bits.
Since a simple modulo is used to transform the hash function to a column index, it is advisable to use a power of two
as the n_features parameter; otherwise the features will not be mapped evenly to the columns.

References:
• Kilian Weinberger, Anirban Dasgupta, John Langford, Alex Smola and Josh Attenberg (2009). Feature hash-
ing for large scale multitask learning. Proc. ICML.
• MurmurHash3.
Text feature extraction
The Bag of Words representation
Text Analysis is a major application field for machine learning algorithms. However the raw data, a sequence of
symbols cannot be fed directly to the algorithms themselves as most of them expect numerical feature vectors with a
fixed size rather than the raw text documents with variable length.
In order to address this, scikit-learn provides utilities for the most common ways to extract numerical features from
text content, namely:
• tokenizing strings and giving an integer id for each possible token, for instance by using white-spaces and
punctuation as token separators.
• counting the occurrences of tokens in each document.
• normalizing and weighting with diminishing importance tokens that occur in the majority of samples / docu-
ments.
In this scheme, features and samples are defined as follows:
• each individual token occurrence frequency (normalized or not) is treated as a feature.
• the vector of all the token frequencies for a given document is considered a multivariate sample.
A corpus of documents can thus be represented by a matrix with one row per document and one column per token
(e.g. word) occurring in the corpus.
We call vectorization the general process of turning a collection of text documents into numerical feature vectors. This
specific strategy (tokenization, counting and normalization) is called the Bag of Words or “Bag of n-grams” represen-
tation. Documents are described by word occurrences while completely ignoring the relative position information of
the words in the document.
Sparsity
As most documents will typically use a very small subset of the words used in the corpus, the resulting matrix will
have many feature values that are zeros (typically more than 99% of them).
For instance a collection of 10,000 short text documents (such as emails) will use a vocabulary with a size in the order
of 100,000 unique words in total while each document will use 100 to 1000 unique words individually.
In order to be able to store such a matrix in memory but also to speed up algebraic operations matrix / vector, imple-
mentations will typically use a sparse representation such as the implementations available in the scipy.sparse
package.
Common Vectorizer usage
CountVectorizer implements both tokenization and occurrence counting in a single class:

This model has many parameters, however the default values are quite reasonable (please see the reference documen-
tation for the details):
>>> vectorizer = CountVectorizer()

>>> vectorizer
CountVectorizer(analyzer=...'word', binary=False, decode_error=...'strict',
dtype=<... 'numpy.int64'>, encoding=...'utf-8', input=...'content',
lowercase=True, max_df=1.0, max_features=None, min_df=1,
ngram_range=(1, 1), preprocessor=None, stop_words=None,
strip_accents=None, token_pattern=...'(?u)\\b\\w\\w+\\b',
tokenizer=None, vocabulary=None)
Let’s use it to tokenize and count the word occurrences of a minimalistic corpus of text documents:
>>> corpus = [
... 'This is the first document.',
... 'This is the second second document.',
... 'And the third one.',
... 'Is this the first document?',
... ]
>>> X = vectorizer.fit_transform(corpus)
>>> X
<4x9 sparse matrix of type '<... 'numpy.int64'>'
The default configuration tokenizes the string by extracting words of at least 2 letters. The specific function that does
this step can be requested explicitly:
>>> analyze = vectorizer.build_analyzer()

>>> analyze("This is a text document to analyze.") == (
... ['this', 'is', 'text', 'document', 'to', 'analyze'])
True
Each term found by the analyzer during the fit is assigned a unique integer index corresponding to a column in the
resulting matrix. This interpretation of the columns can be retrieved as follows:
>>> vectorizer.get_feature_names() == (
... ['and', 'document', 'first', 'is', 'one',
... 'second', 'the', 'third', 'this'])
True
>>> X.toarray()
array([[0, 1, 1, 1, 0, 0, 1, 0, 1],
[0, 1, 0, 1, 0, 2, 1, 0, 1],
[1, 0, 0, 0, 1, 0, 1, 1, 0],
[0, 1, 1, 1, 0, 0, 1, 0, 1]]...)
The converse mapping from feature name to column index is stored in the vocabulary_ attribute of the vectorizer:
>>> vectorizer.vocabulary_.get('document')
1
Hence words that were not seen in the training corpus will be completely ignored in future calls to the transform
method:

>>> vectorizer.transform(['Something completely new.']).toarray()

...
array([[0, 0, 0, 0, 0, 0, 0, 0, 0]]...)
Note that in the previous corpus, the first and the last documents have exactly the same words hence are encoded in
equal vectors. In particular we lose the information that the last document is an interrogative form. To preserve some
of the local ordering information we can extract 2-grams of words in addition to the 1-grams (individual words):
>>> bigram_vectorizer = CountVectorizer(ngram_range=(1, 2),

... token_pattern=r'\b\w+\b', min_df=1)
>>> analyze = bigram_vectorizer.build_analyzer()
>>> analyze('Bi-grams are cool!') == (
... ['bi', 'grams', 'are', 'cool', 'bi grams', 'grams are', 'are cool'])
True
The vocabulary extracted by this vectorizer is hence much bigger and can now resolve ambiguities encoded in local
positioning patterns:
>>> X_2 = bigram_vectorizer.fit_transform(corpus).toarray()

>>> X_2
...
array([[0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 1, 0],
[0, 0, 1, 0, 0, 1, 1, 0, 0, 2, 1, 1, 1, 0, 1, 0, 0, 0, 1, 1, 0],
[1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 1, 1, 0, 0, 0],
[0, 0, 1, 1, 1, 1, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0, 1]]...)
In particular the interrogative form “Is this” is only present in the last document:
>>> feature_index = bigram_vectorizer.vocabulary_.get('is this')

>>> X_2[:, feature_index]
array([0, 0, 0, 1]...)
Using stop words
Stop words are words like “and”, “the”, “him”, which are presumed to be uninformative in representing the content
of a text, and which may be removed to avoid them being construed as signal for prediction. Sometimes, however,
similar words are useful for prediction, such as in classifying writing style or personality.
There are several known issues in our provided ‘english’ stop word list. See [NQY18].
Please take care in choosing a stop word list. Popular stop word lists may include words that are highly informative to
some tasks, such as computer.
You should also make sure that the stop word list has had the same preprocessing and tokenization applied as the one
used in the vectorizer. The word we’ve is split into we and ve by CountVectorizer’s default tokenizer, so if we’ve is in
stop_words, but ve is not, ve will be retained from we’ve in transformed text. Our vectorizers will try to identify
and warn about some kinds of inconsistencies.
References

Tf–idf term weighting
In a large text corpus, some words will be very present (e.g. “the”, “a”, “is” in English) hence carrying very little
meaningful information about the actual contents of the document. If we were to feed the direct count data directly to
a classifier those very frequent terms would shadow the frequencies of rarer yet more interesting terms.
In order to re-weight the count features into floating point values suitable for usage by a classifier it is very common
to use the tf–idf transform.
Tf means term-frequency while tf–idf means term-frequency times inverse document-frequency: tf-idf(t,d) =
tf(t,d) × idf(t).
Using the TfidfTransformer’s default settings, TfidfTransformer(norm='l2', use_idf=True,
smooth_idf=True, sublinear_tf=False) the term frequency, the number of times a term occurs in a
given document, is multiplied with idf component, which is computed as
1+𝑛
idf(𝑡) = log 1+df(𝑡) + 1,
where 𝑛 is the total number of documents in the document set, and df(𝑡) is the number of documents in the document
set that contain term 𝑡. The resulting tf-idf vectors are then normalized by the Euclidean norm:
𝑣 𝑣
𝑣𝑛𝑜𝑟𝑚 = ||𝑣||2 = √
𝑣 1 2 +𝑣 2 2 +···+𝑣 𝑛 2
.
This was originally a term weighting scheme developed for information retrieval (as a ranking function for search
engines results) that has also found good use in document classification and clustering.
The following sections contain further explanations and examples that illustrate how the tf-idfs are computed exactly
and how the tf-idfs computed in scikit-learn’s TfidfTransformer and TfidfVectorizer differ slightly from
the standard textbook notation that defines the idf as
𝑛
idf(𝑡) = log 1+df(𝑡) .
In the TfidfTransformer and TfidfVectorizer with smooth_idf=False, the “1” count is added to the
idf instead of the idf’s denominator:
𝑛
idf(𝑡) = log df(𝑡) +1
This normalization is implemented by the TfidfTransformer class:
>>> from sklearn.feature_extraction.text import TfidfTransformer

>>> transformer = TfidfTransformer(smooth_idf=False)
>>> transformer
TfidfTransformer(norm=...'l2', smooth_idf=False, sublinear_tf=False,
use_idf=True)
Again please see the reference documentation for the details on all the parameters.
Let’s take an example with the following counts. The first term is present 100% of the time hence not very interesting.
The two other features only in less than 50% of the time hence probably more representative of the content of the
documents:
>>> counts = [[3, 0, 1],

... [2, 0, 0],
... [3, 0, 0],
... [4, 0, 0],
... [3, 2, 0],
... [3, 0, 2]]
...
>>> tfidf = transformer.fit_transform(counts)
>>> tfidf

>>> tfidf.toarray()
array([[0.81940995, 0. , 0.57320793],
[1. , 0. , 0. ],
[1. , 0. , 0. ],
[1. , 0. , 0. ],
[0.47330339, 0.88089948, 0. ],
[0.58149261, 0. , 0.81355169]])
Each row is normalized to have unit Euclidean norm:

𝑣 𝑣
𝑣𝑛𝑜𝑟𝑚 = ||𝑣||2 = √
𝑣 1 2 +𝑣 2 2 +···+𝑣 𝑛 2
For example, we can compute the tf-idf of the first term in the first document in the counts array as follows:
𝑛=6
df(𝑡)term1 = 6
𝑛
idf(𝑡)term1 = log df(𝑡) + 1 = log(1) + 1 = 1
tf-idfterm1 = tf × idf = 3 × 1 = 3
Now, if we repeat this computation for the remaining 2 terms in the document, we get
tf-idfterm2 = 0 × (log(6/1) + 1) = 0
tf-idfterm3 = 1 × (log(6/2) + 1) ≈ 2.0986
and the vector of raw tf-idfs:
tf-idfraw = [3, 0, 2.0986].
Then, applying the Euclidean (L2) norm, we obtain the following tf-idfs for document 1:
[3,0,2.0986]
√︂(︀ )︀ = [0.819, 0, 0.573].
32 +02 +2.09862
Furthermore, the default parameter smooth_idf=True adds “1” to the numerator and denominator as if an extra
document was seen containing every term in the collection exactly once, which prevents zero divisions:
1+𝑛
idf(𝑡) = log 1+df(𝑡) +1
Using this modification, the tf-idf of the third term in document 1 changes to 1.8473:
tf-idfterm3 = 1 × log(7/3) + 1 ≈ 1.8473
And the L2-normalized tf-idf changes to
[3,0,1.8473]
√︂(︀ )︀ = [0.8515, 0, 0.5243]:
32 +02 +1.84732
>>> transformer = TfidfTransformer()

>>> transformer.fit_transform(counts).toarray()
array([[0.85151335, 0. , 0.52433293],
[1. , 0. , 0. ],
[1. , 0. , 0. ],
[1. , 0. , 0. ],
[0.55422893, 0.83236428, 0. ],
[0.63035731, 0. , 0.77630514]])
The weights of each feature computed by the fit method call are stored in a model attribute:

>>> transformer.idf_
array([1. ..., 2.25..., 1.84...])
As tf–idf is very often used for text features, there is also another class called TfidfVectorizer that combines all
the options of CountVectorizer and TfidfTransformer in a single model:
>>> from sklearn.feature_extraction.text import TfidfVectorizer

>>> vectorizer = TfidfVectorizer()
>>> vectorizer.fit_transform(corpus)
...
While the tf–idf normalization is often very useful, there might be cases where the binary occurrence markers might
offer better features. This can be achieved by using the binary parameter of CountVectorizer. In particular,
some estimators such as Bernoulli Naive Bayes explicitly model discrete boolean random variables. Also, very short
texts are likely to have noisy tf–idf values while the binary occurrence info is more stable.
As usual the best way to adjust the feature extraction parameters is to use a cross-validated grid search, for instance by
pipelining the feature extractor with a classifier:
Decoding text files
Text is made of characters, but files are made of bytes. These bytes represent characters according to some encoding.
To work with text files in Python, their bytes must be decoded to a character set called Unicode. Common encodings
are ASCII, Latin-1 (Western Europe), KOI8-R (Russian) and the universal encodings UTF-8 and UTF-16. Many
others exist.
Note: An encoding can also be called a ‘character set’, but this term is less accurate: several encodings can exist for
a single character set.
The text feature extractors in scikit-learn know how to decode text files, but only if you tell them what encoding the
files are in. The CountVectorizer takes an encoding parameter for this purpose. For modern text files, the
correct encoding is probably UTF-8, which is therefore the default (encoding="utf-8").
If the text you are loading is not actually encoded with UTF-8, however, you will get a UnicodeDecodeError.
The vectorizers can be told to be silent about decoding errors by setting the decode_error parameter to either
"ignore" or "replace". See the documentation for the Python function bytes.decode for more details (type
help(bytes.decode) at the Python prompt).
If you are having trouble decoding text, here are some things to try:
• Find out what the actual encoding of the text is. The file might come with a header or README that tells you
the encoding, or there might be some standard encoding you can assume based on where the text comes from.
• You may be able to find out what kind of encoding it is in general using the UNIX command file. The Python
chardet module comes with a script called chardetect.py that will guess the specific encoding, though
you cannot rely on its guess being correct.
• You could try UTF-8 and disregard the errors. You can decode byte strings with bytes.
decode(errors='replace') to replace all decoding errors with a meaningless character, or set
decode_error='replace' in the vectorizer. This may damage the usefulness of your features.

• Real text may come from a variety of sources that may have used different encodings, or even be sloppily
decoded in a different encoding than the one it was encoded with. This is common in text retrieved from the
Web. The Python package ftfy can automatically sort out some classes of decoding errors, so you could try
decoding the unknown text as latin-1 and then using ftfy to fix errors.
• If the text is in a mish-mash of encodings that is simply too hard to sort out (which is the case for the 20
Newsgroups dataset), you can fall back on a simple single-byte encoding such as latin-1. Some text may
display incorrectly, but at least the same sequence of bytes will always represent the same feature.
For example, the following snippet uses chardet (not shipped with scikit-learn, must be installed separately) to
figure out the encoding of three texts. It then vectorizes the texts and prints the learned vocabulary. The output is not
shown here.
>>> import chardet

>>> text1 = b"Sei mir gegr\xc3\xbc\xc3\x9ft mein Sauerkraut"
>>> text2 = b"holdselig sind deine Ger\xfcche"
>>> text3 = b"\xff\xfeA\x00u\x00f\x00 \x00F\x00l\x00\xfc\x00g\x00e\x00l\x00n\x00
˓→\x00d\x00e\x00s\x00 \x00G\x00e\x00s\x00a\x00n\x00g\x00e\x00s\x00,\x00
˓→\x00H\x00e\x00r\x00z\x00l\x00i\x00e\x00b\x00c\x00h\x00e\x00n\x00,\x00
˓→\x00t\x00r\x00a\x00g\x00 \x00i\x00c\x00h\x00 \x00d\x00i\x00c\x00h\x00
˓→\x00f\x00o\x00r\x00t\x00"
>>> decoded = [x.decode(chardet.detect(x)['encoding'])

... for x in (text1, text2, text3)]
>>> v = CountVectorizer().fit(decoded).vocabulary_
>>> for term in v: print(v)
(Depending on the version of chardet, it might get the first one wrong.)
For an introduction to Unicode and character encodings in general, see Joel Spolsky’s Absolute Minimum Every
Software Developer Must Know About Unicode.
Applications and examples
The bag of words representation is quite simplistic but surprisingly useful in practice.
In particular in a supervised setting it can be successfully combined with fast and scalable linear models to train
document classifiers, for instance:
In an unsupervised setting it can be used to group similar documents together by applying clustering algorithms such
as K-means:
Finally it is possible to discover the main topics of a corpus by relaxing the hard assignment constraint of clustering,
for instance by using Non-negative matrix factorization (NMF or NNMF):
Limitations of the Bag of Words representation
A collection of unigrams (what bag of words is) cannot capture phrases and multi-word expressions, effectively disre-
garding any word order dependence. Additionally, the bag of words model doesn’t account for potential misspellings
or word derivations.
N-grams to the rescue! Instead of building a simple collection of unigrams (n=1), one might prefer a collection of
bigrams (n=2), where occurrences of pairs of consecutive words are counted.

One might alternatively consider a collection of character n-grams, a representation resilient against misspellings and
derivations.
For example, let’s say we’re dealing with a corpus of two documents: ['words', 'wprds']. The second docu-
ment contains a misspelling of the word ‘words’. A simple bag of words representation would consider these two as
very distinct documents, differing in both of the two possible features. A character 2-gram representation, however,
would find the documents matching in 4 out of 8 features, which may help the preferred classifier decide better:
>>> ngram_vectorizer = CountVectorizer(analyzer='char_wb', ngram_range=(2, 2))
>>> counts = ngram_vectorizer.fit_transform(['words', 'wprds'])
>>> ngram_vectorizer.get_feature_names() == (
... [' w', 'ds', 'or', 'pr', 'rd', 's ', 'wo', 'wp'])
True
>>> counts.toarray().astype(int)
array([[1, 1, 1, 0, 1, 1, 1, 0],
[1, 1, 0, 1, 1, 1, 0, 1]])
In the above example, char_wb analyzer is used, which creates n-grams only from characters inside word boundaries
(padded with space on each side). The char analyzer, alternatively, creates n-grams that span across words:
>>> ngram_vectorizer = CountVectorizer(analyzer='char_wb', ngram_range=(5, 5))
>>> ngram_vectorizer.fit_transform(['jumpy fox'])
...
... [' fox ', ' jump', 'jumpy', 'umpy '])
True
>>> ngram_vectorizer = CountVectorizer(analyzer='char', ngram_range=(5, 5))

>>> ngram_vectorizer.fit_transform(['jumpy fox'])
...
... ['jumpy', 'mpy f', 'py fo', 'umpy ', 'y fox'])
True
The word boundaries-aware variant char_wb is especially interesting for languages that use white-spaces for word
separation as it generates significantly less noisy features than the raw char variant in that case. For such languages
it can increase both the predictive accuracy and convergence speed of classifiers trained using such features while
retaining the robustness with regards to misspellings and word derivations.
While some local positioning information can be preserved by extracting n-grams instead of individual words, bag of
words and bag of n-grams destroy most of the inner structure of the document and hence most of the meaning carried
by that internal structure.
In order to address the wider task of Natural Language Understanding, the local structure of sentences and paragraphs
should thus be taken into account. Many such models will thus be casted as “Structured output” problems which are
currently outside of the scope of scikit-learn.
Vectorizing a large text corpus with the hashing trick
The above vectorization scheme is simple but the fact that it holds an in- memory mapping from the string tokens
to the integer feature indices (the vocabulary_ attribute) causes several problems when dealing with large
datasets:
• the larger the corpus, the larger the vocabulary will grow and hence the memory use too,

• fitting requires the allocation of intermediate data structures of size proportional to that of the original dataset.
• building the word-mapping requires a full pass over the dataset hence it is not possible to fit text classifiers in a
strictly online manner.
• pickling and un-pickling vectorizers with a large vocabulary_ can be very slow (typically much slower than
pickling / un-pickling flat data structures such as a NumPy array of the same size),
• it is not easily possible to split the vectorization work into concurrent sub tasks as the vocabulary_ attribute
would have to be a shared state with a fine grained synchronization barrier: the mapping from token string
to feature index is dependent on ordering of the first occurrence of each token hence would have to be shared,
potentially harming the concurrent workers’ performance to the point of making them slower than the sequential
variant.
It is possible to overcome those limitations by combining the “hashing trick” (Feature hashing) implemented by the
sklearn.feature_extraction.FeatureHasher class and the text preprocessing and tokenization features
of the CountVectorizer.
This combination is implementing in HashingVectorizer, a transformer class that is mostly API compatible with
CountVectorizer. HashingVectorizer is stateless, meaning that you don’t have to call fit on it:
>>> from sklearn.feature_extraction.text import HashingVectorizer

>>> hv = HashingVectorizer(n_features=10)
>>> hv.transform(corpus)
...
You can see that 16 non-zero feature tokens were extracted in the vector output: this is less than the 19 non-zeros
extracted previously by the CountVectorizer on the same toy corpus. The discrepancy comes from hash function
collisions because of the low value of the n_features parameter.
In a real world setting, the n_features parameter can be left to its default value of 2 ** 20 (roughly one million
possible features). If memory or downstream models size is an issue selecting a lower value such as 2 ** 18 might
help without introducing too many additional collisions on typical text classification tasks.
Note that the dimensionality does not affect the CPU training time of algorithms which operate on CSR matrices
(LinearSVC(dual=True), Perceptron, SGDClassifier, PassiveAggressive) but it does for algo-
rithms that work with CSC matrices (LinearSVC(dual=False), Lasso(), etc).
Let’s try again with the default setting:
>>> hv = HashingVectorizer()
>>> hv.transform(corpus)
...
We no longer get the collisions, but this comes at the expense of a much larger dimensionality of the output space. Of
course, other terms than the 19 used here might still collide with each other.
The HashingVectorizer also comes with the following limitations:
• it is not possible to invert the model (no inverse_transform method), nor to access the original string
representation of the features, because of the one-way nature of the hash function that performs the mapping.
• it does not provide IDF weighting as that would introduce statefulness in the model. A TfidfTransformer
can be appended to it in a pipeline if required.

Performing out-of-core scaling with HashingVectorizer
An interesting development of using a HashingVectorizer is the ability to perform out-of-core scaling. This
means that we can learn from data that does not fit into the computer’s main memory.
A strategy to implement out-of-core scaling is to stream data to the estimator in mini-batches. Each mini-batch is
vectorized using HashingVectorizer so as to guarantee that the input space of the estimator has always the same
dimensionality. The amount of memory used at any time is thus bounded by the size of a mini-batch. Although there is
no limit to the amount of data that can be ingested using such an approach, from a practical point of view the learning
time is often limited by the CPU time one wants to spend on the task.
For a full-fledged example of out-of-core scaling in a text classification task see Out-of-core classification of text
documents.
Customizing the vectorizer classes
It is possible to customize the behavior by passing a callable to the vectorizer constructor:

>>> def my_tokenizer(s):
... return s.split()
...
>>> vectorizer = CountVectorizer(tokenizer=my_tokenizer)
>>> vectorizer.build_analyzer()(u"Some... punctuation!") == (
... ['some...', 'punctuation!'])
True
In particular we name:
• preprocessor: a callable that takes an entire document as input (as a single string), and returns a possibly
transformed version of the document, still as an entire string. This can be used to remove HTML tags, lowercase
the entire document, etc.
• tokenizer: a callable that takes the output from the preprocessor and splits it into tokens, then returns a list
of these.
• analyzer: a callable that replaces the preprocessor and tokenizer. The default analyzers all call the prepro-
cessor and tokenizer, but custom analyzers will skip this. N-gram extraction and stop word filtering take place
at the analyzer level, so a custom analyzer may have to reproduce these steps.
(Lucene users might recognize these names, but be aware that scikit-learn concepts may not map one-to-one onto
Lucene concepts.)
To make the preprocessor, tokenizer and analyzers aware of the model parameters it is possible to derive from the class
and override the build_preprocessor, build_tokenizer and build_analyzer factory methods instead
of passing custom functions.
Some tips and tricks:
• If documents are pre-tokenized by an external package, then store them in files (or strings) with the tokens
separated by whitespace and pass analyzer=str.split
• Fancy token-level analysis such as stemming, lemmatizing, compound splitting, filtering based on part-of-
speech, etc. are not included in the scikit-learn codebase, but can be added by customizing either the tokenizer
or the analyzer. Here’s a CountVectorizer with a tokenizer and lemmatizer using NLTK:
>>> from nltk import word_tokenize
>>> from nltk.stem import WordNetLemmatizer
>>> class LemmaTokenizer(object):
... def __init__(self):

... self.wnl = WordNetLemmatizer()

... def __call__(self, doc):
... return [self.wnl.lemmatize(t) for t in word_tokenize(doc)]
...
>>> vect = CountVectorizer(tokenizer=LemmaTokenizer())
(Note that this will not filter out punctuation.)

The following example will, for instance, transform some British spelling to American spelling:
>>> import re
>>> def to_british(tokens):
... for t in tokens:
... t = re.sub(r"(...)our$", r"\1or", t)
... t = re.sub(r"([bt])re$", r"\1er", t)
... t = re.sub(r"([iy])s(e$|ing|ation)", r"\1z\2", t)
... t = re.sub(r"ogue$", "og", t)
... yield t
...
>>> class CustomVectorizer(CountVectorizer):
... def build_tokenizer(self):
... tokenize = super(CustomVectorizer, self).build_tokenizer()
... return lambda doc: list(to_british(tokenize(doc)))
...
>>> print(CustomVectorizer().build_analyzer()(u"color colour"))
[...'color', ...'color']
for other styles of preprocessing; examples include stemming, lemmatization, or normalizing numerical tokens,
with the latter illustrated in:
– Biclustering documents with the Spectral Co-clustering algorithm
Customizing the vectorizer can also be useful when handling Asian languages that do not use an explicit word separator
such as whitespace.
Image feature extraction
Patch extraction
The extract_patches_2d function extracts patches from an image stored as a two-dimensional array, or
three-dimensional with color information along the third axis. For rebuilding an image from all its patches, use
reconstruct_from_patches_2d. For example let use generate a 4x4 pixel picture with 3 color channels (e.g.
in RGB format):
>>> from sklearn.feature_extraction import image
>>> one_image = np.arange(4 * 4 * 3).reshape((4, 4, 3))

>>> one_image[:, :, 0] # R channel of a fake RGB picture
array([[ 0, 3, 6, 9],
[12, 15, 18, 21],
[24, 27, 30, 33],
[36, 39, 42, 45]])
>>> patches = image.extract_patches_2d(one_image, (2, 2), max_patches=2,

... random_state=0)
>>> patches.shape

(2, 2, 2, 3)
>>> patches[:, :, :, 0]
array([[[ 0, 3],
[12, 15]],
[[15, 18],
[27, 30]]])
>>> patches = image.extract_patches_2d(one_image, (2, 2))
>>> patches.shape
(9, 2, 2, 3)
>>> patches[4, :, :, 0]
array([[15, 18],
[27, 30]])
Let us now try to reconstruct the original image from the patches by averaging on overlapping areas:
>>> reconstructed = image.reconstruct_from_patches_2d(patches, (4, 4, 3))

>>> np.testing.assert_array_equal(one_image, reconstructed)
The PatchExtractor class works in the same way as extract_patches_2d, only it supports multiple images
as input. It is implemented as an estimator, so it can be used in pipelines. See:
>>> five_images = np.arange(5 * 4 * 4 * 3).reshape(5, 4, 4, 3)

>>> patches = image.PatchExtractor((2, 2)).transform(five_images)
>>> patches.shape
(45, 2, 2, 3)
Connectivity graph of an image
Several estimators in the scikit-learn can use connectivity information between features or samples. For instance Ward
clustering (Hierarchical clustering) can cluster together only neighboring pixels of an image, thus forming contiguous
patches:
For this purpose, the estimators use a ‘connectivity’ matrix, giving which samples are connected.
The function img_to_graph returns such a matrix from a 2D or 3D image. Similarly, grid_to_graph build a
connectivity matrix for images given the shape of these image.
These matrices can be used to impose connectivity in estimators that use connectivity information, such as Ward
clustering (Hierarchical clustering), but also to build precomputed kernels, or similarity matrices.

Note: Examples
• A demo of structured Ward hierarchical clustering on an image of coins
• Spectral clustering for image segmentation
• Feature agglomeration vs. univariate selection
3.4.3 Preprocessing data
The sklearn.preprocessing package provides several common utility functions and transformer classes to
change raw feature vectors into a representation that is more suitable for the downstream estimators.
In general, learning algorithms benefit from standardization of the data set. If some outliers are present in the set, robust
scalers or transformers are more appropriate. The behaviors of the different scalers, transformers, and normalizers on
a dataset containing marginal outliers is highlighted in Compare the effect of different scalers on data with outliers.
Standardization, or mean removal and variance scaling
Standardization of datasets is a common requirement for many machine learning estimators implemented in
scikit-learn; they might behave badly if the individual features do not more or less look like standard normally dis-
tributed data: Gaussian with zero mean and unit variance.
In practice we often ignore the shape of the distribution and just transform the data to center it by removing the mean
value of each feature, then scale it by dividing non-constant features by their standard deviation.
For instance, many elements used in the objective function of a learning algorithm (such as the RBF kernel of Support
Vector Machines or the l1 and l2 regularizers of linear models) assume that all features are centered around zero and
have variance in the same order. If a feature has a variance that is orders of magnitude larger than others, it might
dominate the objective function and make the estimator unable to learn from other features correctly as expected.
The function scale provides a quick and easy way to perform this operation on a single array-like dataset:

>>> X_train = np.array([[ 1., -1., 2.],
... [ 2., 0., 0.],
... [ 0., 1., -1.]])
>>> X_scaled = preprocessing.scale(X_train)
>>> X_scaled
array([[ 0. ..., -1.22..., 1.33...],
[ 1.22..., 0. ..., -0.26...],
[-1.22..., 1.22..., -1.06...]])
Scaled data has zero mean and unit variance:
>>> X_scaled.mean(axis=0)
array([0., 0., 0.])
>>> X_scaled.std(axis=0)
array([1., 1., 1.])
The preprocessing module further provides a utility class StandardScaler that implements the
Transformer API to compute the mean and standard deviation on a training set so as to be able to later reapply

the same transformation on the testing set. This class is hence suitable for use in the early steps of a sklearn.
pipeline.Pipeline:
>>> scaler = preprocessing.StandardScaler().fit(X_train)

>>> scaler
StandardScaler(copy=True, with_mean=True, with_std=True)
>>> scaler.mean_
array([1. ..., 0. ..., 0.33...])
>>> scaler.scale_
array([0.81..., 0.81..., 1.24...])
>>> scaler.transform(X_train)
array([[ 0. ..., -1.22..., 1.33...],
[ 1.22..., 0. ..., -0.26...],
[-1.22..., 1.22..., -1.06...]])
The scaler instance can then be used on new data to transform it the same way it did on the training set:
>>> X_test = [[-1., 1., 0.]]

>>> scaler.transform(X_test)
array([[-2.44..., 1.22..., -0.26...]])
It is possible to disable either centering or scaling by either passing with_mean=False or with_std=False to

the constructor of StandardScaler.
Scaling features to a range
An alternative standardization is scaling features to lie between a given minimum and maximum value, often between
zero and one, or so that the maximum absolute value of each feature is scaled to unit size. This can be achieved using
MinMaxScaler or MaxAbsScaler, respectively.
The motivation to use this scaling include robustness to very small standard deviations of features and preserving zero
entries in sparse data.
Here is an example to scale a toy data matrix to the [0, 1] range:
>>> X_train = np.array([[ 1., -1., 2.],

... [ 2., 0., 0.],
... [ 0., 1., -1.]])
...
>>> min_max_scaler = preprocessing.MinMaxScaler()
>>> X_train_minmax = min_max_scaler.fit_transform(X_train)
>>> X_train_minmax
array([[0.5 , 0. , 1. ],
[1. , 0.5 , 0.33333333],
[0. , 1. , 0. ]])
The same instance of the transformer can then be applied to some new test data unseen during the fit call: the same
scaling and shifting operations will be applied to be consistent with the transformation performed on the train data:
>>> X_test = np.array([[-3., -1., 4.]])

>>> X_test_minmax = min_max_scaler.transform(X_test)
>>> X_test_minmax
array([[-1.5 , 0. , 1.66666667]])

It is possible to introspect the scaler attributes to find about the exact nature of the transformation learned on the
training data:
>>> min_max_scaler.scale_
array([0.5 , 0.5 , 0.33...])
>>> min_max_scaler.min_
array([0. , 0.5 , 0.33...])
If MinMaxScaler is given an explicit feature_range=(min, max) the full formula is:
X_std = (X - X.min(axis=0)) / (X.max(axis=0) - X.min(axis=0))
X_scaled = X_std * (max - min) + min
MaxAbsScaler works in a very similar fashion, but scales in a way that the training data lies within the range [-1,
1] by dividing through the largest maximum value in each feature. It is meant for data that is already centered at zero
or sparse data.
Here is how to use the toy data from the previous example with this scaler:
>>> X_train = np.array([[ 1., -1., 2.],

... [ 2., 0., 0.],
... [ 0., 1., -1.]])
...
>>> max_abs_scaler = preprocessing.MaxAbsScaler()
>>> X_train_maxabs = max_abs_scaler.fit_transform(X_train)
>>> X_train_maxabs # doctest +NORMALIZE_WHITESPACE^
array([[ 0.5, -1. , 1. ],
[ 1. , 0. , 0. ],
[ 0. , 1. , -0.5]])
>>> X_test = np.array([[ -3., -1., 4.]])
>>> X_test_maxabs = max_abs_scaler.transform(X_test)
>>> X_test_maxabs
array([[-1.5, -1. , 2. ]])
>>> max_abs_scaler.scale_
array([2., 1., 2.])
As with scale, the module further provides convenience functions minmax_scale and maxabs_scale if you
don’t want to create an object.
Scaling sparse data
Centering sparse data would destroy the sparseness structure in the data, and thus rarely is a sensible thing to do.
However, it can make sense to scale sparse inputs, especially if features are on different scales.
MaxAbsScaler and maxabs_scale were specifically designed for scaling sparse data, and are the recommended
way to go about this. However, scale and StandardScaler can accept scipy.sparse matrices as input, as
long as with_mean=False is explicitly passed to the constructor. Otherwise a ValueError will be raised as
silently centering would break the sparsity and would often crash the execution by allocating excessive amounts of
memory unintentionally. RobustScaler cannot be fitted to sparse inputs, but you can use the transform method
on sparse inputs.
Note that the scalers accept both Compressed Sparse Rows and Compressed Sparse Columns format (see scipy.
sparse.csr_matrix and scipy.sparse.csc_matrix). Any other sparse input will be converted to the
Compressed Sparse Rows representation. To avoid unnecessary memory copies, it is recommended to choose the
CSR or CSC representation upstream.

Finally, if the centered data is expected to be small enough, explicitly converting the input to an array using the
toarray method of sparse matrices is another option.
Scaling data with outliers
If your data contains many outliers, scaling using the mean and variance of the data is likely to not work very well.
In these cases, you can use robust_scale and RobustScaler as drop-in replacements instead. They use more
robust estimates for the center and range of your data.
References:
Further discussion on the importance of centering and scaling data is available on this FAQ: Should I normal-
ize/standardize/rescale the data?
Scaling vs Whitening
It is sometimes not enough to center and scale the features independently, since a downstream model can further
make some assumption on the linear independence of the features.
To address this issue you can use sklearn.decomposition.PCA with whiten=True to further remove the
linear correlation across features.
Scaling a 1D array
All above functions (i.e. scale, minmax_scale, maxabs_scale, and robust_scale) accept 1D array
which can be useful in some specific case.
Centering kernel matrices
If you have a kernel matrix of a kernel 𝐾 that computes a dot product in a feature space defined by function 𝑝ℎ𝑖, a
KernelCenterer can transform the kernel matrix so that it contains inner products in the feature space defined by
𝑝ℎ𝑖 followed by removal of the mean in that space.
Non-linear transformation
Mapping to a Uniform distribution
Like scalers, QuantileTransformer puts all features into the same, known range or distribution. However, by
performing a rank transformation, it smooths out unusual distributions and is less influenced by outliers than scaling
methods. It does, however, distort correlations and distances within and across features.
QuantileTransformer and quantile_transform provide a non-parametric transformation based on the
quantile function to map the data to a uniform distribution with values between 0 and 1:



>>> quantile_transformer = preprocessing.QuantileTransformer(random_state=0)
>>> X_train_trans = quantile_transformer.fit_transform(X_train)
>>> X_test_trans = quantile_transformer.transform(X_test)
>>> np.percentile(X_train[:, 0], [0, 25, 50, 75, 100])
array([ 4.3, 5.1, 5.8, 6.5, 7.9])
This feature corresponds to the sepal length in cm. Once the quantile transformation applied, those landmarks approach
closely the percentiles previously defined:
>>> np.percentile(X_train_trans[:, 0], [0, 25, 50, 75, 100])
...
array([ 0.00... , 0.24..., 0.49..., 0.73..., 0.99... ])
This can be confirmed on a independent testing set with similar remarks:

>>> np.percentile(X_test[:, 0], [0, 25, 50, 75, 100])
...
array([ 4.4 , 5.125, 5.75 , 6.175, 7.3 ])
>>> np.percentile(X_test_trans[:, 0], [0, 25, 50, 75, 100])
...
array([ 0.01..., 0.25..., 0.46..., 0.60... , 0.94...])
Mapping to a Gaussian distribution
In many modeling scenarios, normality of the features in a dataset is desirable. Power transforms are a family of
parametric, monotonic transformations that aim to map data from any distribution to as close to a Gaussian distribution
as possible in order to stabilize variance and minimize skewness.
PowerTransformer currently provides two such power transformations, the Yeo-Johnson transform and the Box-
Cox transform.
The Yeo-Johnson transform is given by:
⎧
⎪
⎪ [(𝑥𝑖 + 1)𝜆 − 1]/𝜆 if 𝜆 ̸= 0, 𝑥𝑖 ≥ 0,
⎪
⎪
⎪
if 𝜆 = 0, 𝑥𝑖 ≥ 0
⎪
⎨ln (𝑥𝑖 ) + 1
⎪
(𝜆)
𝑥𝑖 =
⎪−[(−𝑥𝑖 + 1)2−𝜆 − 1]/(2 − 𝜆) if 𝜆 ̸= 2, 𝑥𝑖 < 0,
⎪
⎪
⎪
⎪
⎪
⎪
− ln(−𝑥𝑖 + 1) if 𝜆 = 2, 𝑥𝑖 < 0
⎩
while the Box-Cox transform is given by:

⎧ 𝜆
⎨ 𝑥𝑖 − 1
⎪
if 𝜆 ̸= 0,
(𝜆)
𝑥𝑖 = 𝜆
⎪
⎩ln (𝑥 )
𝑖 if 𝜆 = 0,
Box-Cox can only be applied to strictly positive data. In both methods, the transformation is parameterized by 𝜆,
which is determined through maximum likelihood estimation. Here is an example of using Box-Cox to map samples
drawn from a lognormal distribution to a normal distribution:
>>> pt = preprocessing.PowerTransformer(method='box-cox', standardize=False)
>>> X_lognormal = np.random.RandomState(616).lognormal(size=(3, 3))
>>> X_lognormal
array([[1.28..., 1.18..., 0.84...],

[0.94..., 1.60..., 0.38...],

[1.35..., 0.21..., 1.09...]])
>>> pt.fit_transform(X_lognormal)
array([[ 0.49..., 0.17..., -0.15...],
[-0.05..., 0.58..., -0.57...],
[ 0.69..., -0.84..., 0.10...]])
While the above example sets the standardize option to False, PowerTransformer will apply zero-mean, unit-
variance normalization to the transformed output by default.
Below are examples of Box-Cox and Yeo-Johnson applied to various probability distributions. Note that when applied
to certain distributions, the power transforms achieve very Gaussian-like results, but with others, they are ineffective.
This highlights the importance of visualizing the data before and after transformation.
It is also possible to map data to a normal distribution using QuantileTransformer by setting
output_distribution='normal'. Using the earlier example with the iris dataset:
>>> quantile_transformer = preprocessing.QuantileTransformer(

... output_distribution='normal', random_state=0)
>>> X_trans = quantile_transformer.fit_transform(X)
>>> quantile_transformer.quantiles_
array([[4.3..., 2..., 1..., 0.1...],
[4.31..., 2.02..., 1.01..., 0.1...],
[4.32..., 2.05..., 1.02..., 0.1...],
...,
[7.84..., 4.34..., 6.84..., 2.5...],
[7.87..., 4.37..., 6.87..., 2.5...],
[7.9..., 4.4..., 6.9..., 2.5...]])
Thus the median of the input becomes the mean of the output, centered at 0. The normal output is clipped so that the
input’s minimum and maximum — corresponding to the 1e-7 and 1 - 1e-7 quantiles respectively — do not become
infinite under the transformation.
Normalization
Normalization is the process of scaling individual samples to have unit norm. This process can be useful if you plan
to use a quadratic form such as the dot-product or any other kernel to quantify the similarity of any pair of samples.
This assumption is the base of the Vector Space Model often used in text classification and clustering contexts.
The function normalize provides a quick and easy way to perform this operation on a single array-like dataset,
either using the l1 or l2 norms:
>>> X = [[ 1., -1., 2.],

... [ 2., 0., 0.],
... [ 0., 1., -1.]]
>>> X_normalized = preprocessing.normalize(X, norm='l2')
>>> X_normalized
array([[ 0.40..., -0.40..., 0.81...],
[ 1. ..., 0. ..., 0. ...],
[ 0. ..., 0.70..., -0.70...]])
The preprocessing module further provides a utility class Normalizer that implements the same operation
using the Transformer API (even though the fit method is useless in this case: the class is stateless as this
operation treats samples independently).
This class is hence suitable for use in the early steps of a sklearn.pipeline.Pipeline:


>>> normalizer = preprocessing.Normalizer().fit(X) # fit does nothing

>>> normalizer
Normalizer(copy=True, norm='l2')
The normalizer instance can then be used on sample vectors as any transformer:
>>> normalizer.transform(X)
array([[ 0.40..., -0.40..., 0.81...],
[ 1. ..., 0. ..., 0. ...],
[ 0. ..., 0.70..., -0.70...]])
>>> normalizer.transform([[-1., 1., 0.]])

array([[-0.70..., 0.70..., 0. ...]])
Sparse input
normalize and Normalizer accept both dense array-like and sparse matrices from scipy.sparse as input.
For sparse input the data is converted to the Compressed Sparse Rows representation (see scipy.sparse.
csr_matrix) before being fed to efficient Cython routines. To avoid unnecessary memory copies, it is recom-
mended to choose the CSR representation upstream.
Encoding categorical features
Often features are not given as continuous values but categorical. For example a person could have fea-
tures ["male", "female"], ["from Europe", "from US", "from Asia"], ["uses Firefox",
"uses Chrome", "uses Safari", "uses Internet Explorer"]. Such features can be efficiently
coded as integers, for instance ["male", "from US", "uses Internet Explorer"] could be expressed
as [0, 1, 3] while ["female", "from Asia", "uses Chrome"] would be [1, 2, 1].
To convert categorical features to such integer codes, we can use the OrdinalEncoder. This estimator transforms
each categorical feature to one new feature of integers (0 to n_categories - 1):
>>> enc = preprocessing.OrdinalEncoder()
>>> X = [['male', 'from US', 'uses Safari'], ['female', 'from Europe', 'uses Firefox
˓→']]
>>> enc.fit(X)
OrdinalEncoder(categories='auto', dtype=<... 'numpy.float64'>)
>>> enc.transform([['female', 'from US', 'uses Safari']])
array([[0., 1., 1.]])
Such integer representation can, however, not be used directly with all scikit-learn estimators, as these expect continu-
ous input, and would interpret the categories as being ordered, which is often not desired (i.e. the set of browsers was
ordered arbitrarily).
Another possibility to convert categorical features to features that can be used with scikit-learn estimators is
to use a one-of-K, also known as one-hot or dummy encoding. This type of encoding can be obtained with
the OneHotEncoder, which transforms each categorical feature with n_categories possible values into
n_categories binary features, with one of them 1, and all others 0.
Continuing the example above:
>>> enc = preprocessing.OneHotEncoder()
˓→']]

>>> enc.fit(X)
OneHotEncoder(categorical_features=None, categories=None,
dtype=<... 'numpy.float64'>, handle_unknown='error',
n_values=None, sparse=True)
>>> enc.transform([['female', 'from US', 'uses Safari'],
... ['male', 'from Europe', 'uses Safari']]).toarray()
array([[1., 0., 0., 1., 0., 1.],
[0., 1., 1., 0., 0., 1.]])
By default, the values each feature can take is inferred automatically from the dataset and can be found in the
categories_ attribute:
>>> enc.categories_
[array(['female', 'male'], dtype=object), array(['from Europe', 'from US'],
˓→dtype=object), array(['uses Firefox', 'uses Safari'], dtype=object)]
It is possible to specify this explicitly using the parameter categories. There are two genders, four possible
continents and four web browsers in our dataset:
>>> genders = ['female', 'male']

>>> locations = ['from Africa', 'from Asia', 'from Europe', 'from US']
>>> browsers = ['uses Chrome', 'uses Firefox', 'uses IE', 'uses Safari']
>>> enc = preprocessing.OneHotEncoder(categories=[genders, locations, browsers])
>>> # Note that for there are missing categorical values for the 2nd and 3rd
>>> # feature
˓→']]
>>> enc.fit(X)
OneHotEncoder(categorical_features=None,
categories=[...],
dtype=<... 'numpy.float64'>, handle_unknown='error',
>>> enc.transform([['female', 'from Asia', 'uses Chrome']]).toarray()
array([[1., 0., 0., 1., 0., 0., 1., 0., 0., 0.]])
If there is a possibility that the training data might have missing categorical features, it can often be
better to specify handle_unknown='ignore' instead of setting the categories manually as above.
When handle_unknown='ignore' is specified and unknown categories are encountered during trans-
form, no error will be raised but the resulting one-hot encoded columns for this feature will be all zeros
(handle_unknown='ignore' is only supported for one-hot encoding):
>>> enc = preprocessing.OneHotEncoder(handle_unknown='ignore')

˓→']]
>>> enc.fit(X)
dtype=<... 'numpy.float64'>, handle_unknown='ignore',
>>> enc.transform([['female', 'from Asia', 'uses Chrome']]).toarray()
array([[1., 0., 0., 0., 0., 0.]])
See Loading features from dicts for categorical features that are represented as a dict, not as scalars.
Discretization
Discretization (otherwise known as quantization or binning) provides a way to partition continuous features into dis-

crete values. Certain datasets with continuous features may benefit from discretization, because discretization can
transform the dataset of continuous attributes to one with only nominal attributes.
One-hot encoded discretized features can make a model more expressive, while maintaining interpretability. For
instance, pre-processing with a discretizer can introduce nonlinearity to linear models.
K-bins discretization
KBinsDiscretizer discretizers features into k equal width bins:
>>> X = np.array([[ -3., 5., 15 ],

... [ 0., 6., 14 ],
... [ 6., 3., 11 ]])
>>> est = preprocessing.KBinsDiscretizer(n_bins=[3, 2, 2], encode='ordinal').fit(X)
By default the output is one-hot encoded into a sparse matrix (See Encoding categorical features) and this can be
configured with the encode parameter. For each feature, the bin edges are computed during fit and together with
the number of bins, they will define the intervals. Therefore, for the current example, these intervals are defined as:
• feature 1: [−∞, −1), [−1, 2), [2, ∞)
• feature 2: [−∞, 5), [5, ∞)
• feature 3: [−∞, 14), [14, ∞)
Based on these bin intervals, X is transformed as follows:
>>> est.transform(X)
array([[ 0., 1., 1.],
[ 1., 1., 1.],
[ 2., 0., 0.]])
The resulting dataset contains ordinal attributes which can be further used in a sklearn.pipeline.Pipeline.
Discretization is similar to constructing histograms for continuous data. However, histograms focus on counting
features which fall into particular bins, whereas discretization focuses on assigning feature values to these bins.
KBinsDiscretizer implements different binning strategies, which can be selected with the strategy parame-
ter. The ‘uniform’ strategy uses constant-width bins. The ‘quantile’ strategy uses the quantiles values to have equally
populated bins in each feature. The ‘kmeans’ strategy defines bins based on a k-means clustering procedure performed
on each feature independently.
Examples:
• Using KBinsDiscretizer to discretize continuous features

• Demonstrating the different strategies of KBinsDiscretizer
Feature binarization
Feature binarization is the process of thresholding numerical features to get boolean values. This can be useful for
downstream probabilistic estimators that make assumption that the input data is distributed according to a multi-variate
Bernoulli distribution. For instance, this is the case for the sklearn.neural_network.BernoulliRBM .

It is also common among the text processing community to use binary feature values (probably to simplify the proba-
bilistic reasoning) even if normalized counts (a.k.a. term frequencies) or TF-IDF valued features often perform slightly
better in practice.
As for the Normalizer, the utility class Binarizer is meant to be used in the early stages of sklearn.
pipeline.Pipeline. The fit method does nothing as each sample is treated independently of others:
>>> X = [[ 1., -1., 2.],
... [ 2., 0., 0.],
... [ 0., 1., -1.]]
>>> binarizer = preprocessing.Binarizer().fit(X) # fit does nothing

>>> binarizer
Binarizer(copy=True, threshold=0.0)
>>> binarizer.transform(X)
array([[1., 0., 1.],
[1., 0., 0.],
[0., 1., 0.]])
It is possible to adjust the threshold of the binarizer:

>>> binarizer = preprocessing.Binarizer(threshold=1.1)
>>> binarizer.transform(X)
array([[0., 0., 1.],
[1., 0., 0.],
[0., 0., 0.]])
As for the StandardScaler and Normalizer classes, the preprocessing module provides a companion function
binarize to be used when the transformer API is not necessary.
Note that the Binarizer is similar to the KBinsDiscretizer when k = 2, and when the bin edge is at the
value threshold.
Sparse input
binarize and Binarizer accept both dense array-like and sparse matrices from scipy.sparse as input.
For sparse input the data is converted to the Compressed Sparse Rows representation (see scipy.sparse.
csr_matrix). To avoid unnecessary memory copies, it is recommended to choose the CSR representation up-
stream.
Imputation of missing values
Tools for imputing missing values are discussed at Imputation of missing values.
Generating polynomial features
Often it’s useful to add complexity to the model by considering nonlinear features of the input data. A simple and com-
mon method to use is polynomial features, which can get features’ high-order and interaction terms. It is implemented
in PolynomialFeatures:

>>> X
array([[0, 1],
[2, 3],
[4, 5]])
>>> poly = PolynomialFeatures(2)
array([[ 1., 0., 1., 0., 0., 1.],
[ 1., 2., 3., 4., 6., 9.],
[ 1., 4., 5., 16., 20., 25.]])
The features of X have been transformed from (𝑋1 , 𝑋2 ) to (1, 𝑋1 , 𝑋2 , 𝑋12 , 𝑋1 𝑋2 , 𝑋22 ).
In some cases, only interaction terms among features are required, and it can be gotten with the setting
interaction_only=True:
>>> X
array([[0, 1, 2],
[3, 4, 5],
[6, 7, 8]])
>>> poly = PolynomialFeatures(degree=3, interaction_only=True)
array([[ 1., 0., 1., 2., 0., 0., 2., 0.],
[ 1., 3., 4., 5., 12., 15., 20., 60.],
[ 1., 6., 7., 8., 42., 48., 56., 336.]])
The features of X have been transformed from (𝑋1 , 𝑋2 , 𝑋3 ) to (1, 𝑋1 , 𝑋2 , 𝑋3 , 𝑋1 𝑋2 , 𝑋1 𝑋3 , 𝑋2 𝑋3 , 𝑋1 𝑋2 𝑋3 ).

Note that polynomial features are used implicitly in kernel methods (e.g., sklearn.svm.SVC, sklearn.
decomposition.KernelPCA) when using polynomial Kernel functions.
See Polynomial interpolation for Ridge regression using created polynomial features.
Custom transformers
Often, you will want to convert an existing Python function into a transformer to assist in data cleaning or processing.
You can implement a transformer from an arbitrary function with FunctionTransformer. For example, to build
a transformer that applies a log transformation in a pipeline, do:

>>> from sklearn.preprocessing import FunctionTransformer
>>> transformer = FunctionTransformer(np.log1p, validate=True)
>>> X = np.array([[0, 1], [2, 3]])
>>> transformer.transform(X)
array([[0. , 0.69314718],
[1.09861229, 1.38629436]])
You can ensure that func and inverse_func are the inverse of each other by setting check_inverse=True
and calling fit before transform. Please note that a warning is raised and can be turned into an error with a
filterwarnings:
>>> import warnings

>>> warnings.filterwarnings("error", message=".*check_inverse*.",
... category=UserWarning, append=False)
For a full code example that demonstrates using a FunctionTransformer to do custom feature selection, see
Using FunctionTransformer to select columns

3.4.4 Imputation of missing values
For various reasons, many real world datasets contain missing values, often encoded as blanks, NaNs or other place-
holders. Such datasets however are incompatible with scikit-learn estimators which assume that all values in an array
are numerical, and that all have and hold meaning. A basic strategy to use incomplete datasets is to discard entire rows
and/or columns containing missing values. However, this comes at the price of losing data which may be valuable
(even though incomplete). A better strategy is to impute the missing values, i.e., to infer them from the known part of
the data. See the Glossary of Common Terms and API Elements entry on imputation.
The SimpleImputer class provides basic strategies for imputing missing values. Missing values can be imputed
with a provided constant value, or using the statistics (mean, median or most frequent) of each column in which the
missing values are located. This class also allows for different missing values encodings.
The following snippet demonstrates how to replace missing values, encoded as np.nan, using the mean value of the
columns (axis 0) that contain the missing values:

>>> from sklearn.impute import SimpleImputer
>>> imp = SimpleImputer(missing_values=np.nan, strategy='mean')
>>> imp.fit([[1, 2], [np.nan, 3], [7, 6]])
SimpleImputer(copy=True, fill_value=None, missing_values=nan, strategy='mean',
˓→verbose=0)
>>> X = [[np.nan, 2], [6, np.nan], [7, 6]]

>>> print(imp.transform(X))
[[4. 2. ]
[6. 3.666...]
[7. 6. ]]
The SimpleImputer class also supports sparse matrices:
>>> import scipy.sparse as sp

>>> X = sp.csc_matrix([[1, 2], [0, -1], [8, 4]])
>>> imp = SimpleImputer(missing_values=-1, strategy='mean')
>>> imp.fit(X)
SimpleImputer(copy=True, fill_value=None, missing_values=-1, strategy='mean',
˓→verbose=0)
>>> X_test = sp.csc_matrix([[-1, 2], [6, -1], [7, 6]])

>>> print(imp.transform(X_test).toarray())
[[3. 2.]
[6. 3.]
[7. 6.]]
Note that this format is not meant to be used to implicitly store missing values in the matrix because it would densify
it at transform time. Missing values encoded by 0 must be used with dense input.
The SimpleImputer class also supports categorical data represented as string values or pandas categoricals when
using the 'most_frequent' or 'constant' strategy:
>>> import pandas as pd

>>> df = pd.DataFrame([["a", "x"],
... [np.nan, "y"],
... ["a", np.nan],
... ["b", "y"]], dtype="category")
...
>>> imp = SimpleImputer(strategy="most_frequent")
>>> print(imp.fit_transform(df))
[['a' 'x']
['a' 'y']

['a' 'y']
['b' 'y']]
SimpleImputer can be used in a Pipeline as a way to build a composite estimator that supports imputation. See
Imputing missing values before building an estimator.
Marking imputed values
The MissingIndicator transformer is useful to transform a dataset into corresponding binary matrix indicating
the presence of missing values in the dataset. This transformation is useful in conjunction with imputation. When
using imputation, preserving the information about which values had been missing can be informative.
NaN is usually used as the placeholder for missing values. However, it enforces the data type to be float. The parameter
missing_values allows to specify other placeholder such as integer. In the following example, we will use -1 as
missing values:
>>> from sklearn.impute import MissingIndicator

>>> X = np.array([[-1, -1, 1, 3],
... [4, -1, 0, -1],
... [8, -1, 1, 0]])
>>> indicator = MissingIndicator(missing_values=-1)
>>> mask_missing_values_only = indicator.fit_transform(X)
>>> mask_missing_values_only
array([[ True, True, False],
[False, True, True],
[False, True, False]])
The features parameter is used to choose the features for which the mask is constructed. By default, it is
'missing-only' which returns the imputer mask of the features containing missing values at fit time:
>>> indicator.features_
array([0, 1, 3])
The features parameter can be set to 'all' to returned all features whether or not they contain missing values:
>>> indicator = MissingIndicator(missing_values=-1, features="all")

>>> mask_all = indicator.fit_transform(X)
>>> mask_all
array([[ True, True, False, False],
[False, True, False, True],
[False, True, False, False]])
>>> indicator.features_
array([0, 1, 2, 3])
When using the MissingIndicator in a Pipeline, be sure to use the FeatureUnion or

ColumnTransformer to add the indicator features to the regular features. First we obtain the iris dataset, and
add some missing values to it.

>>> from sklearn.impute import SimpleImputer, MissingIndicator
>>> from sklearn.pipeline import FeatureUnion, make_pipeline
>>> mask = np.random.randint(0, 2, size=X.shape).astype(np.bool)
>>> X[mask] = np.nan

>>> X_train, X_test, y_train, _ = train_test_split(X, y, test_size=100,

... random_state=0)
Now we create a FeatureUnion. All features will be imputed using SimpleImputer, in order to enable classi-
fiers to work with this data. Additionally, it adds the the indicator variables from MissingIndicator.
>>> transformer = FeatureUnion(

... transformer_list=[
... ('features', SimpleImputer(strategy='mean')),
... ('indicators', MissingIndicator())])
>>> transformer = transformer.fit(X_train, y_train)
>>> results = transformer.transform(X_test)
>>> results.shape
(100, 8)
Of course, we cannot use the transformer to make any predictions. We should wrap this in a Pipeline with a
classifier (e.g., a DecisionTreeClassifier) to be able to make predictions.
>>> clf = make_pipeline(transformer, DecisionTreeClassifier())

>>> clf = clf.fit(X_train, y_train)
>>> results = clf.predict(X_test)
>>> results.shape
(100,)
3.4.5 Unsupervised dimensionality reduction
If your number of features is high, it may be useful to reduce it with an unsupervised step prior to supervised steps.
Many of the Unsupervised learning methods implement a transform method that can be used to reduce the dimen-
sionality. Below we discuss two specific example of this pattern that are heavily used.
Pipelining
The unsupervised data reduction and the supervised estimator can be chained in one step. See Pipeline: chaining
estimators.
PCA: principal component analysis
decomposition.PCA looks for a combination of features that capture well the variance of the original features.
See Decomposing signals in components (matrix factorization problems).
Examples
Random projections
The module: random_projection provides several tools for data reduction by random projections. See the
relevant section of the documentation: Random Projection.

Examples
• The Johnson-Lindenstrauss bound for embedding with random projections
Feature agglomeration
cluster.FeatureAgglomeration applies Hierarchical clustering to group together features that behave sim-
ilarly.
Examples

• Feature agglomeration
Feature scaling
Note that if features have very different scaling or statistical properties, cluster.FeatureAgglomeration
may not be able to capture the links between related features. Using a preprocessing.StandardScaler
can be useful in these settings.
3.4.6 Random Projection
The sklearn.random_projection module implements a simple and computationally efficient way to reduce
the dimensionality of the data by trading a controlled amount of accuracy (as additional variance) for faster processing
times and smaller model sizes. This module implements two types of unstructured random matrix: Gaussian random
matrix and sparse random matrix.
The dimensions and distribution of random projections matrices are controlled so as to preserve the pairwise distances
between any two samples of the dataset. Thus random projection is a suitable approximation technique for distance
based method.
References:
• Sanjoy Dasgupta. 2000. Experiments with random projection. In Proceedings of the Sixteenth conference
on Uncertainty in artificial intelligence (UAI‘00), Craig Boutilier and Moisés Goldszmidt (Eds.). Morgan
Kaufmann Publishers Inc., San Francisco, CA, USA, 143-151.
• Ella Bingham and Heikki Mannila. 2001. Random projection in dimensionality reduction: applications to
image and text data. In Proceedings of the seventh ACM SIGKDD international conference on Knowledge
discovery and data mining (KDD ‘01). ACM, New York, NY, USA, 245-250.
The Johnson-Lindenstrauss lemma
The main theoretical result behind the efficiency of random projection is the Johnson-Lindenstrauss lemma (quoting
Wikipedia):

In mathematics, the Johnson-Lindenstrauss lemma is a result concerning low-distortion embeddings of

points from high-dimensional into low-dimensional Euclidean space. The lemma states that a small set
of points in a high-dimensional space can be embedded into a space of much lower dimension in such a
way that distances between the points are nearly preserved. The map used for the embedding is at least
Lipschitz, and can even be taken to be an orthogonal projection.
Knowing only the number of samples, the sklearn.random_projection.
johnson_lindenstrauss_min_dim estimates conservatively the minimal size of the random subspace
to guarantee a bounded distortion introduced by the random projection:
>>> from sklearn.random_projection import johnson_lindenstrauss_min_dim

>>> johnson_lindenstrauss_min_dim(n_samples=1e6, eps=0.5)
663
>>> johnson_lindenstrauss_min_dim(n_samples=1e6, eps=[0.5, 0.1, 0.01])
array([ 663, 11841, 1112658])
>>> johnson_lindenstrauss_min_dim(n_samples=[1e4, 1e5, 1e6], eps=0.1)
array([ 7894, 9868, 11841])
Example:
• See The Johnson-Lindenstrauss bound for embedding with random projections for a theoretical explication
on the Johnson-Lindenstrauss lemma and an empirical validation using sparse random matrices.
References:
• Sanjoy Dasgupta and Anupam Gupta, 1999. An elementary proof of the Johnson-Lindenstrauss Lemma.

Gaussian random projection
The sklearn.random_projection.GaussianRandomProjection reduces the dimensionality by pro-

jecting the original input space on a randomly generated matrix where components are drawn from the following
1
distribution 𝑁 (0, 𝑛𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠 ).
Here a small excerpt which illustrates how to use the Gaussian random projection transformer:
>>> X = np.random.rand(100, 10000)
>>> transformer = random_projection.GaussianRandomProjection()
>>> X_new.shape
(100, 3947)
Sparse random projection
The sklearn.random_projection.SparseRandomProjection reduces the dimensionality by projecting

the original input space using a sparse random matrix.
Sparse random matrices are an alternative to dense Gaussian random projection matrix that guarantees similar embed-
ding quality while being much more memory efficient and allowing faster computation of the projected data.
If we define s = 1 / density, the elements of the random matrix are drawn from
⎧ √︁
𝑠
⎪ − 𝑛components
⎪
⎨ 1/2𝑠
√︁ 0 with probability 1 − 1/𝑠

𝑠
⎪
⎩ +
⎪ 1/2𝑠
𝑛components
where 𝑛components is the size of the projected subspace. By default the density of non zero elements is set to the
√
minimum density as recommended by Ping Li et al.: 1/ 𝑛features .

Here a small excerpt which illustrates how to use the sparse random projection transformer:

>>> X = np.random.rand(100, 10000)
>>> transformer = random_projection.SparseRandomProjection()
>>> X_new.shape
(100, 3947)
References:
• D. Achlioptas. 2003. Database-friendly random projections: Johnson-Lindenstrauss with binary coins. Jour-
nal of Computer and System Sciences 66 (2003) 671–687
• Ping Li, Trevor J. Hastie, and Kenneth W. Church. 2006. Very sparse random projections. In Proceedings
of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining (KDD ‘06).
ACM, New York, NY, USA, 287-296.
3.4.7 Kernel Approximation
This submodule contains functions that approximate the feature mappings that correspond to certain kernels, as they
are used for example in support vector machines (see Support Vector Machines). The following feature functions
perform non-linear transformations of the input, which can serve as a basis for linear classification or other algorithms.
The advantage of using approximate explicit feature maps compared to the kernel trick, which makes use of feature
maps implicitly, is that explicit mappings can be better suited for online learning and can significantly reduce the
cost of learning with very large datasets. Standard kernelized SVMs do not scale well to large datasets, but using an
approximate kernel map it is possible to use much more efficient linear SVMs. In particular, the combination of kernel
map approximations with SGDClassifier can make non-linear learning on large datasets possible.
Since there has not been much empirical work using approximate embeddings, it is advisable to compare results
against exact kernel methods when possible.
See also:
Polynomial regression: extending linear models with basis functions for an exact polynomial transformation.
Nystroem Method for Kernel Approximation
The Nystroem method, as implemented in Nystroem is a general method for low-rank approximations of kernels.
It achieves this by essentially subsampling the data on which the kernel is evaluated. By default Nystroem uses the
rbf kernel, but it can use any kernel function or a precomputed kernel matrix. The number of samples used - which
is also the dimensionality of the features computed - is given by the parameter n_components.
Radial Basis Function Kernel
The RBFSampler constructs an approximate mapping for the radial basis function kernel, also known as Random
Kitchen Sinks [RR2007]. This transformation can be used to explicitly model a kernel map, prior to applying a linear
algorithm, for example a linear SVM:

>>> from sklearn.kernel_approximation import RBFSampler

>>> X = [[0, 0], [1, 1], [1, 0], [0, 1]]
>>> y = [0, 0, 1, 1]
>>> rbf_feature = RBFSampler(gamma=1, random_state=1)
>>> X_features = rbf_feature.fit_transform(X)
>>> clf = SGDClassifier(max_iter=5)
>>> clf.fit(X_features, y)
>>> clf.score(X_features, y)
1.0
The mapping relies on a Monte Carlo approximation to the kernel values. The fit function performs the Monte Carlo
sampling, whereas the transform method performs the mapping of the data. Because of the inherent randomness
of the process, results may vary between different calls to the fit function.
The fit function takes two arguments: n_components, which is the target dimensionality of the feature transform,
and gamma, the parameter of the RBF-kernel. A higher n_components will result in a better approximation of the
kernel and will yield results more similar to those produced by a kernel SVM. Note that “fitting” the feature function
does not actually depend on the data given to the fit function. Only the dimensionality of the data is used. Details
on the method can be found in [RR2007].
For a given value of n_components RBFSampler is often less accurate as Nystroem. RBFSampler is cheaper
to compute, though, making use of larger feature spaces more efficient.
Fig. 3.9: Comparing an exact RBF kernel (left) with the approximation (right)
Examples:

Additive Chi Squared Kernel
The additive chi squared kernel is a kernel on histograms, often used in computer vision.
The additive chi squared kernel as used here is given by
∑︁ 2𝑥𝑖 𝑦𝑖
𝑘(𝑥, 𝑦) =
𝑖
𝑥𝑖 + 𝑦𝑖
This is not exactly the same as sklearn.metrics.additive_chi2_kernel. The authors of [VZ2010] prefer
the version above as it is always positive definite. Since the kernel is additive, it is possible to treat all components
𝑥𝑖 separately for embedding. This makes it possible to sample the Fourier transform in regular intervals, instead of
approximating using Monte Carlo sampling.
The class AdditiveChi2Sampler implements this component wise deterministic sampling. Each component
is sampled 𝑛 times, yielding 2𝑛 + 1 dimensions per input dimension (the multiple of two stems from the real and
complex part of the Fourier transform). In the literature, 𝑛 is usually chosen to be 1 or 2, transforming the dataset to
size n_samples * 5 * n_features (in the case of 𝑛 = 2).
The approximate feature map provided by AdditiveChi2Sampler can be combined with the approximate feature
map provided by RBFSampler to yield an approximate feature map for the exponentiated chi squared kernel. See
the [VZ2010] for details and [VVZ2010] for combination with the RBFSampler.
Skewed Chi Squared Kernel
The skewed chi squared kernel is given by:

∏︁ 2√𝑥𝑖 + 𝑐√𝑦𝑖 + 𝑐
𝑘(𝑥, 𝑦) =
𝑖
𝑥𝑖 + 𝑦𝑖 + 2𝑐
It has properties that are similar to the exponentiated chi squared kernel often used in computer vision, but allows for
a simple Monte Carlo approximation of the feature map.
The usage of the SkewedChi2Sampler is the same as the usage described above for the RBFSampler. The only
difference is in the free parameter, that is called 𝑐. For a motivation for this mapping and the mathematical details see
[LS2010].
Mathematical Details
Kernel methods like support vector machines or kernelized PCA rely on a property of reproducing kernel Hilbert
spaces. For any positive definite kernel function 𝑘 (a so called Mercer kernel), it is guaranteed that there exists a
mapping 𝜑 into a Hilbert space ℋ, such that
𝑘(𝑥, 𝑦) = ⟨𝜑(𝑥), 𝜑(𝑦)⟩
Where ⟨·, ·⟩ denotes the inner product in the Hilbert space.

If an algorithm, such as a linear support vector machine or PCA, relies only on the scalar product of data points 𝑥𝑖 ,
one may use the value of 𝑘(𝑥𝑖 , 𝑥𝑗 ), which corresponds to applying the algorithm to the mapped data points 𝜑(𝑥𝑖 ). The
advantage of using 𝑘 is that the mapping 𝜑 never has to be calculated explicitly, allowing for arbitrary large features
(even infinite).
One drawback of kernel methods is, that it might be necessary to store many kernel values 𝑘(𝑥𝑖 , 𝑥𝑗 ) during optimiza-
tion. If a kernelized classifier is applied to new data 𝑦𝑗 , 𝑘(𝑥𝑖 , 𝑦𝑗 ) needs to be computed to make predictions, possibly
for many different 𝑥𝑖 in the training set.
The classes in this submodule allow to approximate the embedding 𝜑, thereby working explicitly with the representa-
tions 𝜑(𝑥𝑖 ), which obviates the need to apply the kernel or store training examples.

References:
3.4.8 Pairwise metrics, Affinities and Kernels
The sklearn.metrics.pairwise submodule implements utilities to evaluate pairwise distances or affinity of

sets of samples.
This module contains both distance metrics and kernels. A brief summary is given on the two here.
Distance metrics are functions d(a, b) such that d(a, b) < d(a, c) if objects a and b are considered “more
similar” than objects a and c. Two objects exactly alike would have a distance of zero. One of the most popular
examples is Euclidean distance. To be a ‘true’ metric, it must obey the following four conditions:
1. d(a, b) >= 0, for all a and b

2. d(a, b) == 0, if and only if a = b, positive definiteness
3. d(a, b) == d(b, a), symmetry
4. d(a, c) <= d(a, b) + d(b, c), the triangle inequality
Kernels are measures of similarity, i.e. s(a, b) > s(a, c) if objects a and b are considered “more similar” than
objects a and c. A kernel must also be positive semi-definite.
There are a number of ways to convert between a distance metric and a similarity measure, such as a kernel. Let D be
the distance, and S be the kernel:
1. S = np.exp(-D * gamma), where one heuristic for choosing gamma is 1 / num_features
2. S = 1. / (D / np.max(D))
The distances between the row vectors of X and the row vectors of Y can be evaluated using pairwise_distances.
If Y is omitted the pairwise distances of the row vectors of X are calculated. Similarly, pairwise.
pairwise_kernels can be used to calculate the kernel between X and Y using different kernel functions. See
the API reference for more details.

>>> from sklearn.metrics.pairwise import pairwise_kernels
>>> X = np.array([[2, 3], [3, 5], [5, 8]])
>>> Y = np.array([[1, 0], [2, 1]])
>>> pairwise_distances(X, Y, metric='manhattan')
array([[ 4., 2.],
[ 7., 5.],
[12., 10.]])
>>> pairwise_distances(X, metric='manhattan')
array([[0., 3., 8.],
[3., 0., 5.],
[8., 5., 0.]])
>>> pairwise_kernels(X, Y, metric='linear')
array([[ 2., 7.],
[ 3., 11.],
[ 5., 18.]])
Cosine similarity
cosine_similarity computes the L2-normalized dot product of vectors. That is, if 𝑥 and 𝑦 are row vectors, their

cosine similarity 𝑘 is defined as:
𝑥𝑦 ⊤
𝑘(𝑥, 𝑦) =
‖𝑥‖‖𝑦‖
This is called cosine similarity, because Euclidean (L2) normalization projects the vectors onto the unit sphere, and
their dot product is then the cosine of the angle between the points denoted by the vectors.
This kernel is a popular choice for computing the similarity of documents represented as tf-idf vectors.
cosine_similarity accepts scipy.sparse matrices. (Note that the tf-idf functionality in sklearn.
feature_extraction.text can produce normalized vectors, in which case cosine_similarity is equiv-
alent to linear_kernel, only slower.)
References:
• C.D. Manning, P. Raghavan and H. Schütze (2008). Introduction to Information Retrieval. Cambridge Uni-
versity Press. http://nlp.stanford.edu/IR-book/html/htmledition/the-vector-space-model-for-scoring-1.html
Linear kernel
The function linear_kernel computes the linear kernel, that is, a special case of polynomial_kernel with
degree=1 and coef0=0 (homogeneous). If x and y are column vectors, their linear kernel is:
𝑘(𝑥, 𝑦) = 𝑥⊤ 𝑦
Polynomial kernel
The function polynomial_kernel computes the degree-d polynomial kernel between two vectors. The polyno-
mial kernel represents the similarity between two vectors. Conceptually, the polynomial kernels considers not only
the similarity between vectors under the same dimension, but also across dimensions. When used in machine learning
algorithms, this allows to account for feature interaction.
The polynomial kernel is defined as:
𝑘(𝑥, 𝑦) = (𝛾𝑥⊤ 𝑦 + 𝑐0 )𝑑
where:
• x, y are the input vectors
• d is the kernel degree
If 𝑐0 = 0 the kernel is said to be homogeneous.
Sigmoid kernel
The function sigmoid_kernel computes the sigmoid kernel between two vectors. The sigmoid kernel is also
known as hyperbolic tangent, or Multilayer Perceptron (because, in the neural network field, it is often used as neuron
activation function). It is defined as:
𝑘(𝑥, 𝑦) = tanh(𝛾𝑥⊤ 𝑦 + 𝑐0 )
where:

• x, y are the input vectors

• 𝛾 is known as slope
• 𝑐0 is known as intercept
RBF kernel
The function rbf_kernel computes the radial basis function (RBF) kernel between two vectors. This kernel is
defined as:
𝑘(𝑥, 𝑦) = exp(−𝛾‖𝑥 − 𝑦‖2 )
where x and y are the input vectors. If 𝛾 = 𝜎 −2 the kernel is known as the Gaussian kernel of variance 𝜎 2 .
Laplacian kernel
The function laplacian_kernel is a variant on the radial basis function kernel defined as:
𝑘(𝑥, 𝑦) = exp(−𝛾‖𝑥 − 𝑦‖1 )
where x and y are the input vectors and ‖𝑥 − 𝑦‖1 is the Manhattan distance between the input vectors.
It has proven useful in ML applied to noiseless data. See e.g. Machine learning for quantum mechanics in a nutshell.
Chi-squared kernel
The chi-squared kernel is a very popular choice for training non-linear SVMs in computer vision applications. It can
be computed using chi2_kernel and then passed to an sklearn.svm.SVC with kernel="precomputed":

>>> from sklearn.metrics.pairwise import chi2_kernel
>>> X = [[0, 1], [1, 0], [.2, .8], [.7, .3]]
>>> y = [0, 1, 0, 1]
>>> K = chi2_kernel(X, gamma=.5)
>>> K
array([[1. , 0.36787944, 0.89483932, 0.58364548],
[0.36787944, 1. , 0.51341712, 0.83822343],
[0.89483932, 0.51341712, 1. , 0.7768366 ],
[0.58364548, 0.83822343, 0.7768366 , 1. ]])
>>> svm = SVC(kernel='precomputed').fit(K, y)

>>> svm.predict(K)
array([0, 1, 0, 1])
It can also be directly used as the kernel argument:
>>> svm = SVC(kernel=chi2_kernel).fit(X, y)

>>> svm.predict(X)
array([0, 1, 0, 1])
The chi squared kernel is given by

(︃ )︃
∑︁ (𝑥[𝑖] − 𝑦[𝑖])2
𝑘(𝑥, 𝑦) = exp −𝛾
𝑖
𝑥[𝑖] + 𝑦[𝑖]

The data is assumed to be non-negative, and is often normalized to have an L1-norm of one. The normalization is
rationalized with the connection to the chi squared distance, which is a distance between discrete probability distribu-
tions.
The chi squared kernel is most commonly used on histograms (bags) of visual words.
References:
• Zhang, J. and Marszalek, M. and Lazebnik, S. and Schmid, C. Local features and kernels for classification
of texture and object categories: A comprehensive study International Journal of Computer Vision 2007
https://research.microsoft.com/en-us/um/people/manik/projects/trade-off/papers/ZhangIJCV06.pdf
3.4.9 Transforming the prediction target (y)
These are transformers that are not intended to be used on features, only on supervised learning targets. See also
Transforming target in regression if you want to transform the prediction target for learning, but evaluate the model in
the original (untransformed) space.
Label binarization
LabelBinarizer is a utility class to help create a label indicator matrix from a list of multi-class labels:

>>> lb = preprocessing.LabelBinarizer()
>>> lb.fit([1, 2, 6, 4, 2])
LabelBinarizer(neg_label=0, pos_label=1, sparse_output=False)
>>> lb.classes_
array([1, 2, 4, 6])
>>> lb.transform([1, 6])
array([[1, 0, 0, 0],
[0, 0, 0, 1]])
For multiple labels per instance, use MultiLabelBinarizer:
>>> lb = preprocessing.MultiLabelBinarizer()
>>> lb.fit_transform([(1, 2), (3,)])
array([[1, 1, 0],
[0, 0, 1]])
>>> lb.classes_
array([1, 2, 3])
Label encoding
LabelEncoder is a utility class to help normalize labels such that they contain only values between 0 and n_classes-
1. This is sometimes useful for writing efficient Cython routines. LabelEncoder can be used as follows:

>>> le = preprocessing.LabelEncoder()
>>> le.fit([1, 2, 2, 6])
LabelEncoder()
>>> le.classes_
array([1, 2, 6])

>>> le.transform([1, 1, 2, 6])

array([0, 0, 1, 2])
>>> le.inverse_transform([0, 0, 1, 2])
array([1, 1, 2, 6])
It can also be used to transform non-numerical labels (as long as they are hashable and comparable) to numerical
labels:
>>> le.fit(["paris", "paris", "tokyo", "amsterdam"])
LabelEncoder()
>>> list(le.classes_)
['amsterdam', 'paris', 'tokyo']
>>> le.transform(["tokyo", "tokyo", "paris"])
array([2, 2, 1])
>>> list(le.inverse_transform([2, 2, 1]))
['tokyo', 'tokyo', 'paris']
3.5 Dataset loading utilities
The sklearn.datasets package embeds some small toy datasets as introduced in the Getting Started section.
This package also features helpers to fetch larger datasets commonly used by the machine learning community to
benchmark algorithms on data that comes from the ‘real world’.
To evaluate the impact of the scale of the dataset (n_samples and n_features) while controlling the statistical
properties of the data (typically the correlation and informativeness of the features), it is also possible to generate
synthetic data.
3.5.1 General dataset API
There are three main kinds of dataset interfaces that can be used to get datasets depending on the desired type of
dataset.
The dataset loaders. They can be used to load small standard datasets, described in the Toy datasets section.
The dataset fetchers. They can be used to download and load larger datasets, described in the Real world datasets
section.
Both loaders and fetchers functions return a dictionary-like object holding at least two items: an array of shape
n_samples * n_features with key data (except for 20newsgroups) and a numpy array of length n_samples,
containing the target values, with key target.
It’s also possible for almost all of these function to constrain the output to be a tuple containing only the data and the
target, by setting the return_X_y parameter to True.
The datasets also contain a full description in their DESCR attribute and some contain feature_names and
target_names. See the dataset descriptions below for details.
The dataset generation functions. They can be used to generate controlled synthetic datasets, described in the
Generated datasets section.
These functions return a tuple (X, y) consisting of a n_samples * n_features numpy array X and an array of
length n_samples containing the targets y.
In addition, there are also miscellanous tools to load datasets of other formats or from other locations, described in the
Loading other datasets section.
3.5. Dataset loading utilities 583

3.5.2 Toy datasets
scikit-learn comes with a few small standard datasets that do not require to download any file from some external
website.
They can be loaded using the following functions:
load_boston([return_X_y]) Load and return the boston house-prices dataset (regres-

sion).
load_iris([return_X_y]) Load and return the iris dataset (classification).
load_diabetes([return_X_y]) Load and return the diabetes dataset (regression).
load_digits([n_class, return_X_y]) Load and return the digits dataset (classification).
load_linnerud([return_X_y]) Load and return the linnerud dataset (multivariate regres-
sion).
load_wine([return_X_y]) Load and return the wine dataset (classification).
load_breast_cancer([return_X_y]) Load and return the breast cancer wisconsin dataset (clas-
sification).
These datasets are useful to quickly illustrate the behavior of the various algorithms implemented in scikit-learn. They
are however often too small to be representative of real world machine learning tasks.
Boston house prices dataset
Data Set Characteristics:

Number of Instances 506
Number of Attributes 13 numeric/categorical predictive. Median Value (attribute 14) is usu-
ally the target.
Attribute Information (in order)
• CRIM per capita crime rate by town
• ZN proportion of residential land zoned for lots over 25,000 sq.ft.
• INDUS proportion of non-retail business acres per town
• CHAS Charles River dummy variable (= 1 if tract bounds river; 0 otherwise)
• NOX nitric oxides concentration (parts per 10 million)
• RM average number of rooms per dwelling
• AGE proportion of owner-occupied units built prior to 1940
• DIS weighted distances to five Boston employment centres
• RAD index of accessibility to radial highways
• TAX full-value property-tax rate per $10,000
• PTRATIO pupil-teacher ratio by town
• B 1000(Bk - 0.63)^2 where Bk is the proportion of blacks by town
• LSTAT % lower status of the population
• MEDV Median value of owner-occupied homes in $1000’s
Missing Attribute Values None

Creator Harrison, D. and Rubinfeld, D.L.

This is a copy of UCI ML housing dataset. https://archive.ics.uci.edu/ml/machine-learning-databases/housing/
This dataset was taken from the StatLib library which is maintained at Carnegie Mellon University.
The Boston house-price data of Harrison, D. and Rubinfeld, D.L. ‘Hedonic prices and the demand for clean air’, J.
Environ. Economics & Management, vol.5, 81-102, 1978. Used in Belsley, Kuh & Welsch, ‘Regression diagnostics
. . . ’, Wiley, 1980. N.B. Various transformations are used in the table on pages 244-261 of the latter.
The Boston house-price data has been used in many machine learning papers that address regression problems.
References
• Belsley, Kuh & Welsch, ‘Regression diagnostics: Identifying Influential Data and Sources of Collinearity’,
Wiley, 1980. 244-261.
• Quinlan,R. (1993). Combining Instance-Based and Model-Based Learning. In Proceedings on the Tenth
International Conference of Machine Learning, 236-243, University of Massachusetts, Amherst. Morgan
Kaufmann.
Iris plants dataset

Number of Instances 150 (50 in each of three classes)
Number of Attributes 4 numeric, predictive attributes and the class
Attribute Information
• sepal length in cm
• sepal width in cm
• petal length in cm
• petal width in cm
• class:
– Iris-Setosa
– Iris-Versicolour
– Iris-Virginica
Summary Statistics
sepal length: 4.3 7.9 5.84 0.83 0.7826

sepal width: 2.0 4.4 3.05 0.43 -0.4194
petal length: 1.0 6.9 3.76 1.76 0.9490 (high!)
petal width: 0.1 2.5 1.20 0.76 0.9565 (high!)

Class Distribution 33.3% for each of 3 classes.
Creator R.A. Fisher
Donor Michael Marshall (MARSHALL%PLU@io.arc.nasa.gov)

Date July, 1988

The famous Iris database, first used by Sir R.A. Fisher. The dataset is taken from Fisher’s paper. Note that it’s the
same as in R, but not as in the UCI Machine Learning Repository, which has two wrong data points.
This is perhaps the best known database to be found in the pattern recognition literature. Fisher’s paper is a classic in
the field and is referenced frequently to this day. (See Duda & Hart, for example.) The data set contains 3 classes of
50 instances each, where each class refers to a type of iris plant. One class is linearly separable from the other 2; the
latter are NOT linearly separable from each other.
References
• Fisher, R.A. “The use of multiple measurements in taxonomic problems” Annual Eugenics, 7, Part II, 179-188
(1936); also in “Contributions to Mathematical Statistics” (John Wiley, NY, 1950).
• Duda, R.O., & Hart, P.E. (1973) Pattern Classification and Scene Analysis. (Q327.D83) John Wiley & Sons.
ISBN 0-471-22361-1. See page 218.
• Dasarathy, B.V. (1980) “Nosing Around the Neighborhood: A New System Structure and Classification Rule
for Recognition in Partially Exposed Environments”. IEEE Transactions on Pattern Analysis and Machine
Intelligence, Vol. PAMI-2, No. 1, 67-71.
• Gates, G.W. (1972) “The Reduced Nearest Neighbor Rule”. IEEE Transactions on Information Theory, May
1972, 431-433.
• See also: 1988 MLC Proceedings, 54-64. Cheeseman et al”s AUTOCLASS II conceptual clustering system
finds 3 classes in the data.
• Many, many more . . .
Diabetes dataset
Ten baseline variables, age, sex, body mass index, average blood pressure, and six blood serum measurements were
obtained for each of n = 442 diabetes patients, as well as the response of interest, a quantitative measure of disease
progression one year after baseline.
Number of Attributes First 10 columns are numeric predictive values
Target Column 11 is a quantitative measure of disease progression one year after baseline
• Age
• Sex
• Body mass index
• Average blood pressure
• S1
• S2
• S3
• S4

• S5
• S6
Note: Each of these 10 feature variables have been mean centered and scaled by the standard deviation times n_samples
(i.e. the sum of squares of each column totals 1).
Source URL: http://www4.stat.ncsu.edu/~boos/var.select/diabetes.html
For more information see: Bradley Efron, Trevor Hastie, Iain Johnstone and Robert Tibshirani (2004) “Least An-
gle Regression,” Annals of Statistics (with discussion), 407-499. (http://web.stanford.edu/~hastie/Papers/LARS/
LeastAngle_2002.pdf)
Optical recognition of handwritten digits dataset

Number of Attributes 64
Attribute Information 8x8 image of integer pixels in the range 0..16.
Creator
5. Alpaydin (alpaydin ‘@’ boun.edu.tr)
Date July; 1998
This is a copy of the test set of the UCI ML hand-written digits datasets http://archive.ics.uci.edu/ml/datasets/Optical+
Recognition+of+Handwritten+Digits
The data set contains images of hand-written digits: 10 classes where each class refers to a digit.
Preprocessing programs made available by NIST were used to extract normalized bitmaps of handwritten digits from
a preprinted form. From a total of 43 people, 30 contributed to the training set and different 13 to the test set. 32x32
bitmaps are divided into nonoverlapping blocks of 4x4 and the number of on pixels are counted in each block. This
generates an input matrix of 8x8 where each element is an integer in the range 0..16. This reduces dimensionality and
gives invariance to small distortions.
For info on NIST preprocessing routines, see M. D. Garris, J. L. Blue, G. T. Candela, D. L. Dimmick, J. Geist, P. J.
Grother, S. A. Janet, and C. L. Wilson, NIST Form-Based Handprint Recognition System, NISTIR 5469, 1994.
References
• C. Kaynak (1995) Methods of Combining Multiple Classifiers and Their Applications to Handwritten Digit
Recognition, MSc Thesis, Institute of Graduate Studies in Science and Engineering, Bogazici University.
• 5. Alpaydin, C. Kaynak (1998) Cascading Classifiers, Kybernetika.
• Ken Tang and Ponnuthurai N. Suganthan and Xi Yao and A. Kai Qin. Linear dimensionalityreduction using
relevance weighted LDA. School of Electrical and Electronic Engineering Nanyang Technological University.
2005.
• Claudio Gentile. A New Approximate Maximal Margin Classification Algorithm. NIPS. 2000.

Linnerrud dataset

Number of Attributes 3
The Linnerud dataset constains two small dataset:
• physiological - CSV containing 20 observations on 3 exercise variables: Weight, Waist and Pulse.
• exercise - CSV containing 20 observations on 3 physiological variables: Chins, Situps and Jumps.
References
• Tenenhaus, M. (1998). La regression PLS: theorie et pratique. Paris: Editions Technic.
Wine recognition dataset

Number of Instances 178 (50 in each of three classes)
• Alcohol
• Malic acid
• Ash
• Alcalinity of ash
• Magnesium
• Total phenols
• Flavanoids
• Nonflavanoid phenols
• Proanthocyanins
• Color intensity
• Hue
• OD280/OD315 of diluted wines
• Proline
• class:
– class_0
– class_1
– class_2
Summary Statistics

Alcohol: 11.0 14.8 13.0 0.8

Malic Acid: 0.74 5.80 2.34 1.12
Ash: 1.36 3.23 2.36 0.27
Alcalinity of Ash: 10.6 30.0 19.5 3.3
Magnesium: 70.0 162.0 99.7 14.3
Total Phenols: 0.98 3.88 2.29 0.63
Flavanoids: 0.34 5.08 2.03 1.00
Nonflavanoid Phenols: 0.13 0.66 0.36 0.12
Proanthocyanins: 0.41 3.58 1.59 0.57
Colour Intensity: 1.3 13.0 5.1 2.3
Hue: 0.48 1.71 0.96 0.23
OD280/OD315 of diluted wines: 1.27 4.00 2.61 0.71
Proline: 278 1680 746 315

Class Distribution class_0 (59), class_1 (71), class_2 (48)
Creator R.A. Fisher
Donor Michael Marshall (MARSHALL%PLU@io.arc.nasa.gov)
Date July, 1988
This is a copy of UCI ML Wine recognition datasets. https://archive.ics.uci.edu/ml/machine-learning-databases/wine/
wine.data
The data is the results of a chemical analysis of wines grown in the same region in Italy by three different cultivators.
There are thirteen different measurements taken for different constituents found in the three types of wine.
Original Owners:
Forina, M. et al, PARVUS - An Extendible Package for Data Exploration, Classification and Correlation. Institute of
Pharmaceutical and Food Analysis and Technologies, Via Brigata Salerno, 16147 Genoa, Italy.
Citation:
Lichman, M. (2013). UCI Machine Learning Repository [http://archive.ics.uci.edu/ml]. Irvine, CA: University of
California, School of Information and Computer Science.
References
(1) S. Aeberhard, D. Coomans and O. de Vel, Comparison of Classifiers in High Dimensional Settings, Tech. Rep.
no. 92-02, (1992), Dept. of Computer Science and Dept. of Mathematics and Statistics, James Cook University of
North Queensland. (Also submitted to Technometrics).
The data was used with many others for comparing various classifiers. The classes are separable, though only RDA
has achieved 100% correct classification. (RDA : 100%, QDA 99.4%, LDA 98.9%, 1NN 96.1% (z-transformed
data)) (All results using the leave-one-out technique)
(2) S. Aeberhard, D. Coomans and O. de Vel, “THE CLASSIFICATION PERFORMANCE OF RDA” Tech. Rep.
no. 92-01, (1992), Dept. of Computer Science and Dept. of Mathematics and Statistics, James Cook University of
North Queensland. (Also submitted to Journal of Chemometrics).

Breast cancer wisconsin (diagnostic) dataset

• radius (mean of distances from center to points on the perimeter)
• texture (standard deviation of gray-scale values)
• perimeter
• area
• smoothness (local variation in radius lengths)
• compactness (perimeter^2 / area - 1.0)
• concavity (severity of concave portions of the contour)
• concave points (number of concave portions of the contour)
• symmetry
• fractal dimension (“coastline approximation” - 1)
The mean, standard error, and “worst” or largest (mean of the three largest values) of these
features were computed for each image, resulting in 30 features. For instance, field 3 is
Mean Radius, field 13 is Radius SE, field 23 is Worst Radius.
• class:
– WDBC-Malignant
– WDBC-Benign
Summary Statistics
radius (mean): 6.981 28.11

texture (mean): 9.71 39.28
perimeter (mean): 43.79 188.5
area (mean): 143.5 2501.0
smoothness (mean): 0.053 0.163
compactness (mean): 0.019 0.345
concavity (mean): 0.0 0.427
concave points (mean): 0.0 0.201
symmetry (mean): 0.106 0.304
fractal dimension (mean): 0.05 0.097
radius (standard error): 0.112 2.873
texture (standard error): 0.36 4.885
perimeter (standard error): 0.757 21.98
area (standard error): 6.802 542.2
smoothness (standard error): 0.002 0.031
compactness (standard error): 0.002 0.135
concavity (standard error): 0.0 0.396
concave points (standard error): 0.0 0.053

symmetry (standard error): 0.008 0.079

fractal dimension (standard error): 0.001 0.03
radius (worst): 7.93 36.04
texture (worst): 12.02 49.54
perimeter (worst): 50.41 251.2
area (worst): 185.2 4254.0
smoothness (worst): 0.071 0.223
compactness (worst): 0.027 1.058
concavity (worst): 0.0 1.252
concave points (worst): 0.0 0.291
symmetry (worst): 0.156 0.664
fractal dimension (worst): 0.055 0.208

Class Distribution 212 - Malignant, 357 - Benign
Creator Dr. William H. Wolberg, W. Nick Street, Olvi L. Mangasarian
Donor Nick Street
Date November, 1995
This is a copy of UCI ML Breast Cancer Wisconsin (Diagnostic) datasets. https://goo.gl/U2Uwz2
Features are computed from a digitized image of a fine needle aspirate (FNA) of a breast mass. They describe charac-
teristics of the cell nuclei present in the image.
Separating plane described above was obtained using Multisurface Method-Tree (MSM-T) [K. P. Bennett, “Decision
Tree Construction Via Linear Programming.” Proceedings of the 4th Midwest Artificial Intelligence and Cognitive
Science Society, pp. 97-101, 1992], a classification method which uses linear programming to construct a decision
tree. Relevant features were selected using an exhaustive search in the space of 1-4 features and 1-3 separating planes.
The actual linear program used to obtain the separating plane in the 3-dimensional space is that described in: [K.
P. Bennett and O. L. Mangasarian: “Robust Linear Programming Discrimination of Two Linearly Inseparable Sets”,
Optimization Methods and Software 1, 1992, 23-34].
This database is also available through the UW CS ftp server:
ftp ftp.cs.wisc.edu cd math-prog/cpo-dataset/machine-learn/WDBC/
References
• W.N. Street, W.H. Wolberg and O.L. Mangasarian. Nuclear feature extraction for breast tumor diagnosis.
IS&T/SPIE 1993 International Symposium on Electronic Imaging: Science and Technology, volume 1905,
pages 861-870, San Jose, CA, 1993.
• O.L. Mangasarian, W.N. Street and W.H. Wolberg. Breast cancer diagnosis and prognosis via linear program-
ming. Operations Research, 43(4), pages 570-577, July-August 1995.
• W.H. Wolberg, W.N. Street, and O.L. Mangasarian. Machine learning techniques to diagnose breast cancer
from fine-needle aspirates. Cancer Letters 77 (1994) 163-171.
3.5.3 Real world datasets
scikit-learn provides tools to load larger datasets, downloading them if necessary.

They can be loaded using the following functions:
fetch_olivetti_faces([data_home, shuffle, . . . ]) Load the Olivetti faces data-set from AT&T (classification).
fetch_20newsgroups([data_home, subset, . . . ]) Load the filenames and data from the 20 newsgroups
dataset (classification).
fetch_20newsgroups_vectorized([subset, . . . ]) Load the 20 newsgroups dataset and vectorize it into token
counts (classification).
fetch_lfw_people([data_home, funneled, . . . ]) Load the Labeled Faces in the Wild (LFW) people dataset
(classification).
fetch_lfw_pairs([subset, data_home, . . . ]) Load the Labeled Faces in the Wild (LFW) pairs dataset
(classification).
fetch_covtype([data_home, . . . ]) Load the covertype dataset (classification).
fetch_rcv1([data_home, subset, . . . ]) Load the RCV1 multilabel dataset (classification).
fetch_kddcup99([subset, data_home, shuffle, . . . ]) Load the kddcup99 dataset (classification).
fetch_california_housing([data_home, . . . ]) Load the California housing dataset (regression).
The Olivetti faces dataset
This dataset contains a set of face images taken between April 1992 and April 1994 at AT&T Laboratories Cam-
bridge. The sklearn.datasets.fetch_olivetti_faces function is the data fetching / caching function
that downloads the data archive from AT&T.
As described on the original website:
There are ten different images of each of 40 distinct subjects. For some subjects, the images were taken
at different times, varying the lighting, facial expressions (open / closed eyes, smiling / not smiling) and
facial details (glasses / no glasses). All the images were taken against a dark homogeneous background
with the subjects in an upright, frontal position (with tolerance for some side movement).
Classes 40
Samples total 400
Dimensionality 4096
Features real, between 0 and 1
The image is quantized to 256 grey levels and stored as unsigned 8-bit integers; the loader will convert these to floating
point values on the interval [0, 1], which are easier to work with for many algorithms.
The “target” for this database is an integer from 0 to 39 indicating the identity of the person pictured; however, with
only 10 examples per class, this relatively small dataset is more interesting from an unsupervised or semi-supervised
perspective.
The original dataset consisted of 92 x 112, while the version available here consists of 64x64 images.
When using these images, please give credit to AT&T Laboratories Cambridge.
The 20 newsgroups text dataset
The 20 newsgroups dataset comprises around 18000 newsgroups posts on 20 topics split in two subsets: one for
training (or development) and the other one for testing (or for performance evaluation). The split between the train
and test set is based upon a messages posted before and after a specific date.

This module contains two loaders. The first one, sklearn.datasets.fetch_20newsgroups, returns a list
of the raw texts that can be fed to text feature extractors such as sklearn.feature_extraction.text.
CountVectorizer with custom parameters so as to extract feature vectors. The second one, sklearn.
datasets.fetch_20newsgroups_vectorized, returns ready-to-use features, i.e., it is not necessary to use
a feature extractor.
Classes 20
Samples total 18846
Dimensionality 1
Features text
Usage
The sklearn.datasets.fetch_20newsgroups function is a data fetching / caching functions that down-

loads the data archive from the original 20 newsgroups website, extracts the archive contents in the ~/
scikit_learn_data/20news_home folder and calls the sklearn.datasets.load_files on either the
training or testing set folder, or both of them:
>>> from sklearn.datasets import fetch_20newsgroups

>>> newsgroups_train = fetch_20newsgroups(subset='train')
>>> from pprint import pprint

>>> pprint(list(newsgroups_train.target_names))
['alt.atheism',
'comp.graphics',
'comp.os.ms-windows.misc',
'comp.sys.ibm.pc.hardware',
'comp.sys.mac.hardware',
'comp.windows.x',
'misc.forsale',
'rec.autos',
'rec.motorcycles',
'rec.sport.baseball',
'rec.sport.hockey',
'sci.crypt',
'sci.electronics',
'sci.med',
'sci.space',
'soc.religion.christian',
'talk.politics.guns',
'talk.politics.mideast',
'talk.politics.misc',
'talk.religion.misc']
The real data lies in the filenames and target attributes. The target attribute is the integer index of the category:
>>> newsgroups_train.filenames.shape
(11314,)
>>> newsgroups_train.target.shape
(11314,)
>>> newsgroups_train.target[:10]
array([ 7, 4, 4, 1, 14, 16, 13, 3, 2, 4])

It is possible to load only a sub-selection of the categories by passing the list of the categories to load to the sklearn.
datasets.fetch_20newsgroups function:
>>> cats = ['alt.atheism', 'sci.space']
>>> newsgroups_train = fetch_20newsgroups(subset='train', categories=cats)
>>> list(newsgroups_train.target_names)
['alt.atheism', 'sci.space']
>>> newsgroups_train.filenames.shape
(1073,)
>>> newsgroups_train.target.shape
(1073,)
>>> newsgroups_train.target[:10]
array([0, 1, 1, 1, 0, 1, 1, 0, 0, 0])
Converting text to vectors
In order to feed predictive or clustering models with the text data, one first need to turn the text into vectors
of numerical values suitable for statistical analysis. This can be achieved with the utilities of the sklearn.
feature_extraction.text as demonstrated in the following example that extract TF-IDF vectors of unigram
tokens from a subset of 20news:
>>> categories = ['alt.atheism', 'talk.religion.misc',
... 'comp.graphics', 'sci.space']
>>> newsgroups_train = fetch_20newsgroups(subset='train',
... categories=categories)
>>> vectors = vectorizer.fit_transform(newsgroups_train.data)
>>> vectors.shape
(2034, 34118)
The extracted TF-IDF vectors are very sparse, with an average of 159 non-zero components by sample in a more than
30000-dimensional space (less than .5% non-zero features):
>>> vectors.nnz / float(vectors.shape[0])
159.01327...
sklearn.datasets.fetch_20newsgroups_vectorized is a function which returns ready-to-use token

counts features instead of file names.
Filtering text for more realistic training
It is easy for a classifier to overfit on particular things that appear in the 20 Newsgroups data, such as newsgroup
headers. Many classifiers achieve very high F-scores, but their results would not generalize to other documents that
aren’t from this window of time.
For example, let’s look at the results of a multinomial Naive Bayes classifier, which is fast to train and achieves a
decent F-score:
>>> newsgroups_test = fetch_20newsgroups(subset='test',
>>> vectors_test = vectorizer.transform(newsgroups_test.data)

>>> clf = MultinomialNB(alpha=.01)

>>> clf.fit(vectors, newsgroups_train.target)
MultinomialNB(alpha=0.01, class_prior=None, fit_prior=True)
>>> pred = clf.predict(vectors_test)

>>> metrics.f1_score(newsgroups_test.target, pred, average='macro')
0.88213...
(The example Classification of text documents using sparse features shuffles the training and test data, instead of
segmenting by time, and in that case multinomial Naive Bayes gets a much higher F-score of 0.88. Are you suspicious
yet of what’s going on inside this classifier?)
Let’s take a look at what the most informative features are:
>>> def show_top10(classifier, vectorizer, categories):
... feature_names = np.asarray(vectorizer.get_feature_names())
... for i, category in enumerate(categories):
... top10 = np.argsort(classifier.coef_[i])[-10:]
... print("%s: %s" % (category, " ".join(feature_names[top10])))
...
>>> show_top10(clf, vectorizer, newsgroups_train.target_names)
alt.atheism: edu it and in you that is of to the
comp.graphics: edu in graphics it is for and of to the
sci.space: edu it that is in and space to of the
talk.religion.misc: not it you in is that and to of the
You can now see many things that these features have overfit to:
• Almost every group is distinguished by whether headers such as NNTP-Posting-Host: and
Distribution: appear more or less often.
• Another significant feature involves whether the sender is affiliated with a university, as indicated either by their
headers or their signature.
• The word “article” is a significant feature, based on how often people quote previous posts like this: “In article
[article ID], [name] <[e-mail address]> wrote:”
• Other features match the names and e-mail addresses of particular people who were posting at the time.
With such an abundance of clues that distinguish newsgroups, the classifiers barely have to identify topics from text at
all, and they all perform at the same high level.
For this reason, the functions that load 20 Newsgroups data provide a parameter called remove, telling it what
kinds of information to strip out of each file. remove should be a tuple containing any subset of ('headers',
'footers', 'quotes'), telling it to remove headers, signature blocks, and quotation blocks respectively.
>>> newsgroups_test = fetch_20newsgroups(subset='test',
... remove=('headers', 'footers', 'quotes'),
>>> metrics.f1_score(pred, newsgroups_test.target, average='macro')
0.77310...
This classifier lost over a lot of its F-score, just because we removed metadata that has little to do with topic classifi-
cation. It loses even more if we also strip this metadata from the training data:
>>> newsgroups_train = fetch_20newsgroups(subset='train',
... remove=('headers', 'footers', 'quotes'),

>>> vectors = vectorizer.fit_transform(newsgroups_train.data)
>>> clf = MultinomialNB(alpha=.01)
>>> clf.fit(vectors, newsgroups_train.target)

>>> metrics.f1_score(newsgroups_test.target, pred, average='macro')
0.76995...
Some other classifiers cope better with this harder version of the task. Try running Sample pipeline for text feature
extraction and evaluation with and without the --filter option to compare the results.
Recommendation
When evaluating text classifiers on the 20 Newsgroups data, you should strip newsgroup-related metadata. In
scikit-learn, you can do this by setting remove=('headers', 'footers', 'quotes'). The F-score will
be lower because it is more realistic.
Examples

The Labeled Faces in the Wild face recognition dataset
This dataset is a collection of JPEG pictures of famous people collected over the internet, all details are available on
the official website:
http://vis-www.cs.umass.edu/lfw/
Each picture is centered on a single face. The typical task is called Face Verification: given a pair of two pictures, a
binary classifier must predict whether the two images are from the same person.
An alternative task, Face Recognition or Face Identification is: given the picture of the face of an unknown person,
identify the name of the person by referring to a gallery of previously seen pictures of identified persons.
Both Face Verification and Face Recognition are tasks that are typically performed on the output of a model trained to
perform Face Detection. The most popular model for Face Detection is called Viola-Jones and is implemented in the
OpenCV library. The LFW faces were extracted by this face detector from various online websites.
Classes 5749
Samples total 13233
Dimensionality 5828

Usage
scikit-learn provides two loaders that will automatically download, cache, parse the metadata files, decode the
jpeg and convert the interesting slices into memmapped numpy arrays. This dataset size is more than 200 MB. The first
load typically takes more than a couple of minutes to fully decode the relevant part of the JPEG files into numpy arrays.
If the dataset has been loaded once, the following times the loading times less than 200ms by using a memmapped
version memoized on the disk in the ~/scikit_learn_data/lfw_home/ folder using joblib.
The first loader is used for the Face Identification task: a multi-class classification task (hence supervised learning):
>>> from sklearn.datasets import fetch_lfw_people
>>> lfw_people = fetch_lfw_people(min_faces_per_person=70, resize=0.4)
>>> for name in lfw_people.target_names:

... print(name)
...
Ariel Sharon
Colin Powell
Donald Rumsfeld
George W Bush
Gerhard Schroeder
Hugo Chavez
Tony Blair
The default slice is a rectangular shape around the face, removing most of the background:
>>> lfw_people.data.dtype
dtype('float32')
>>> lfw_people.data.shape
(1288, 1850)
>>> lfw_people.images.shape
(1288, 50, 37)
Each of the 1140 faces is assigned to a single person id in the target array:
>>> lfw_people.target.shape
(1288,)
>>> list(lfw_people.target[:10])
[5, 6, 3, 1, 0, 1, 3, 4, 3, 0]
The second loader is typically used for the face verification task: each sample is a pair of two picture belonging or not
to the same person:
>>> from sklearn.datasets import fetch_lfw_pairs
>>> lfw_pairs_train = fetch_lfw_pairs(subset='train')
>>> list(lfw_pairs_train.target_names)
['Different persons', 'Same person']
>>> lfw_pairs_train.pairs.shape
(2200, 2, 62, 47)
>>> lfw_pairs_train.data.shape
(2200, 5828)

>>> lfw_pairs_train.target.shape
(2200,)
Both for the sklearn.datasets.fetch_lfw_people and sklearn.datasets.fetch_lfw_pairs

function it is possible to get an additional dimension with the RGB color channels by passing color=True, in
that case the shape will be (2200, 2, 62, 47, 3).
The sklearn.datasets.fetch_lfw_pairs datasets is subdivided into 3 subsets: the development train
set, the development test set and an evaluation 10_folds set meant to compute performance metrics using a
10-folds cross validation scheme.
References:
• Labeled Faces in the Wild: A Database for Studying Face Recognition in Unconstrained Environments. Gary
B. Huang, Manu Ramesh, Tamara Berg, and Erik Learned-Miller. University of Massachusetts, Amherst,
Technical Report 07-49, October, 2007.
Examples
Faces recognition example using eigenfaces and SVMs
Forest covertypes
The samples in this dataset correspond to 30×30m patches of forest in the US, collected for the task of predicting
each patch’s cover type, i.e. the dominant species of tree. There are seven covertypes, making this a multiclass
classification problem. Each sample has 54 features, described on the dataset’s homepage. Some of the features are
boolean indicators, while others are discrete or continuous measurements.
Classes 7
Samples total 581012
Dimensionality 54
Features int
sklearn.datasets.fetch_covtype will load the covertype dataset; it returns a dictionary-like object with
the feature matrix in the data member and the target values in target. The dataset will be downloaded from the
web if necessary.
RCV1 dataset
Reuters Corpus Volume I (RCV1) is an archive of over 800,000 manually categorized newswire stories made available
by Reuters, Ltd. for research purposes. The dataset is extensively described in1 .
1 Lewis, D. D., Yang, Y., Rose, T. G., & Li, F. (2004). RCV1: A new benchmark collection for text categorization research. The Journal of
Machine Learning Research, 5, 361-397.

Classes 103
Dimensionality 47236
sklearn.datasets.fetch_rcv1 will load the following version: RCV1-v2, vectors, full sets, topics multil-
abels:
>>> from sklearn.datasets import fetch_rcv1

>>> rcv1 = fetch_rcv1()
It returns a dictionary-like object, with the following attributes:

data: The feature matrix is a scipy CSR sparse matrix, with 804414 samples and 47236 features. Non-zero values
contains cosine-normalized, log TF-IDF vectors. A nearly chronological split is proposed in1 : The first 23149 samples
are the training set. The last 781265 samples are the testing set. This follows the official LYRL2004 chronological
split. The array has 0.16% of non zero values:
>>> rcv1.data.shape
(804414, 47236)
target: The target values are stored in a scipy CSR sparse matrix, with 804414 samples and 103 categories. Each
sample has a value of 1 in its categories, and 0 in others. The array has 3.15% of non zero values:
>>> rcv1.target.shape
(804414, 103)
sample_id: Each sample can be identified by its ID, ranging (with gaps) from 2286 to 810596:
>>> rcv1.sample_id[:3]
array([2286, 2287, 2288], dtype=uint32)
target_names: The target values are the topics of each sample. Each sample belongs to at least one topic, and
to up to 17 topics. There are 103 topics, each represented by a string. Their corpus frequencies span five orders of
magnitude, from 5 occurrences for ‘GMIL’, to 381327 for ‘CCAT’:
>>> rcv1.target_names[:3].tolist()
['E11', 'ECAT', 'M11']
The dataset will be downloaded from the rcv1 homepage if necessary. The compressed size is about 656 MB.
References
Kddcup 99 dataset
The KDD Cup ‘99 dataset was created by processing the tcpdump portions of the 1998 DARPA Intrusion Detection
System (IDS) Evaluation dataset, created by MIT Lincoln Lab [1]. The artificial data (described on the dataset’s
homepage) was generated using a closed network and hand-injected attacks to produce a large number of different
types of attack with normal activity in the background. As the initial goal was to produce a large training set for
supervised learning algorithms, there is a large proportion (80.1%) of abnormal data which is unrealistic in real world,
and inappropriate for unsupervised anomaly detection which aims at detecting ‘abnormal’ data, ie
1. qualitatively different from normal data

2. in large minority among the observations.

We thus transform the KDD Data set into two different data sets: SA and SF.
-SA is obtained by simply selecting all the normal data, and a small proportion of abnormal data to gives an anomaly
proportion of 1%.
-SF is obtained as in [2] by simply picking up the data whose attribute logged_in is positive, thus focusing on the
intrusion attack, which gives a proportion of 0.3% of attack.
-http and smtp are two subsets of SF corresponding with third feature equal to ‘http’ (resp. to ‘smtp’)
General KDD structure :

Dimensionality 41
Features discrete (int) or continuous (float)
Targets str, ‘normal.’ or name of the anomaly type
SA structure :

Dimensionality 41
SF structure :

Dimensionality 4
http structure :

Dimensionality 3
smtp structure :
Samples total 95373

Dimensionality 3
sklearn.datasets.fetch_kddcup99 will load the kddcup99 dataset; it returns a dictionary-like object with
the feature matrix in the data member and the target values in target. The dataset will be downloaded from the
web if necessary.

California Housing dataset

Number of Attributes 8 numeric, predictive attributes and the target
• MedInc median income in block
• HouseAge median house age in block
• AveRooms average number of rooms
• AveBedrms average number of bedrooms
• Population block population
• AveOccup average house occupancy
• Latitude house block latitude
• Longitude house block longitude
This dataset was obtained from the StatLib repository. http://lib.stat.cmu.edu/datasets/
The target variable is the median house value for California districts.
This dataset was derived from the 1990 U.S. census, using one row per census block group. A block group is the
smallest geographical unit for which the U.S. Census Bureau publishes sample data (a block group typically has a
population of 600 to 3,000 people).
It can be downloaded/loaded using the sklearn.datasets.fetch_california_housing function.
References
• Pace, R. Kelley and Ronald Barry, Sparse Spatial Autoregressions, Statistics and Probability Letters, 33
(1997) 291-297
3.5.4 Generated datasets
In addition, scikit-learn includes various random sample generators that can be used to build artificial datasets of
controlled size and complexity.
Generators for classification and clustering
These generators produce a matrix of features and corresponding discrete targets.
Single label
Both make_blobs and make_classification create multiclass datasets by allocating each class one or more
normally-distributed clusters of points. make_blobs provides greater control regarding the centers and standard de-
viations of each cluster, and is used to demonstrate clustering. make_classification specialises in introducing

noise by way of: correlated, redundant and uninformative features; multiple Gaussian clusters per class; and linear
transformations of the feature space.
make_gaussian_quantiles divides a single Gaussian cluster into near-equal-size classes separated
by concentric hyperspheres. make_hastie_10_2 generates a similar binary, 10-dimensional problem.
make_circles and make_moons gener-

ate 2d binary classification datasets that are challenging to certain algorithms (e.g. centroid-based clustering or linear
classification), including optional Gaussian noise. They are useful for visualisation. produces Gaussian data with a
spherical decision boundary for binary classification.
Multilabel
make_multilabel_classification generates random samples with multiple labels, reflecting a bag of words
drawn from a mixture of topics. The number of topics for each document is drawn from a Poisson distribution, and the
topics themselves are drawn from a fixed random distribution. Similarly, the number of words is drawn from Poisson,
with words drawn from a multinomial, where each topic defines a probability distribution over words. Simplifications
with respect to true bag-of-words mixtures include:
• Per-topic word distributions are independently drawn, where in reality all would be affected by a sparse base
distribution, and would be correlated.
• For a document generated from multiple topics, all topics are weighted equally in generating its bag of words.
• Documents without labels words at random, rather than from a base distribution.

Biclustering
make_biclusters(shape, n_clusters[, noise, . . . ]) Generate an array with constant block diagonal structure
for biclustering.
make_checkerboard(shape, n_clusters[, . . . ]) Generate an array with block checkerboard structure for bi-
clustering.
Generators for regression
make_regression produces regression targets as an optionally-sparse random linear combination of random fea-
tures, with noise. Its informative features may be uncorrelated, or low rank (few features account for most of the
variance).
Other regression generators generate functions deterministically from randomized features.
make_sparse_uncorrelated produces a target as a linear combination of four features with fixed coef-
ficients. Others encode explicitly non-linear relations: make_friedman1 is related by polynomial and sine
transforms; make_friedman2 includes feature multiplication and reciprocation; and make_friedman3 is
similar with an arctan transformation on the target.
Generators for manifold learning
make_s_curve([n_samples, noise, random_state]) Generate an S curve dataset.

make_swiss_roll([n_samples, noise, random_state]) Generate a swiss roll dataset.
Generators for decomposition
make_low_rank_matrix([n_samples, . . . ]) Generate a mostly low rank matrix with bell-shaped singu-

lar values
make_sparse_coded_signal(n_samples, . . . [, . . . ]) Generate a signal as a sparse combination of dictionary el-
ements.
make_spd_matrix(n_dim[, random_state]) Generate a random symmetric, positive-definite matrix.
make_sparse_spd_matrix([dim, alpha, . . . ]) Generate a sparse symmetric definite positive matrix.

3.5.5 Loading other datasets
Sample images
Scikit-learn also embed a couple of sample JPEG images published under Creative Commons license by their authors.
Those image can be useful to test algorithms and pipeline on 2D data.
load_sample_images() Load sample images for image manipulation.

load_sample_image(image_name) Load the numpy array of a single sample image
Warning: The default coding of images is based on the uint8 dtype to spare memory. Often machine learning
algorithms work best if the input is converted to a floating point representation first. Also, if you plan to use
matplotlib.pyplpt.imshow don’t forget to scale to the range 0 - 1 as done in the following example.
Examples:
• Color Quantization using K-Means
Datasets in svmlight / libsvm format
scikit-learn includes utility functions for loading datasets in the svmlight / libsvm format. In this format, each line
takes the form <label> <feature-id>:<feature-value> <feature-id>:<feature-value> ..
.. This format is especially suitable for sparse datasets. In this module, scipy sparse CSR matrices are used for X and
numpy arrays are used for y.
You may load a dataset like as follows:
>>> from sklearn.datasets import load_svmlight_file

>>> X_train, y_train = load_svmlight_file("/path/to/train_dataset.txt")
...
You may also load two (or more) datasets at once:
>>> X_train, y_train, X_test, y_test = load_svmlight_files(

... ("/path/to/train_dataset.txt", "/path/to/test_dataset.txt"))
...
In this case, X_train and X_test are guaranteed to have the same number of features. Another way to achieve the
same result is to fix the number of features:

>>> X_test, y_test = load_svmlight_file(

... "/path/to/test_dataset.txt", n_features=X_train.shape[1])
...
Related links:
Public datasets in svmlight / libsvm format: https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets

Faster API-compatible implementation: https://github.com/mblondel/svmlight-loader
Downloading datasets from the openml.org repository
openml.org is a public repository for machine learning data and experiments, that allows everybody to upload open
datasets.
The sklearn.datasets package is able to download datasets from the repository using the function sklearn.
datasets.fetch_openml.
For example, to download a dataset of gene expressions in mice brains:
>>> from sklearn.datasets import fetch_openml

>>> mice = fetch_openml(name='miceprotein', version=4)
To fully specify a dataset, you need to provide a name and a version, though the version is optional, see Dataset
Versions below. The dataset contains a total of 1080 examples belonging to 8 different classes:
>>> mice.data.shape
(1080, 77)
>>> mice.target.shape
(1080,)
>>> np.unique(mice.target)
array(['c-CS-m', 'c-CS-s', 'c-SC-m', 'c-SC-s', 't-CS-m', 't-CS-s', 't-SC-m', 't-SC-s
˓→'], dtype=object)
You can get more information on the dataset by looking at the DESCR and details attributes:
>>> print(mice.DESCR)
**Author**: Clara Higuera, Katheleen J. Gardiner, Krzysztof J. Cios
**Source**: [UCI](https://archive.ics.uci.edu/ml/datasets/Mice+Protein+Expression) -
˓→2015
**Please cite**: Higuera C, Gardiner KJ, Cios KJ (2015) Self-Organizing

Feature Maps Identify Proteins Critical to Learning in a Mouse Model of Down
Syndrome. PLoS ONE 10(6): e0129126...
>>> mice.details
{'id': '40966', 'name': 'MiceProtein', 'version': '4', 'format': 'ARFF',
'upload_date': '2017-11-08T16:00:15', 'licence': 'Public',
'url': 'https://www.openml.org/data/v1/download/17928620/MiceProtein.arff',
'file_id': '17928620', 'default_target_attribute': 'class',
'row_id_attribute': 'MouseID',
'ignore_attribute': ['Genotype', 'Treatment', 'Behavior'],
'tag': ['OpenML-CC18', 'study_135', 'study_98', 'study_99'],
'visibility': 'public', 'status': 'active',
'md5_checksum': '3c479a6885bfa0438971388283a1ce32'}

The DESCR contains a free-text description of the data, while details contains a dictionary of meta-data stored
by openml, like the dataset id. For more details, see the OpenML documentation The data_id of the mice protein
dataset is 40966, and you can use this (or the name) to get more information on the dataset on the openml website:
>>> mice.url
'https://www.openml.org/d/40966'
The data_id also uniquely identifies a dataset from OpenML:
>>> mice = fetch_openml(data_id=40966)

>>> mice.details
{'id': '4550', 'name': 'MiceProtein', 'version': '1', 'format': 'ARFF',
'creator': ...,
'upload_date': '2016-02-17T14:32:49', 'licence': 'Public', 'url':
'https://www.openml.org/data/v1/download/1804243/MiceProtein.ARFF', 'file_id':
'1804243', 'default_target_attribute': 'class', 'citation': 'Higuera C,
Gardiner KJ, Cios KJ (2015) Self-Organizing Feature Maps Identify Proteins
Critical to Learning in a Mouse Model of Down Syndrome. PLoS ONE 10(6):
e0129126. [Web Link] journal.pone.0129126', 'tag': ['OpenML100', 'study_14',
'study_34'], 'visibility': 'public', 'status': 'active', 'md5_checksum':
'3c479a6885bfa0438971388283a1ce32'}
Dataset Versions
A dataset is uniquely specified by its data_id, but not necessarily by its name. Several different “versions” of a
dataset with the same name can exist which can contain entirely different datasets. If a particular version of a dataset
has been found to contain significant issues, it might be deactivated. Using a name to specify a dataset will yield the
earliest version of a dataset that is still active. That means that fetch_openml(name="miceprotein") can
yield different results at different times if earlier versions become inactive. You can see that the dataset with data_id
40966 that we fetched above is the version 1 of the “miceprotein” dataset:
>>> mice.details['version']
'1'
In fact, this dataset only has one version. The iris dataset on the other hand has multiple versions:
>>> iris = fetch_openml(name="iris")

>>> iris.details['version']
'1'
>>> iris.details['id']
'61'
>>> iris_61 = fetch_openml(data_id=61)

>>> iris_61.details['version']
'1'
>>> iris_61.details['id']
'61'
>>> iris_969 = fetch_openml(data_id=969)

>>> iris_969.details['version']
'3'
>>> iris_969.details['id']
'969'
Specifying the dataset by the name “iris” yields the lowest version, version 1, with the data_id 61. To make sure
you always get this exact dataset, it is safest to specify it by the dataset data_id. The other dataset, with data_id

969, is version 3 (version 2 has become inactive), and contains a binarized version of the data:
>>> np.unique(iris_969.target)
array(['N', 'P'], dtype=object)
You can also specify both the name and the version, which also uniquely identifies the dataset:
>>> iris_version_3 = fetch_openml(name="iris", version=3)

>>> iris_version_3.details['version']
'3'
>>> iris_version_3.details['id']
'969'
References:
• Vanschoren, van Rijn, Bischl and Torgo “OpenML: networked science in machine learning”, ACM SIGKDD
Explorations Newsletter, 15(2), 49-60, 2014.
Loading from external datasets
scikit-learn works on any numeric data stored as numpy arrays or scipy sparse matrices. Other types that are convertible
to numeric arrays such as pandas DataFrame are also acceptable.
Here are some recommended ways to load standard columnar data into a format usable by scikit-learn:
• pandas.io provides tools to read data from common formats including CSV, Excel, JSON and SQL. DataFrames
may also be constructed from lists of tuples or dicts. Pandas handles heterogeneous data smoothly and provides
tools for manipulation and conversion into a numeric array suitable for scikit-learn.
• scipy.io specializes in binary formats often used in scientific computing context such as .mat and .arff
• numpy/routines.io for standard loading of columnar data into numpy arrays
• scikit-learn’s datasets.load_svmlight_file for the svmlight or libSVM sparse format
• scikit-learn’s datasets.load_files for directories of text files where the name of each directory is the
name of each category and each file inside of each directory corresponds to one sample from that category
For some miscellaneous data such as images, videos, and audio, you may wish to refer to:
• skimage.io or Imageio for loading images and videos into numpy arrays
• scipy.io.wavfile.read for reading WAV files into a numpy array
Categorical (or nominal) features stored as strings (common in pandas DataFrames) will need converting to
numerical features using sklearn.preprocessing.OneHotEncoder or sklearn.preprocessing.
OrdinalEncoder or similar. See Preprocessing data.
Note: if you manage your own numerical data it is recommended to use an optimized file format such as HDF5 to
reduce data load times. Various libraries such as H5Py, PyTables and pandas provides a Python interface for reading
and writing data in that format.

3.6 Computing with scikit-learn
3.6.1 Strategies to scale computationally: bigger data
For some applications the amount of examples, features (or both) and/or the speed at which they need to be processed
are challenging for traditional approaches. In these cases scikit-learn has a number of options you can consider to
make your system scale.
Scaling with instances using out-of-core learning
Out-of-core (or “external memory”) learning is a technique used to learn from data that cannot fit in a computer’s main
memory (RAM).
Here is a sketch of a system designed to achieve this goal:
1. a way to stream instances
2. a way to extract features from instances
3. an incremental algorithm
Streaming instances
Basically, 1. may be a reader that yields instances from files on a hard drive, a database, from a network stream etc.
However, details on how to achieve this are beyond the scope of this documentation.
Extracting features
2. could be any relevant way to extract features among the different feature extraction methods supported by scikit-
learn. However, when working with data that needs vectorization and where the set of features or values is not
known in advance one should take explicit care. A good example is text classification where unknown terms are
likely to be found during training. It is possible to use a stateful vectorizer if making multiple passes over the data
is reasonable from an application point of view. Otherwise, one can turn up the difficulty by using a stateless feature
extractor. Currently the preferred way to do this is to use the so-called hashing trick as implemented by sklearn.
feature_extraction.FeatureHasher for datasets with categorical variables represented as list of Python
dicts or sklearn.feature_extraction.text.HashingVectorizer for text documents.
Incremental learning
Finally, for 3. we have a number of options inside scikit-learn. Although not all algorithms can learn incrementally
(i.e. without seeing all the instances at once), all estimators implementing the partial_fit API are candidates.
Actually, the ability to learn incrementally from a mini-batch of instances (sometimes called “online learning”) is key
to out-of-core learning as it guarantees that at any given time there will be only a small amount of instances in the
main memory. Choosing a good size for the mini-batch that balances relevancy and memory footprint could involve
some tuning1 .
Here is a list of incremental estimators for different tasks:
• Classification
1 Depending on the algorithm the mini-batch size can influence results or not. SGD*, PassiveAggressive*, and discrete NaiveBayes are truly
online and are not affected by batch size. Conversely, MiniBatchKMeans convergence rate is affected by the batch size. Also, its memory footprint
can vary dramatically with batch size.

– sklearn.naive_bayes.MultinomialNB
– sklearn.naive_bayes.BernoulliNB
– sklearn.linear_model.Perceptron
– sklearn.linear_model.SGDClassifier
– sklearn.linear_model.PassiveAggressiveClassifier
• Regression
– sklearn.linear_model.SGDRegressor
– sklearn.linear_model.PassiveAggressiveRegressor
– sklearn.neural_network.MLPRegressor
• Clustering
– sklearn.cluster.MiniBatchKMeans
– sklearn.cluster.Birch
• Decomposition / feature Extraction
– sklearn.decomposition.MiniBatchDictionaryLearning
– sklearn.decomposition.IncrementalPCA
– sklearn.decomposition.LatentDirichletAllocation
• Preprocessing
– sklearn.preprocessing.StandardScaler
– sklearn.preprocessing.MinMaxScaler
– sklearn.preprocessing.MaxAbsScaler
For classification, a somewhat important thing to note is that although a stateless feature extraction routine may be
able to cope with new/unseen attributes, the incremental learner itself may be unable to cope with new/unseen targets
classes. In this case you have to pass all the possible classes to the first partial_fit call using the classes=
parameter.
Another aspect to consider when choosing a proper algorithm is that not all of them put the same importance on each
example over time. Namely, the Perceptron is still sensitive to badly labeled examples even after many examples
whereas the SGD* and PassiveAggressive* families are more robust to this kind of artifacts. Conversely, the
latter also tend to give less importance to remarkably different, yet properly labeled examples when they come late in
the stream as their learning rate decreases over time.
Examples
Finally, we have a full-fledged example of Out-of-core classification of text documents. It is aimed at providing a
starting point for people wanting to build out-of-core learning systems and demonstrates most of the notions discussed
above.
Furthermore, it also shows the evolution of the performance of different algorithms with the number of processed
examples.
3.6. Computing with scikit-learn 609

Now looking at the computation time of the different parts, we see that the vectorization is much more expensive
than learning itself. From the different algorithms, MultinomialNB is the most expensive, but its overhead can be
mitigated by increasing the size of the mini-batches (exercise: change minibatch_size to 100 and 10000 in the
program and compare).

Notes
3.6.2 Computational Performance
For some applications the performance (mainly latency and throughput at prediction time) of estimators is crucial. It
may also be of interest to consider the training throughput but this is often less important in a production setup (where
it often takes place offline).
We will review here the orders of magnitude you can expect from a number of scikit-learn estimators in different
contexts and provide some tips and tricks for overcoming performance bottlenecks.
Prediction latency is measured as the elapsed time necessary to make a prediction (e.g. in micro-seconds). Latency
is often viewed as a distribution and operations engineers often focus on the latency at a given percentile of this
distribution (e.g. the 90 percentile).
Prediction throughput is defined as the number of predictions the software can deliver in a given amount of time (e.g.
in predictions per second).
An important aspect of performance optimization is also that it can hurt prediction accuracy. Indeed, simpler models
(e.g. linear instead of non-linear, or with fewer parameters) often run faster but are not always able to take into account
the same exact properties of the data as more complex ones.
Prediction Latency
One of the most straight-forward concerns one may have when using/choosing a machine learning toolkit is the latency
at which predictions can be made in a production environment.
The main factors that influence the prediction latency are
1. Number of features
2. Input data representation and sparsity
3. Model complexity
4. Feature extraction
A last major parameter is also the possibility to do predictions in bulk or one-at-a-time mode.
Bulk versus Atomic mode
In general doing predictions in bulk (many instances at the same time) is more efficient for a number of reasons
(branching predictability, CPU cache, linear algebra libraries optimizations etc.). Here we see on a setting with few
features that independently of estimator choice the bulk mode is always faster, and for some of them by 1 to 2 orders
of magnitude:

To benchmark different estimators for your case you can simply change the n_features parameter in this example:
Prediction Latency. This should give you an estimate of the order of magnitude of the prediction latency.

Configuring Scikit-learn for reduced validation overhead
Scikit-learn does some validation on data that increases the overhead per call to predict and similar functions.
In particular, checking that features are finite (not NaN or infinite) involves a full pass over the data. If you en-
sure that your data is acceptable, you may suppress checking for finiteness by setting the environment variable
SKLEARN_ASSUME_FINITE to a non-empty string before importing scikit-learn, or configure it in Python with
sklearn.set_config. For more control than these global settings, a config_context allows you to set this
configuration within a specified context:
>>> import sklearn

>>> with sklearn.config_context(assume_finite=True):
... pass # do learning/prediction here with reduced validation
Note that this will affect all uses of sklearn.utils.assert_all_finite within the context.
Influence of the Number of Features
Obviously when the number of features increases so does the memory consumption of each example. Indeed, for a
matrix of 𝑀 instances with 𝑁 features, the space complexity is in 𝑂(𝑁 𝑀 ). From a computing perspective it also
means that the number of basic operations (e.g., multiplications for vector-matrix products in linear models) increases
too. Here is a graph of the evolution of the prediction latency with the number of features:
Overall you can expect the prediction time to increase at least linearly with the number of features (non-linear cases
can happen depending on the global memory footprint and estimator).

Influence of the Input Data Representation
Scipy provides sparse matrix data structures which are optimized for storing sparse data. The main feature of sparse
formats is that you don’t store zeros so if your data is sparse then you use much less memory. A non-zero value in
a sparse (CSR or CSC) representation will only take on average one 32bit integer position + the 64 bit floating point
value + an additional 32bit per row or column in the matrix. Using sparse input on a dense (or sparse) linear model
can speedup prediction by quite a bit as only the non zero valued features impact the dot product and thus the model
predictions. Hence if you have 100 non zeros in 1e6 dimensional space, you only need 100 multiply and add operation
instead of 1e6.
Calculation over a dense representation, however, may leverage highly optimised vector operations and multithreading
in BLAS, and tends to result in fewer CPU cache misses. So the sparsity should typically be quite high (10% non-zeros
max, to be checked depending on the hardware) for the sparse input representation to be faster than the dense input
representation on a machine with many CPUs and an optimized BLAS implementation.
Here is sample code to test the sparsity of your input:
def sparsity_ratio(X):
return 1.0 - np.count_nonzero(X) / float(X.shape[0] * X.shape[1])
print("input sparsity ratio:", sparsity_ratio(X))
As a rule of thumb you can consider that if the sparsity ratio is greater than 90% you can probably benefit from sparse
formats. Check Scipy’s sparse matrix formats documentation for more information on how to build (or convert your
data to) sparse matrix formats. Most of the time the CSR and CSC formats work best.
Influence of the Model Complexity
Generally speaking, when model complexity increases, predictive power and latency are supposed to increase. In-
creasing predictive power is usually interesting, but for many applications we would better not increase prediction
latency too much. We will now review this idea for different families of supervised models.
For sklearn.linear_model (e.g. Lasso, ElasticNet, SGDClassifier/Regressor, Ridge & RidgeClassifier, Pas-
siveAggressiveClassifier/Regressor, LinearSVC, LogisticRegression. . . ) the decision function that is applied at pre-
diction time is the same (a dot product) , so latency should be equivalent.
Here is an example using sklearn.linear_model.stochastic_gradient.SGDClassifier with the
elasticnet penalty. The regularization strength is globally controlled by the alpha parameter. With a sufficiently
high alpha, one can then increase the l1_ratio parameter of elasticnet to enforce various levels of sparsity
in the model coefficients. Higher sparsity here is interpreted as less model complexity as we need fewer coefficients to
describe it fully. Of course sparsity influences in turn the prediction time as the sparse dot-product takes time roughly
proportional to the number of non-zero coefficients.

For the sklearn.svm family of algorithms with a non-linear kernel, the latency is tied to the number of support
vectors (the fewer the faster). Latency and throughput should (asymptotically) grow linearly with the number of
support vectors in a SVC or SVR model. The kernel will also influence the latency as it is used to compute the
projection of the input vector once per support vector. In the following graph the nu parameter of sklearn.svm.
classes.NuSVR was used to influence the number of support vectors.
For sklearn.ensemble of trees (e.g. RandomForest, GBT, ExtraTrees etc) the number of trees and their
depth play the most important role. Latency and throughput should scale linearly with the number of trees. In
this case we used directly the n_estimators parameter of sklearn.ensemble.gradient_boosting.
GradientBoostingRegressor.

In any case be warned that decreasing model complexity can hurt accuracy as mentioned above. For instance a non-
linearly separable problem can be handled with a speedy linear model but prediction power will very likely suffer in
the process.
Feature Extraction Latency
Most scikit-learn models are usually pretty fast as they are implemented either with compiled Cython extensions or
optimized computing libraries. On the other hand, in many real world applications the feature extraction process (i.e.
turning raw data like database rows or network packets into numpy arrays) governs the overall prediction time. For
example on the Reuters text classification task the whole preparation (reading and parsing SGML files, tokenizing the
text and hashing it into a common vector space) is taking 100 to 500 times more time than the actual prediction code,
depending on the chosen model.

In many cases it is thus recommended to carefully time and profile your feature extraction code as it may be a good
place to start optimizing when your overall latency is too slow for your application.
Prediction Throughput
Another important metric to care about when sizing production systems is the throughput i.e. the number of predictions
you can make in a given amount of time. Here is a benchmark from the Prediction Latency example that measures this
quantity for a number of estimators on synthetic data:

These throughputs are achieved on a single process. An obvious way to increase the throughput of your application
is to spawn additional instances (usually processes in Python because of the GIL) that share the same model. One
might also add machines to spread the load. A detailed explanation on how to achieve this is beyond the scope of this
documentation though.
Tips and Tricks
Linear algebra libraries
As scikit-learn relies heavily on Numpy/Scipy and linear algebra in general it makes sense to take explicit care of the
versions of these libraries. Basically, you ought to make sure that Numpy is built using an optimized BLAS / LAPACK
library.
Not all models benefit from optimized BLAS and Lapack implementations. For instance models based on (random-
ized) decision trees typically do not rely on BLAS calls in their inner loops, nor do kernel SVMs (SVC, SVR, NuSVC,
NuSVR). On the other hand a linear model implemented with a BLAS DGEMM call (via numpy.dot) will typically
benefit hugely from a tuned BLAS implementation and lead to orders of magnitude speedup over a non-optimized
BLAS.
You can display the BLAS / LAPACK implementation used by your NumPy / SciPy / scikit-learn install with the
following commands:
from numpy.distutils.system_info import get_info

print(get_info('blas_opt'))
print(get_info('lapack_opt'))
Optimized BLAS / LAPACK implementations include:

• Atlas (need hardware specific tuning by rebuilding on the target machine)

• OpenBLAS
• MKL
• Apple Accelerate and vecLib frameworks (OSX only)
More information can be found on the Scipy install page and in this blog post from Daniel Nouri which has some nice
step by step install instructions for Debian / Ubuntu.
Warning: Multithreaded BLAS libraries sometimes conflict with Python’s multiprocessing module, which
is used by e.g. GridSearchCV and most other estimators that take an n_jobs argument (with the exception of
SGDClassifier, SGDRegressor, Perceptron, PassiveAggressiveClassifier and tree-based
methods such as random forests). This is true of Apple’s Accelerate and OpenBLAS when built with OpenMP
support.
Besides scikit-learn, NumPy and SciPy also use BLAS internally, as explained earlier.
If you experience hanging subprocesses with n_jobs>1 or n_jobs=-1, make sure you have a single-threaded
BLAS library, or set n_jobs=1, or upgrade to Python 3.4 which has a new version of multiprocessing that
should be immune to this problem.
Limiting Working Memory
Some calculations when implemented using standard numpy vectorized operations involve using a large amount of
temporary memory. This may potentially exhaust system memory. Where computations can be performed in fixed-
memory chunks, we attempt to do so, and allow the user to hint at the maximum size of this working memory (de-
faulting to 1GB) using sklearn.set_config or config_context. The following suggests to limit temporary
working memory to 128 MiB:
>>> import sklearn

>>> with sklearn.config_context(working_memory=128):
... pass # do chunked work here
An example of a chunked operation adhering to this setting is metric.pairwise_distances_chunked, which

facilitates computing row-wise reductions of a pairwise distance matrix.
Model Compression
Model compression in scikit-learn only concerns linear models for the moment. In this context it means that we want
to control the model sparsity (i.e. the number of non-zero coordinates in the model vectors). It is generally a good
idea to combine model sparsity with sparse input data representation.
Here is sample code that illustrates the use of the sparsify() method:
clf = SGDRegressor(penalty='elasticnet', l1_ratio=0.25)

clf.fit(X_train, y_train).sparsify()
clf.predict(X_test)
In this example we prefer the elasticnet penalty as it is often a good compromise between model compactness
and prediction power. One can also further tune the l1_ratio parameter (in combination with the regularization
strength alpha) to control this tradeoff.
A typical benchmark on synthetic data yields a >30% decrease in latency when both the model and input are sparse
(with 0.000024 and 0.027400 non-zero coefficients ratio respectively). Your mileage may vary depending on the

sparsity and size of your data and model. Furthermore, sparsifying can be very useful to reduce the memory usage of
predictive models deployed on production servers.
Model Reshaping
Model reshaping consists in selecting only a portion of the available features to fit a model. In other words, if a
model discards features during the learning phase we can then strip those from the input. This has several benefits.
Firstly it reduces memory (and therefore time) overhead of the model itself. It also allows to discard explicit feature
selection components in a pipeline once we know which features to keep from a previous run. Finally, it can help
reduce processing time and I/O usage upstream in the data access and feature extraction layers by not collecting and
building features that are discarded by the model. For instance if the raw data come from a database, it can make it
possible to write simpler and faster queries or reduce I/O usage by making the queries return lighter records. At the
moment, reshaping needs to be performed manually in scikit-learn. In the case of sparse input (particularly in CSR
format), it is generally sufficient to not generate the relevant features, leaving their columns empty.
Links
• scikit-learn developer performance documentation

• Scipy sparse matrix formats documentation
3.6.3 Parallelism, resource management, and configuration
Parallel and distributed computing
Scikit-learn uses the joblib library to enable parallel computing inside its estimators. See the joblib documentation for
the switches to control parallel computing.
Note that, by default, scikit-learn uses its embedded (vendored) version of joblib. A configuration switch (documented
below) controls this behavior.
Configuration switches
Python runtime
sklearn.set_config controls the following behaviors:

assume_finite used to skip validation, which enables faster computations but may lead to segmentation
faults if the data contains NaNs.
working_memory the optimal size of temporary arrays used by some algoritms.
Environment variables
These environment variables should be set before importing scikit-learn.

SKLEARN_SITE_JOBLIB When this environment variable is set to a non zero value, scikit-learn uses
the site joblib rather than its vendored version. Consequently, joblib must be installed for scikit-learn
to run. Note that using the site joblib is at your own risks: the versions of scikit-learn and joblib
need to be compatible. Currently, joblib 0.11+ is supported. In addition, dumps from joblib.Memory
might be incompatible, and you might loose some caches and have to redownload some datasets.

SKLEARN_ASSUME_FINITE Sets the default value for the assume_finite argument of sklearn.
set_config.
SKLEARN_WORKING_MEMORY Sets the default value for the working_memory argument of
sklearn.set_config.
SKLEARN_SEED Sets the seed of the global random generator when running the tests, for reproducibil-
ity.
SKLEARN_SKIP_NETWORK_TESTS When this environment variable is set to a non zero value, the
tests that need network access are skipped.


CHAPTER
FOUR
GLOSSARY OF COMMON TERMS AND API ELEMENTS
This glossary hopes to definitively represent the tacit and explicit conventions applied in Scikit-learn and its API, while
providing a reference for users and contributors. It aims to describe the concepts and either detail their corresponding
API or link to other relevant parts of the documentation which do so. By linking to glossary entries from the API
Reference and User Guide, we may minimize redundancy and inconsistency.
We begin by listing general concepts (and any that didn’t fit elsewhere), but more specific sets of related terms are listed
below: Class APIs and Estimator Types, Target Types, Methods, Parameters, Attributes, Data and sample properties.
4.1 General Concepts
1d
1d array One-dimensional array. A NumPy array whose .shape has length 1. A vector.
2d
2d array Two-dimensional array. A NumPy array whose .shape has length 2. Often represents a matrix.
API Refers to both the specific interfaces for estimators implemented in Scikit-learn and the generalized conventions
across types of estimators as described in this glossary and overviewed in the contributor documentation.
The specific interfaces that constitute Scikit-learn’s public API are largely documented in API Reference. How-
ever we less formally consider anything as public API if none of the identifiers required to access it begins with
_. We generally try to maintain backwards compatibility for all objects in the public API.
Private API, including functions, modules and methods beginning _ are not assured to be stable.
array-like The most common data format for input to Scikit-learn estimators and functions, array-like is any type
object for which numpy.asarray will produce an array of appropriate shape (usually 1 or 2-dimensional) of
appropriate dtype (usually numeric).
This includes:
• a numpy array
• a list of numbers
• a list of length-k lists of numbers for some fixed length k
• a pandas.DataFrame with all columns numeric
• a numeric pandas.Series
It excludes:
• a sparse matrix
• an iterator
623
• a generator
Note that output from scikit-learn estimators and functions (e.g. predictions) should generally be arrays or sparse
matrices, or lists thereof (as in multi-output tree.DecisionTreeClassifier’s predict_proba). An
estimator where predict() returns a list or a pandas.Series is not valid.
attribute
attributes We mostly use attribute to refer to how model information is stored on an estimator during fitting. Any pub-
lic attribute stored on an estimator instance is required to begin with an alphabetic character and end in a single
underscore if it is set in fit or partial_fit. These are what is documented under an estimator’s Attributes documen-
tation. The information stored in attributes is usually either: sufficient statistics used for prediction or transfor-
mation; transductive outputs such as labels_ or embedding_; or diagnostic data, such as feature_importances_.
Common attributes are listed below.
A public attribute may have the same name as a constructor parameter, with a _ appended. This is used to
store a validated or estimated version of the user’s input. For example, decomposition.PCA is constructed
with an n_components parameter. From this, together with other parameters and the data, PCA estimates the
attribute n_components_.
Further private attributes used in prediction/transformation/etc. may also be set when fitting. These begin with
a single underscore and are not assured to be stable for public access.
A public attribute on an estimator instance that does not end in an underscore should be the stored, unmodified
value of an __init__ parameter of the same name. Because of this equivalence, these are documented under
an estimator’s Parameters documentation.
backwards compatibility We generally try to maintain backwards compatibility (i.e. interfaces and behaviors may
be extended but not changed or removed) from release to release but this comes with some exceptions:
Public API only The behaviour of objects accessed through private identifiers (those beginning _) may be
changed arbitrarily between versions.
As documented We will generally assume that the users have adhered to the documented parameter types and
ranges. If the documentation asks for a list and the user gives a tuple, we do not assure consistent behavior
from version to version.
Deprecation Behaviors may change following a deprecation period (usually two releases long). Warnings are
issued using Python’s warnings module.
Keyword arguments We may sometimes assume that all optional parameters (other than X and y to fit and
similar methods) are passed as keyword arguments only and may be positionally reordered.
Bug fixes and enhancements Bug fixes and – less often – enhancements may change the behavior of estima-
tors, including the predictions of an estimator trained on the same data and random_state. When this
happens, we attempt to note it clearly in the changelog.
Serialization We make no assurances that pickling an estimator in one version will allow it to be unpickled
to an equivalent model in the subsequent version. (For estimators in the sklearn package, we issue a
warning when this unpickling is attempted, even if it may happen to work.) See Security & maintainability
limitations.
utils.estimator_checks.check_estimator We provide limited backwards compatibility assur-
ances for the estimator checks: we may add extra requirements on estimators tested with this function,
usually when these were informally assumed but not formally tested.
Despite this informal contract with our users, the software is provided as is, as stated in the licence. When
a release inadvertently introduces changes that are not backwards compatible, these are known as software
regressions.
callable A function, class or an object which implements the __call__ method; anything that returns True when
the argument of callable().
624 Chapter 4. Glossary of Common Terms and API Elements

categorical feature A categorical or nominal feature is one that has a finite set of discrete values across the pop-
ulation of data. These are commonly represented as columns of integers or strings. Strings will be rejected
by most scikit-learn estimators, and integers will be treated as ordinal or count-valued. For the use with most
estimators, categorical variables should be one-hot encoded. Notable exceptions include tree-based models such
as random forests and gradient boosting models that often work better and faster with integer-coded categor-
ical variables. OrdinalEncoder helps encoding string-valued categorical features as ordinal integers, and
OneHotEncoder can be used to one-hot encode categorical features. See also Encoding categorical features
and the http://contrib.scikit-learn.org/categorical-encoding package for tools related to encoding categorical fea-
tures.
clone
cloned To copy an estimator instance and create a new one with identical parameters, but without any fitted attributes,
using clone.
When fit is called, a meta-estimator usually clones a wrapped estimator instance before fitting the cloned
instance. (Exceptions, for legacy reasons, include Pipeline and FeatureUnion.)
common tests This refers to the tests run on almost every estimator class in Scikit-learn to check they comply
with basic API conventions. They are available for external use through utils.estimator_checks.
check_estimator, with most of the implementation in sklearn/utils/estimator_checks.py.
Note: Some exceptions to the common testing regime are currently hard-coded into the library, but we hope to
replace this by marking exceptional behaviours on the estimator using semantic estimator tags.
deprecation We use deprecation to slowly violate our backwards compatibility assurances, usually to to:
• change the default value of a parameter; or
• remove a parameter, attribute, method, class, etc.
We will ordinarily issue a warning when a deprecated element is used, although there may be limitations to this.
For instance, we will raise a warning when someone sets a parameter that has been deprecated, but may not
when they access that parameter’s attribute on the estimator instance.
See the Contributors’ Guide.
dimensionality May be used to refer to the number of features (i.e. n_features), or columns in a 2d feature matrix.
Dimensions are, however, also used to refer to the length of a NumPy array’s shape, distinguishing a 1d array
from a 2d matrix.
docstring The embedded documentation for a module, class, function, etc., usually in code as a string at the beginning
of the object’s definition, and accessible as the object’s __doc__ attribute.
We try to adhere to PEP257, and follow NumpyDoc conventions.
double underscore
double underscore notation When specifying parameter names for nested estimators, __ may be used to separate
between parent and child in some contexts. The most common use is when setting parameters through a meta-
estimator with set_params and hence in specifying a search grid in parameter search. See parameter. It is also
used in pipeline.Pipeline.fit for passing sample properties to the fit methods of estimators in the
pipeline.
dtype
data type NumPy arrays assume a homogeneous data type throughout, available in the .dtype attribute of an array
(or sparse matrix). We generally assume simple data types for scikit-learn data: float or integer. We may support
object or string data types for arrays before encoding or vectorizing. Our estimators do not work with struct
arrays, for instance.
TODO: Mention efficiency and precision issues; casting policy.
4.1. General Concepts 625

duck typing We try to apply duck typing to determine how to handle some input values (e.g. checking whether a
given estimator is a classifier). That is, we avoid using isinstance where possible, and rely on the presence
or absence of attributes to determine an object’s behaviour. Some nuance is required when following this
approach:
• For some estimators, an attribute may only be available once it is fitted. For instance, we cannot a priori
determine if predict_proba is available in a grid search where the grid includes alternating between a
probabilistic and a non-probabilistic predictor in the final step of the pipeline. In the following, we can
only determine if clf is probabilistic after fitting it on some data:

>>> clf = GridSearchCV(SGDClassifier(),
... param_grid={'loss': ['log', 'hinge']})
This means that we can only check for duck-typed attributes after fitting, and that we must be careful to
make meta-estimators only present attributes according to the state of the underlying estimator after fitting.
• Checking if an attribute is present (using hasattr) is in general just as expensive as getting the attribute
(getattr or dot notation). In some cases, getting the attribute may indeed be expensive (e.g. for some
implementations of feature_importances_, which may suggest this is an API design flaw). So code which
does hasattr followed by getattr should be avoided; getattr within a try-except block is pre-
ferred.
• For determining some aspects of an estimator’s expectations or support for some feature, we use estimator
tags instead of duck typing.
early stopping This consists in stopping an iterative optimization method before the convergence of the training loss,
to avoid over-fitting. This is generally done by monitoring the generalization score on a validation set. When
available, it is activated through the parameter early_stopping or by setting a positive n_iter_no_change.
estimator instance We sometimes use this terminology to distinguish an estimator class from a constructed instance.
For example, in the following, cls is an estimator class, while est1 and est2 are instances:
cls = RandomForestClassifier
est1 = cls()
est2 = RandomForestClassifier()
examples We try to give examples of basic usage for most functions and classes in the API:
• as doctests in their docstrings (i.e. within the sklearn/ library code itself).
• as examples in the example gallery rendered (using sphinx-gallery) from scripts in the examples/ direc-
tory, exemplifying key features or parameters of the estimator/function. These should also be referenced
from the User Guide.
• sometimes in the User Guide (built from doc/) alongside a technical description of the estimator.
evaluation metric
evaluation metrics Evaluation metrics give a measure of how well a model performs. We may use this term specif-
ically to refer to the functions in metrics (disregarding metrics.pairwise), as distinct from the score
method and the scoring API used in cross validation. See Model evaluation: quantifying the quality of predic-
tions.
These functions usually accept a ground truth (or the raw data where the metric evaluates clustering without
a ground truth) and a prediction, be it the output of predict (y_pred), of predict_proba (y_proba), or of
an arbitrary score function including decision_function (y_score). Functions are usually named to end with
_score if a greater score indicates a better model, and _loss if a lesser score indicates a better model. This
diversity of interface motivates the scoring API.

Note that some estimators can calculate metrics that are not included in metrics and are estimator-specific,
notably model likelihoods.
estimator tags A proposed feature (e.g. #8022) by which the capabilities of an estimator are described through a set
of semantic tags. This would enable some runtime behaviors based on estimator inspection, but it also allows
each estimator to be tested for appropriate invariances while being excepted from other common tests.
Some aspects of estimator tags are currently determined through the duck typing of methods like
predict_proba and through some special attributes on estimator objects:
_estimator_type This string-valued attribute identifies an estimator as being a classifier, regressor, etc. It
is set by mixins such as base.ClassifierMixin, but needs to be more explicitly adopted on a meta-
estimator. Its value should usually be checked by way of a helper such as base.is_classifier.
_pairwise This boolean attribute indicates whether the data (X) passed to fit and similar methods consists
of pairwise measures over samples rather than a feature representation for each sample. It is usually
True where an estimator has a metric or affinity or kernel parameter with value ‘precomputed’.
Its primary purpose is that when a meta-estimator extracts a sub-sample of data intended for a pairwise
estimator, the data needs to be indexed on both axes, while other data is indexed only on the first axis.
feature
features
feature vector In the abstract, a feature is a function (in its mathematical sense) mapping a sampled object to a nu-
meric or categorical quantity. “Feature” is also commonly used to refer to these quantities, being the individual
elements of a vector representing a sample. In a data matrix, features are represented as columns: each column
contains the result of applying a feature function to a set of samples.
Elsewhere features are known as attributes, predictors, regressors, or independent variables.
Nearly all estimators in scikit-learn assume that features are numeric, finite and not missing, even when they
have semantically distinct domains and distributions (categorical, ordinal, count-valued, real-valued, interval).
See also categorical feature and missing values.
n_features indicates the number of features in a dataset.
fitting Calling fit (or fit_transform, fit_predict, etc.) on an estimator.
fitted The state of an estimator after fitting.
There is no conventional procedure for checking if an estimator is fitted. However, an estimator that is not fitted:
• should raise exceptions.NotFittedError when a prediction method (predict, transform, etc.) is
called. (utils.validation.check_is_fitted is used internally for this purpose.)
• should not have any attributes beginning with an alphabetic character and ending with an underscore.
(Note that a descriptor for the attribute may still be present on the class, but hasattr should return False)
function We provide ad hoc function interfaces for many algorithms, while estimator classes provide a more consis-
tent interface.
In particular, Scikit-learn may provide a function interface that fits a model to some data and returns the learnt
model parameters, as in linear_model.enet_path. For transductive models, this also returns the em-
bedding or cluster labels, as in manifold.spectral_embedding or cluster.dbscan. Many prepro-
cessing transformers also provide a function interface, akin to calling fit_transform, as in preprocessing.
maxabs_scale. Users should be careful to avoid data leakage when making use of these fit_transform-
equivalent functions.
We do not have a strict policy about when to or when not to provide function forms of estimators, but maintainers
should consider consistency with existing interfaces, and whether providing a function would lead users astray
from best practices (as regards data leakage, etc.)

gallery See examples.

hyperparameter
hyper-parameter See parameter.
impute
imputation Most machine learning algorithms require that their inputs have no missing values, and will not work
if this requirement is violated. Algorithms that attempt to fill in (or impute) missing values are referred to as
imputation algorithms.
indexable An array-like, sparse matrix, pandas DataFrame or sequence (usually a list).
induction
inductive Inductive (contrasted with transductive) machine learning builds a model of some data that can then be
applied to new instances. Most estimators in Scikit-learn are inductive, having predict and/or transform methods.
joblib A Python library (https://joblib.readthedocs.io) used in Scikit-learn to facilite simple parallelism and caching.
Joblib is oriented towards efficiently working with numpy arrays, such as through use of memory mapping. See
Parallel and distributed computing for more information.
label indicator matrix
multilabel indicator matrix
multilabel indicator matrices The format used to represent multilabel data, where each row of a 2d array or sparse
matrix corresponds to a sample, each column corresponds to a class, and each element is 1 if the sample is
labeled with the class and 0 if not.
leakage
data leakage A problem in cross validation where generalization performance can be over-estimated since knowl-
edge of the test data was inadvertently included in training a model. This is a risk, for instance, when applying
a transformer to the entirety of a dataset rather than each training portion in a cross validation split.
We aim to provide interfaces (such as pipeline and model_selection) that shield the user from data
leakage.
memmapping
memory map
memory mapping A memory efficiency strategy that keeps data on disk rather than copying it into main memory.
Memory maps can be created for arrays that can be read, written, or both, using numpy.memmap. When using
joblib to parallelize operations in Scikit-learn, it may automatically memmap large arrays to reduce memory
duplication overhead in multiprocessing.
missing values Most Scikit-learn estimators do not work with missing values. When they do (e.g. in impute.
SimpleImputer), NaN is the preferred representation of missing values in float arrays. If the array has
integer dtype, NaN cannot be represented. For this reason, we support specifying another missing_values
value when imputation or learning can be performed in integer space. Unlabeled data is a special case of missing
values in the target.
n_features The number of features.
n_outputs The number of outputs in the target.
n_samples The number of samples.
n_targets Synonym for n_outputs.
narrative docs

narrative documentation An alias for User Guide, i.e. documentation written in doc/modules/. Unlike the API
reference provided through docstrings, the User Guide aims to:
• group tools provided by Scikit-learn together thematically or in terms of usage;
• motivate why someone would use each particular tool, often through comparison;
• provide both intuitive and technical descriptions of tools;
• provide or link to examples of using key features of a tool.
np A shorthand for Numpy due to the conventional import statement:
import numpy as np
online learning Where a model is iteratively updated by receiving each batch of ground truth targets soon after
making predictions on corresponding batch of data. Intrinsically, the model must be usable for prediction after
each batch. See partial_fit.
out-of-core An efficiency strategy where not all the data is stored in main memory at once, usually by performing
learning on batches of data. See partial_fit.
outputs Individual scalar/categorical variables per sample in the target. For example, in multilabel classification each
possible label corresponds to a binary output. Also called responses, tasks or targets. See multiclass multioutput
and continuous multioutput.
pair A tuple of length two.
parameter
parameters
param
params We mostly use parameter to refer to the aspects of an estimator that can be specified in its construction. For
example, max_depth and random_state are parameters of RandomForestClassifier. Parameters
to an estimator’s constructor are stored unmodified as attributes on the estimator instance, and conventionally
start with an alphabetic character and end with an alphanumeric character. Each estimator’s constructor param-
eters are described in the estimator’s docstring.
We do not use parameters in the statistical sense, where parameters are values that specify a model and can be
estimated from data. What we call parameters might be what statisticians call hyperparameters to the model:
aspects for configuring model structure that are often not directly learnt from data. However, our parameters
are also used to prescribe modeling operations that do not affect the learnt model, such as n_jobs for controlling
parallelism.
When talking about the parameters of a meta-estimator, we may also be including the parameters of
the estimators wrapped by the meta-estimator. Ordinarily, these nested parameters are denoted by using
a double underscore (__) to separate between the estimator-as-parameter and its parameter. Thus clf
= BaggingClassifier(base_estimator=DecisionTreeClassifier(max_depth=3)) has
a deep parameter base_estimator__max_depth with value 3, which is accessible with clf.
base_estimator.max_depth or clf.get_params()['base_estimator__max_depth'].
The list of parameters and their current values can be retrieved from an estimator instance using its get_params
method.
Between construction and fitting, parameters may be modified using set_params. To enable this, parameters are
not ordinarily validated or altered when the estimator is constructed, or when each parameter is set. Parameter
validation is performed when fit is called.
Common parameters are listed below.
pairwise metric

pairwise metrics In its broad sense, a pairwise metric defines a function for measuring similarity or dissimilar-
ity between two samples (with each ordinarily represented as a feature vector). We particularly provide im-
plementations of distance metrics (as well as improper metrics like Cosine Distance) through metrics.
pairwise_distances, and of kernel functions (a constrained class of similarity functions) in metrics.
pairwise_kernels. These can compute pairwise distance matrices that are symmetric and hence store data
redundantly.
See also precomputed and metric.
Note that for most distance metrics, we rely on implementations from scipy.spatial.distance, but may
reimplement for efficiency in our context. The neighbors module also duplicates some metric implementa-
tions for integration with efficient binary tree search data structures.
pd A shorthand for Pandas due to the conventional import statement:
import pandas as pd
precomputed Where algorithms rely on pairwise metrics, and can be computed from pairwise metrics alone, we
often allow the user to specify that the X provided is already in the pairwise (dis)similarity space, rather than
in a feature space. That is, when passed to fit, it is a square, symmetric matrix, with each vector indicating
(dis)similarity to every sample, and when passed to prediction/transformation methods, each row corresponds
to a testing sample and each column to a training sample.
Use of precomputed X is usually indicated by setting a metric, affinity or kernel parameter to the
string ‘precomputed’. An estimator should mark itself as being _pairwise if this is the case.
rectangular Data that can be represented as a matrix with samples on the first axis and a fixed, finite set of features
on the second is called rectangular.
This term excludes samples with non-vectorial structure, such as text, an image of arbitrary size, a time series of
arbitrary length, a set of vectors, etc. The purpose of a vectorizer is to produce rectangular forms of such data.
sample
samples We usually use this term as a noun to indicate a single feature vector. Elsewhere a sample is called an
instance, data point, or observation. n_samples indicates the number of samples in a dataset, being the
number of rows in a data array X.
sample property
sample properties A sample property is data for each sample (e.g. an array of length n_samples) passed to an
estimator method or a similar function, alongside but distinct from the features (X) and target (y). The most
prominent example is sample_weight; see others at Data and sample properties.
As of version 0.19 we do not have a consistent approach to handling sample properties and their routing in
meta-estimators, though a fit_params parameter is often used.
scikit-learn-contrib A venue for publishing Scikit-learn-compatible libraries that are broadly authorized by the
core developers and the contrib community, but not maintained by the core developer team. See https:
//scikit-learn-contrib.github.io.
semi-supervised
semi-supervised learning
semisupervised Learning where the expected prediction (label or ground truth) is only available for some samples
provided as training data when fitting the model. We conventionally apply the label -1 to unlabeled samples in
semi-supervised classification.
sparse matrix A representation of two-dimensional numeric data that is more memory efficient the corresponding
dense numpy array where almost all elements are zero. We use the scipy.sparse framework, which provides
several underlying sparse data representations, or formats. Some formats are more efficient than others for

particular tasks, and when a particular format provides especial benefit, we try to document this fact in Scikit-
learn parameter descriptions.
Some sparse matrix formats (notably CSR, CSC, COO and LIL) distinguish between implicit and explicit zeros.
Explicit zeros are stored (i.e. they consume memory in a data array) in the data structure, while implicit zeros
correspond to every element not otherwise defined in explicit storage.
Two semantics for sparse matrices are used in Scikit-learn:
matrix semantics The sparse matrix is interpreted as an array with implicit and explicit zeros being interpreted
as the number 0. This is the interpretation most often adopted, e.g. when sparse matrices are used for
feature matrices or multilabel indicator matrices.
graph semantics As with scipy.sparse.csgraph, explicit zeros are interpreted as the number 0, but
implicit zeros indicate a masked or absent value, such as the absence of an edge between two ver-
tices of a graph, where an explicit value indicates an edge’s weight. This interpretation is adopted to
represent connectivity in clustering, in representations of nearest neighborhoods (e.g. neighbors.
kneighbors_graph), and for precomputed distance representation where only distances in the neigh-
borhood of each point are required.
When working with sparse matrices, we assume that it is sparse for a good reason, and avoid writing code that
densifies a user-provided sparse matrix, instead maintaining sparsity or raising an error if not possible (i.e. if an
estimator does not / cannot support sparse matrices).
supervised
supervised learning Learning where the expected prediction (label or ground truth) is available for each sample
when fitting the model, provided as y. This is the approach taken in a classifier or regressor among other
estimators.
target
targets The dependent variable in supervised (and semisupervised) learning, passed as y to an estimator’s fit method.
Also known as dependent variable, outcome variable, response variable, ground truth or label. Scikit-learn
works with targets that have minimal structure: a class from a finite set, a finite real-valued number, multiple
classes, or multiple numbers. See Target Types.
transduction
transductive A transductive (contrasted with inductive) machine learning method is designed to model a spe-
cific dataset, but not to apply that model to unseen data. Examples include manifold.TSNE, cluster.
AgglomerativeClustering and neighbors.LocalOutlierFactor.
unlabeled
unlabeled data Samples with an unknown ground truth when fitting; equivalently, missing values in the target. See
also semisupervised and unsupervised learning.
unsupervised
unsupervised learning Learning where the expected prediction (label or ground truth) is not available for each
sample when fitting the model, as in clusterers and outlier detectors. Unsupervised estimators ignore any y
passed to fit.
4.2 Class APIs and Estimator Types
classifier
classifiers A supervised (or semi-supervised) predictor with a finite set of discrete possible output values.
4.2. Class APIs and Estimator Types 631

A classifier supports modeling some of binary, multiclass, multilabel, or multiclass multioutput targets. Within
scikit-learn, all classifiers support multi-class classification, defaulting to using a one-vs-rest strategy over the
binary classification problem.
Classifiers must store a classes_ attribute after fitting, and usually inherit from base.ClassifierMixin,
which sets their _estimator_type attribute.
A classifier can be distinguished from other estimators with is_classifier.
A classifier must implement:
• fit
• predict
• score
It may also be appropriate to implement decision_function, predict_proba and predict_log_proba.
clusterer
clusterers A unsupervised predictor with a finite set of discrete output values.
A clusterer usually stores labels_ after fitting, and must do so if it is transductive.
A clusterer must implement:
• fit
• fit_predict if transductive
• predict if inductive
density estimator TODO
estimator
estimators An object which manages the estimation and decoding of a model. The model is estimated as a determin-
istic function of:
• parameters provided in object construction or with set_params;
• the global numpy.random random state if the estimator’s random_state parameter is set to None; and
• any data or sample properties passed to the most recent call to fit, fit_transform or fit_predict, or data
similarly passed in a sequence of calls to partial_fit.
The estimated model is stored in public and private attributes on the estimator instance, facilitating decoding
through prediction and transformation methods.
Estimators must provide a fit method, and should provide set_params and get_params, although these are usually
provided by inheritance from base.BaseEstimator.
The core functionality of some estimators may also be available as a function.
feature extractor
feature extractors A transformer which takes input where each sample is not represented as an array-like object of
fixed length, and produces an array-like object of features for each sample (and thus a 2-dimensional array-like
for a set of samples). In other words, it (lossily) maps a non-rectangular data representation into rectangular
data.
Feature extractors must implement at least:
• fit
• transform

• get_feature_names
meta-estimator
meta-estimators
metaestimator
metaestimators An estimator which takes another estimator as a parameter. Examples include pipeline.
Pipeline, model_selection.GridSearchCV , feature_selection.SelectFromModel and
ensemble.BaggingClassifier.
In a meta-estimator’s fit method, any contained estimators should be cloned before they are fit (although FIXME:
Pipeline and FeatureUnion do not do this currently). An exception to this is that an estimator may explic-
itly document that it accepts a prefitted estimator (e.g. using prefit=True in feature_selection.
SelectFromModel). One known issue with this is that the prefitted estimator will lose its model if the
meta-estimator is cloned. A meta-estimator should have fit called before prediction, even if all contained
estimators are prefitted.
In cases where a meta-estimator’s primary behaviors (e.g. predict or transform implementation) are functions
of prediction/transformation methods of the provided base estimator (or multiple base estimators), a meta-
estimator should provide at least the standard methods provided by the base estimator. It may not be possible
to identify which methods are provided by the underlying estimator until the meta-estimator has been fitted
(see also duck typing), for which utils.metaestimators.if_delegate_has_method may help. It
should also provide (or modify) the estimator tags and classes_ attribute provided by the base estimator.
Meta-estimators should be careful to validate data as minimally as possible before passing it to an underlying
estimator. This saves computation time, and may, for instance, allow the underlying estimator to easily work
with data that is not rectangular.
outlier detector
outlier detectors An unsupervised binary predictor which models the distinction between core and outlying samples.
Outlier detectors must implement:
• fit
• fit_predict if transductive
• predict if inductive
Inductive outlier detectors may also implement decision_function to give a normalized inlier score where outliers
have score below 0. score_samples may provide an unnormalized score per sample.
predictor
predictors An estimator supporting predict and/or fit_predict. This encompasses classifier, regressor, outlier detec-
tor and clusterer.
In statistics, “predictors” refers to features.
regressor
regressors A supervised (or semi-supervised) predictor with continuous output values.
Regressors usually inherit from base.RegressorMixin, which sets their _estimator_type attribute.
A regressor can be distinguished from other estimators with is_regressor.
A regressor must implement:
• fit
• predict
4.2. Class APIs and Estimator Types 633

• score
transformer
transformers An estimator supporting transform and/or fit_transform. A purely transductive transformer, such as
manifold.TSNE, may not implement transform.
vectorizer
vectorizers See feature extractor.
There are further APIs specifically related to a small family of estimators, such as:
cross-validation splitter
CV splitter
cross-validation generator A non-estimator family of classes used to split a dataset into a sequence of train and test
portions (see Cross-validation: evaluating estimator performance), by providing split and get_n_splits meth-
ods. Note that unlike estimators, these do not have fit methods and do not provide set_params or get_params.
Parameter validation may be performed in __init__.
cross-validation estimator An estimator that has built-in cross-validation capabilities to automatically select the best
hyper-parameters (see the User Guide). Some example of cross-validation estimators are ElasticNetCV
and LogisticRegressionCV . Cross-validation estimators are named EstimatorCV and tend to be roughly
equivalent to GridSearchCV(Estimator(), . . . ). The advantage of using a cross-validation estimator over the
canonical Estimator class along with grid search is that they can take advantage of warm-starting by reusing
precomputed results in the previous steps of the cross-validation process. This generally leads to speed improve-
ments. An exception is the RidgeCV class, which can instead perform efficient Leave-One-Out CV.
scorer A non-estimator callable object which evaluates an estimator on given test data, returning a number. Unlike
evaluation metrics, a greater returned number must correspond with a better score. See The scoring parameter:
defining model evaluation rules.
Further examples:
• neighbors.DistanceMetric
• gaussian_process.kernels.Kernel
• tree.Criterion
4.3 Target Types
binary A classification problem consisting of two classes. A binary target may represented as for a multiclass
problem but with only two labels. A binary decision function is represented as a 1d array.
Semantically, one class is often considered the “positive” class. Unless otherwise specified (e.g. using pos_label
in evaluation metrics), we consider the class label with the greater value (numerically or lexicographically) as
the positive class: of labels [0, 1], 1 is the positive class; of [1, 2], 2 is the positive class; of [‘no’, ‘yes’], ‘yes’
is the positive class; of [‘no’, ‘YES’], ‘no’ is the positive class. This affects the output of decision_function, for
instance.
Note that a dataset sampled from a multiclass y or a continuous y may appear to be binary.
type_of_target will return ‘binary’ for binary input, or a similar array with only a single class present.
continuous A regression problem where each sample’s target is a finite floating point number, represented as a
1-dimensional array of floats (or sometimes ints).
type_of_target will return ‘continuous’ for continuous input, but if the data is all integers, it will be
identified as ‘multiclass’.

continuous multioutput
multioutput continuous A regression problem where each sample’s target consists of n_outputs outputs, each
one a finite floating point number, for a fixed int n_outputs > 1 in a particular dataset.
Continuous multioutput targets are represented as multiple continuous targets, horizontally stacked into an array
of shape (n_samples, n_outputs).
type_of_target will return ‘continuous-multioutput’ for continuous multioutput input, but if the data is all
integers, it will be identified as ‘multiclass-multioutput’.
multiclass A classification problem consisting of more than two classes. A multiclass target may be represented as
a 1-dimensional array of strings or integers. A 2d column vector of integers (i.e. a single output in multioutput
terms) is also accepted.
We do not officially support other orderable, hashable objects as class labels, even if estimators may happen to
work when given classification targets of such type.
For semi-supervised classification, unlabeled samples should have the special label -1 in y.
Within sckit-learn, all estimators supporting binary classification also support multiclass classification, using
One-vs-Rest by default.
A preprocessing.LabelEncoder helps to canonicalize multiclass targets as integers.
type_of_target will return ‘multiclass’ for multiclass input. The user may also want to handle ‘binary’
input identically to ‘multiclass’.
multiclass multioutput
multioutput multiclass A classification problem where each sample’s target consists of n_outputs outputs, each
a class label, for a fixed int n_outputs > 1 in a particular dataset. Each output has a fixed set of available
classes, and each sample is labelled with a class for each output. An output may be binary or multiclass, and in
the case where all outputs are binary, the target is multilabel.
Multiclass multioutput targets are represented as multiple multiclass targets, horizontally stacked into an array
of shape (n_samples, n_outputs).
XXX: For simplicity, we may not always support string class labels for multiclass multioutput, and integer class
labels should be used.
multioutput provides estimators which estimate multi-output problems using multiple single-output estima-
tors. This may not fully account for dependencies among the different outputs, which methods natively handling
the multioutput case (e.g. decision trees, nearest neighbors, neural networks) may do better.
type_of_target will return ‘multiclass-multioutput’ for multiclass multioutput input.
multilabel A multiclass multioutput target where each output is binary. This may be represented as a 2d (dense)
array or sparse matrix of integers, such that each column is a separate binary target, where positive labels are
indicated with 1 and negative labels are usually -1 or 0. Sparse multilabel targets are not supported everywhere
that dense multilabel targets are supported.
Semantically, a multilabel target can be thought of as a set of labels for each sample. While not used inter-
nally, preprocessing.MultiLabelBinarizer is provided as a utility to convert from a list of sets
representation to a 2d array or sparse matrix. One-hot encoding a multiclass target with preprocessing.
LabelBinarizer turns it into a multilabel problem.
type_of_target will return ‘multilabel-indicator’ for multilabel input, whether sparse or dense.
multioutput
multi-output A target where each sample has multiple classification/regression labels. See multiclass multioutput
and continuous multioutput. We do not currently support modelling mixed classification and regression targets.
4.3. Target Types 635

4.4 Methods
decision_function In a fitted classifier or outlier detector, predicts a “soft” score for each sample in relation
to each class, rather than the “hard” categorical prediction produced by predict. Its input is usually only some
observed data, X.
If the estimator was not already fitted, calling this method should raise a exceptions.NotFittedError.
Output conventions:
binary classification A 1-dimensional array, where values strictly greater than zero indicate the positive class
(i.e. the last class in classes_).
multiclass classification A 2-dimensional array, where the row-wise arg-maximum is the predicted class.
Columns are ordered according to classes_.
multilabel classification Scikit-learn is inconsistent in its representation of multilabel decision functions.
Some estimators represent it like multiclass multioutput, i.e. a list of 2d arrays, each with two columns.
Others represent it with a single 2d array, whose columns correspond to the individual binary classification
decisions. The latter representation is ambiguously identical to the multiclass classification format, though
its semantics differ: it should be interpreted, like in the binary case, by thresholding at 0.
TODO: This gist higlights the use of the different formats for multilabel.
multioutput classification A list of 2d arrays, corresponding to each multiclass decision function.
outlier detection A 1-dimensional array, where a value greater than or equal to zero indicates an inlier.
fit The fit method is provided on every estimator. It usually takes some samples X, targets y if the model is
supervised, and potentially other sample properties such as sample_weight. It should:
• clear any prior attributes stored on the estimator, unless warm_start is used;
• validate and interpret any parameters, ideally raising an error if invalid;
• validate the input data;
• estimate and store model attributes from the estimated parameters and provided data; and
• return the now fitted estimator to facilitate method chaining.
Target Types describes possible formats for y.
fit_predict Used especially for unsupervised, transductive estimators, this fits the model and returns the pre-
dictions (similar to predict) on the training data. In clusterers, these predictions are also stored in the labels_
attribute, and the output of .fit_predict(X) is usually equivalent to .fit(X).predict(X). The pa-
rameters to fit_predict are the same as those to fit.
fit_transform A method on transformers which fits the estimator and returns the transformed training
data. It takes parameters as in fit and its output should have the same shape as calling .fit(X, ...).
transform(X). There are nonetheless rare cases where .fit_transform(X, ...) and .fit(X, ..
.).transform(X) do not return the same value, wherein training data needs to be handled differently (due
to model blending in stacked ensembles, for instance; such cases should be clearly documented). Transductive
transformers may also provide fit_transform but not transform.
One reason to implement fit_transform is that performing fit and transform separately would be less
efficient than together. base.TransformerMixin provides a default implementation, providing a consis-
tent interface across transformers where fit_transform is or is not specialised.
In inductive learning – where the goal is to learn a generalised model that can be applied to new data – users
should be careful not to apply fit_transform to the entirety of a dataset (i.e. training and test data together)
before further modelling, as this results in data leakage.

get_feature_names Primarily for feature extractors, but also used for other transformers to provide string
names for each column in the output of the estimator’s transform method. It outputs a list of strings, and may
take a list of strings as input, corresponding to the names of input columns from which output column names
can be generated. By default input features are named x0, x1, . . . .
get_n_splits On a CV splitter (not an estimator), returns the number of elements one would get if iterating
through the return value of split given the same parameters. Takes the same parameters as split.
get_params Gets all parameters, and their values, that can be set using set_params. A parameter deep can be
used, when set to False to only return those parameters not including __, i.e. not due to indirection via contained
estimators.
Most estimators adopt the definition from base.BaseEstimator, which simply adopts the parameters de-
fined for __init__. pipeline.Pipeline, among others, reimplements get_params to declare the
estimators named in its steps parameters as themselves being parameters.
partial_fit Facilitates fitting an estimator in an online fashion. Unlike fit, repeatedly calling partial_fit
does not clear the model, but updates it with respect to the data provided. The portion of data provided to
partial_fit may be called a mini-batch. Each mini-batch must be of consistent shape, etc.
partial_fit may also be used for out-of-core learning, although usually limited to the case where learning
can be performed online, i.e. the model is usable after each partial_fit and there is no separate processing
needed to finalize the model. cluster.Birch introduces the convention that calling partial_fit(X)
will produce a model that is not finalized, but the model can be finalized by calling partial_fit() i.e.
without passing a further mini-batch.
Generally, estimator parameters should not be modified between calls to partial_fit, although
partial_fit should validate them as well as the new mini-batch of data. In contrast, warm_start is
used to repeatedly fit the same estimator with the same data but varying parameters.
Like fit, partial_fit should return the estimator object.
To clear the model, a new estimator should be constructed, for instance with base.clone.
predict Makes a prediction for each sample, usually only taking X as input (but see under regressor output con-
ventions below). In a classifier or regressor, this prediction is in the same target space used in fitting (e.g. one
of {‘red’, ‘amber’, ‘green’} if the y in fitting consisted of these strings). Despite this, even when y passed to fit
is a list or other array-like, the output of predict should always be an array or sparse matrix. In a clusterer or
outlier detector the prediction is an integer.
Output conventions:
classifier An array of shape (n_samples,) (n_samples, n_outputs). Multilabel data may be rep-
resented as a sparse matrix if a sparse matrix was used in fitting. Each element should be one of the values
in the classifier’s classes_ attribute.
clusterer An array of shape (n_samples,) where each value is from 0 to n_clusters - 1 if the corre-
sponding sample is clustered, and -1 if the sample is not clustered, as in cluster.dbscan.
outlier detector An array of shape (n_samples,) where each value is -1 for an outlier and 1 otherwise.
regressor A numeric array of shape (n_samples,), usually float64. Some regressors have extra options in
their predict method, allowing them to return standard deviation (return_std=True) or covariance
(return_cov=True) relative to the predicted value. In this case, the return value is a tuple of arrays
corresponding to (prediction mean, std, cov) as required.
predict_log_proba The natural logarithm of the output of predict_proba, provided to facilitate numerical sta-
bility.
4.4. Methods 637

predict_proba A method in classifiers and clusterers that are able to return probability estimates for each
class/cluster. Its input is usually only some observed data, X.
Output conventions are like those for decision_function except in the binary classification case, where one
column is output for each class (while decision_function outputs a 1d array). For binary and multiclass
predictions, each row should add to 1.
Like other methods, predict_proba should only be present when the estimator can make probabilistic pre-
dictions (see duck typing). This means that the presence of the method may depend on estimator parameters (e.g.
in linear_model.SGDClassifier) or training data (e.g. in model_selection.GridSearchCV )
and may only appear after fitting.
score A method on an estimator, usually a predictor, which evaluates its predictions on a given dataset, and returns
a single numerical score. A greater return value should indicate better predictions; accuracy is used for classifiers
and R^2 for regressors by default.
Some estimators implement a custom, estimator-specific score function, often the likelihood of the data under
the model.
score_samples TODO
set_params Available in any estimator, takes keyword arguments corresponding to keys in get_params. Each is
provided a new value to assign such that calling get_params after set_params will reflect the changed pa-
rameters. Most estimators use the implementation in base.BaseEstimator, which handles nested parame-
ters and otherwise sets the parameter as an attribute on the estimator. The method is overridden in pipeline.
Pipeline and related estimators.
split On a CV splitter (not an estimator), this method accepts parameters (X, y, groups), where all may be op-
tional, and returns an iterator over (train_idx, test_idx) pairs. Each of {train,test}_idx is a 1d integer
array, with values from 0 from X.shape[0] - 1 of any length, such that no values appear in both some
train_idx and its corresponding test_idx.
transform In a transformer, transforms the input, usually only X, into some transformed space (conventionally
notated as Xt). Output is an array or sparse matrix of length n_samples and with number of columns fixed after
fitting.
4.5 Parameters
These common parameter names, specifically used in estimator construction (see concept parameter), sometimes also
appear as parameters of functions or non-estimator constructors.
class_weight Used to specify sample weights when fitting classifiers as a function of the target class. Where
sample_weight is also supported and given, it is multiplied by the class_weight contribution. Similarly,
where class_weight is used in a multioutput (including multilabel) tasks, the weights are multiplied across
outputs (i.e. columns of y).
By default all samples have equal weight such that classes are effectively weighted by their their prevalence in
the training data. This could be achieved explicitly with class_weight={label1: 1, label2: 1,
...} for all class labels.
More generally, class_weight is specified as a dict mapping class labels to weights ({class_label:
weight}), such that each sample of the named class is given that weight.

class_weight='balanced' can be used to give all classes equal weight by giving each sample a
weight inversely related to its class’s prevalence in the training data: n_samples / (n_classes * np.
bincount(y)). Note however that this rebalancing does not take the weight of samples in each class into
account.
For multioutput classification, a list of dicts is used to specify weights for each output. For example, for four-
class multilabel classification weights should be [{0: 1, 1: 1}, {0: 1, 1: 5}, {0: 1,
1: 1}, {0: 1, 1: 1}] instead of [{1:1}, {2:5}, {3:1}, {4:1}].
The class_weight parameter is validated and interpreted with utils.compute_class_weight.
cv Determines a cross validation splitting strategy, as used in cross-validation based routines. cv
is also available in estimators such as multioutput.ClassifierChain or calibration.
CalibratedClassifierCV which use the predictions of one estimator as training data for another, to
not overfit the training supervision.
Possible inputs for cv are usually:
• An integer, specifying the number of folds in K-fold cross validation. K-fold will be stratified over classes
if the estimator is a classifier (determined by base.is_classifier) and the targets may represent a
binary or multiclass (but not multioutput) classification problem (determined by utils.multiclass.
type_of_target).
• A cross-validation splitter instance. Refer to the User Guide for splitters available within Scikit-learn.
• An iterable yielding train/test splits.
With some exceptions (especially where not using cross validation at all is an option), the default is 3-fold and
will change to 5-fold in version 0.22.
cv values are validated and interpreted with utils.check_cv.
kernel TODO
max_iter For estimators involving iterative optimization, this determines the maximum number of itera-
tions to be performed in fit. If max_iter iterations are run without convergence, a exceptions.
ConvergenceWarning should be raised. Note that the interpretation of “a single iteration” is inconsistent
across estimators: some, but not all, use it to mean a single epoch (i.e. a pass over every sample in the data).
FIXME perhaps we should have some common tests about the relationship between ConvergenceWarning and
max_iter.
memory Some estimators make use of joblib.Memory to store partial solutions during fitting. Thus when fit
is called again, those partial solutions have been memoized and can be reused.
A memory parameter can be specified as a string with a path to a directory, or a joblib.Memory instance
(or an object with a similar interface, i.e. a cache method) can be used.
memory values are validated and interpreted with utils.validation.check_memory.
metric As a parameter, this is the scheme for determining the distance between two data points. See metrics.
pairwise_distances. In practice, for some algorithms, an improper distance metric (one that does not
obey the triangle inequality, such as Cosine Distance) may be used.
XXX: hierarchical clustering uses affinity with this meaning.
We also use metric to refer to evaluation metrics, but avoid using this sense as a parameter name.
n_components The number of features which a transformer should transform the input into. See components_ for
the special case of affine projection.
n_iter_no_change Number of iterations with no improvement to wait before stopping the iterative procedure.
This is also known as a patience parameter. It is typically used with early stopping to avoid stopping too early.
4.5. Parameters 639

n_jobs This is used to specify how many concurrent processes/threads should be used for parallelized routines.
Scikit-learn uses one processor for its processing by default, although it also makes use of NumPy, which may
be configured to use a threaded numerical processor library (like MKL; see FAQ).
n_jobs is an int, specifying the maximum number of concurrently running jobs. If set to -1, all CPUs are used.
If 1 is given, no joblib level parallelism is used at all, which is useful for debugging. Even with n_jobs =
1, parallelism may occur due to numerical processing libraries (see FAQ). For n_jobs below -1, (n_cpus + 1 +
n_jobs) are used. Thus for n_jobs = -2, all CPUs but one are used.
n_jobs=None means unset; it will generally be interpreted as n_jobs=1, unless the current joblib.
Parallel backend context specifies otherwise.
The use of n_jobs-based parallelism in estimators varies:
• Most often parallelism happens in fitting, but sometimes parallelism happens in prediction (e.g. in random
forests).
• Some parallelism uses a multi-threading backend by default, some a multi-processing backend. It is pos-
sible to override the default backend by using sklearn.utils.parallel_backend.
• Whether parallel processing is helpful at improving runtime depends on many factors, and it’s usually a
good idea to experiment rather than assuming that increasing the number of jobs is always a good thing.
It can be highly detrimental to performance to run multiple copies of some estimators or functions in
parallel.
Nested uses of n_jobs-based parallelism with the same backend will result in an exception. So
GridSearchCV(OneVsRestClassifier(SVC(), n_jobs=2), n_jobs=2) won’t work.
When n_jobs is not 1, the estimator being parallelized must be picklable. This means, for instance, that
lambdas cannot be used as estimator parameters.
random_state Whenever randomization is part of a Scikit-learn algorithm, a random_state parameter may
be provided to control the random number generator used. Note that the mere presence of random_state
doesn’t mean that randomization is always used, as it may be dependent on another parameter, e.g. shuffle,
being set.
random_state’s value may be:
None (default) Use the global random state from numpy.random.
An integer Use a new random number generator seeded by the given integer. To make a randomized al-
gorithm deterministic (i.e. running it multiple times will produce the same result), an arbitrary integer
random_state can be used. However, it may be worthwhile checking that your results are stable across
a number of different distinct random seeds. Popular integer random seeds are 0 and 42.
A numpy.random.RandomState instance Use the provided random state, only affecting other users of
the same random state instance. Calling fit multiple times will reuse the same instance, and will produce
different results.
utils.check_random_state is used internally to validate the input random_state and return a
RandomState instance.
scoring Specifies the score function to be maximized (usually by cross validation), or – in some cases – multiple
score functions to be reported. The score function can be a string accepted by metrics.get_scorer or a
callable scorer, not to be confused with an evaluation metric, as the latter have a more diverse API. scoring
may also be set to None, in which case the estimator’s score method is used. See The scoring parameter:
defining model evaluation rules in the User Guide.
Where multiple metrics can be evaluated, scoring may be given either as a list of unique strings or a dict with
names as keys and callables as values. Note that this does not specify which score function is to be maximised,
and another parameter such as refit may be used for this purpose.

The scoring parameter is validated and interpreted using metrics.check_scoring.

verbose Logging is not handled very consistently in Scikit-learn at present, but when it is provided as an option,
the verbose parameter is usually available to choose no logging (set to False). Any True value should enable
some logging, but larger integers (e.g. above 10) may be needed for full verbosity. Verbose logs are usually
printed to Standard Output. Estimators should not produce any output on Standard Output with the default
verbose setting.
warm_start When fitting an estimator repeatedly on the same dataset, but for multiple parameter values (such as
to find the value maximizing performance as in grid search), it may be possible to reuse aspects of the model
learnt from the previous parameter value, saving time. When warm_start is true, the existing fitted model
attributes are used to initialise the new model in a subsequent call to fit.
Note that this is only applicable for some models and some parameters, and even some orders of parameter
values. For example, warm_start may be used when building random forests to add more trees to the forest
(increasing n_estimators) but not to reduce their number.
partial_fit also retains the model between calls, but differs: with warm_start the parameters change and the
data is (more-or-less) constant across calls to fit; with partial_fit, the mini-batch of data changes and
model parameters stay fixed.
There are cases where you want to use warm_start to fit on different, but closely related data. For exam-
ple, one may initially fit to a subset of the data, then fine-tune the parameter search on the full dataset. For
classification, all data in a sequence of warm_start calls to fit must include samples from each class.
4.6 Attributes
See concept attribute.

classes_ A list of class labels known to the classifier, mapping each label to a numerical index used in the
model representation our output. For instance, the array output from predict_proba has columns aligned with
classes_. For multi-output classifiers, classes_ should be a list of lists, with one class listing for each
output. For each output, the classes should be sorted (numerically, or lexicographically for strings).
classes_ and the mapping to indices is often managed with preprocessing.LabelEncoder.
components_ An affine transformation matrix of shape (n_components, n_features) used in many lin-
ear transformers where n_components is the number of output features and n_features is the number of input
features.
See also components_ which is a similar attribute for linear predictors.
coef_ The weight/coefficient matrix of a generalised linear model predictor, of shape (n_features,) for binary
classification and single-output regression, (n_classes, n_features) for multiclass classification and
(n_targets, n_features) for multi-output regression. Note this does not include the intercept (or bias)
term, which is stored in intercept_.
When available, feature_importances_ is not usually provided as well, but can be calculated as the norm
of each feature’s entry in coef_.
See also components_ which is a similar attribute for linear transformers.
embedding_ An embedding of the training data in manifold learning estimators, with shape (n_samples,
n_components), identical to the output of fit_transform. See also labels_.
n_iter_ The number of iterations actually performed when fitting an iterative estimator that may stop upon con-
vergence. See also max_iter.
feature_importances_ A vector of shape (n_features,) available in some predictors to provide a relative
measure of the importance of each feature in the predictions of the model.
4.6. Attributes 641

labels_ A vector containing a cluster label for each sample of the training data in clusterers, identical to the output
of fit_predict. See also embedding_.
4.7 Data and sample properties
See concept sample property.

groups Used in cross validation routines to identify samples which are correlated. Each value is an identifier such
that, in a supporting CV splitter, samples from some groups value may not appear in both a training set and
its corresponding test set. See Cross-validation iterators for grouped data..
sample_weight A relative weight for each sample. Intuitively, if all weights are integers, a weighted model or
score should be equivalent to that calculated when repeating the sample the number of times specified in the
weight. Weights may be specified as floats, so that sample weights are usually equivalent up to a constant
positive scaling factor.
FIXME Is this interpretation always the case in practice? We have no common tests.
Some estimators, such as decision trees, support negative weights. FIXME: This feature or its absence may not
be tested or documented in many estimators.
This is not entirely the case where other parameters of the model consider the number of samples in a region,
as with min_samples in cluster.DBSCAN . In this case, a count of samples becomes to a sum of their
weights.
In classification, sample weights can also be specified as a function of class with the class_weight estimator
parameter.
X Denotes data that is observed at training and prediction time, used as independent variables in learning. The
notation is uppercase to denote that it is ordinarily a matrix (see rectangular). When a matrix, each sample may
be represented by a feature vector, or a vector of precomputed (dis)similarity with each training sample. X may
also not be a matrix, and may require a feature extractor or a pairwise metric to turn it into one before learning
a model.
Xt Shorthand for “transformed X”.
y
Y Denotes data that may be observed at training time as the dependent variable in learning, but which is unavailable
at prediction time, and is usually the target of prediction. The notation may be uppercase to denote that it is
a matrix, representing multi-output targets, for instance; but usually we use y and sometimes do so even when
multiple outputs are assumed.

CHAPTER
FIVE
EXAMPLES
5.1 Miscellaneous examples
Miscellaneous and introductory examples for scikit-learn.
Note: Click here to download the full example code
5.1.1 Isotonic Regression
An illustration of the isotonic regression on generated data. The isotonic regression finds a non-decreasing approx-
imation of a function while minimizing the mean squared error on the training data. The benefit of such a model is
that it does not assume any form for the target function such as linearity. For comparison a linear regression is also
presented.
643
print(__doc__)
# Author: Nelle Varoquaux <nelle.varoquaux@gmail.com>

# Alexandre Gramfort <alexandre.gramfort@inria.fr>
# License: BSD
import numpy as np
from matplotlib.collections import LineCollection
from sklearn.linear_model import LinearRegression

from sklearn.isotonic import IsotonicRegression
from sklearn.utils import check_random_state
n = 100
x = np.arange(n)
rs = check_random_state(0)
y = rs.randint(-50, 50, size=(n,)) + 50. * np.log1p(np.arange(n))
# #############################################################################
# Fit IsotonicRegression and LinearRegression models
ir = IsotonicRegression()
y_ = ir.fit_transform(x, y)
644 Chapter 5. Examples

lr = LinearRegression()
lr.fit(x[:, np.newaxis], y) # x needs to be 2d for LinearRegression
# #############################################################################
# Plot result
segments = [[[i, y[i]], [i, y_[i]]] for i in range(n)]

lc = LineCollection(segments, zorder=0)
lc.set_array(np.ones(len(y)))
lc.set_linewidths(np.full(n, 0.5))
fig = plt.figure()
plt.plot(x, y, 'r.', markersize=12)
plt.plot(x, y_, 'g.-', markersize=12)
plt.plot(x, lr.predict(x[:, np.newaxis]), 'b-')
plt.gca().add_collection(lc)
plt.legend(('Data', 'Isotonic Fit', 'Linear Fit'), loc='lower right')
plt.title('Isotonic regression')
plt.show()
Total running time of the script: ( 0 minutes 0.113 seconds)
5.1.2 Face completion with a multi-output estimators
This example shows the use of multi-output estimator to complete images. The goal is to predict the lower half of a
face given its upper half.
The first column of images shows true faces. The next columns illustrate how extremely randomized trees, k nearest
neighbors, linear regression and ridge regression complete the lower half of those faces.
5.1. Miscellaneous examples 645

Out:
downloading Olivetti faces from https://ndownloader.figshare.com/files/5976027 to /

˓→home/circleci/scikit_learn_data

print(__doc__)
import numpy as np
from sklearn.datasets import fetch_olivetti_faces

from sklearn.utils.validation import check_random_state
from sklearn.ensemble import ExtraTreesRegressor

from sklearn.neighbors import KNeighborsRegressor
from sklearn.linear_model import RidgeCV
# Load the faces datasets

data = fetch_olivetti_faces()
targets = data.target
data = data.images.reshape((len(data.images), -1))

train = data[targets < 30]
test = data[targets >= 30] # Test on independent people
# Test on a subset of people

n_faces = 5
rng = check_random_state(4)
face_ids = rng.randint(test.shape[0], size=(n_faces, ))
test = test[face_ids, :]
n_pixels = data.shape[1]
# Upper half of the faces
X_train = train[:, :(n_pixels + 1) // 2]
# Lower half of the faces
y_train = train[:, n_pixels // 2:]
X_test = test[:, :(n_pixels + 1) // 2]
y_test = test[:, n_pixels // 2:]
# Fit estimators
ESTIMATORS = {
"Extra trees": ExtraTreesRegressor(n_estimators=10, max_features=32,
random_state=0),
"K-nn": KNeighborsRegressor(),
"Linear regression": LinearRegression(),
"Ridge": RidgeCV(),
}
y_test_predict = dict()
for name, estimator in ESTIMATORS.items():
estimator.fit(X_train, y_train)
y_test_predict[name] = estimator.predict(X_test)
# Plot the completed faces

image_shape = (64, 64)
n_cols = 1 + len(ESTIMATORS)
plt.figure(figsize=(2. * n_cols, 2.26 * n_faces))
plt.suptitle("Face completion with multi-output estimators", size=16)
for i in range(n_faces):
true_face = np.hstack((X_test[i], y_test[i]))

if i:
sub = plt.subplot(n_faces, n_cols, i * n_cols + 1)
else:
sub = plt.subplot(n_faces, n_cols, i * n_cols + 1,
title="true faces")
sub.axis("off")
sub.imshow(true_face.reshape(image_shape),
cmap=plt.cm.gray,
interpolation="nearest")
for j, est in enumerate(sorted(ESTIMATORS)):

completed_face = np.hstack((X_test[i], y_test_predict[est][i]))
if i:
sub = plt.subplot(n_faces, n_cols, i * n_cols + 2 + j)
else:
sub = plt.subplot(n_faces, n_cols, i * n_cols + 2 + j,
title=est)
sub.axis("off")
sub.imshow(completed_face.reshape(image_shape),
cmap=plt.cm.gray,
interpolation="nearest")
plt.show()
5.1.3 Multilabel classification
This example simulates a multi-label document classification problem. The dataset is generated randomly based on
the following process:
• pick the number of labels: n ~ Poisson(n_labels)
• n times, choose a class c: c ~ Multinomial(theta)
• pick the document length: k ~ Poisson(length)
• k times, choose a word: w ~ Multinomial(theta_c)
In the above process, rejection sampling is used to make sure that n is more than 2, and that the document length is
never zero. Likewise, we reject classes which have already been chosen. The documents that are assigned to both
classes are plotted surrounded by two colored circles.
The classification is performed by projecting to the first two principal components found by PCA and CCA for visual-
isation purposes, followed by using the sklearn.multiclass.OneVsRestClassifier metaclassifier using
two SVCs with linear kernels to learn a discriminative model for each class. Note that PCA is used to perform an
unsupervised dimensionality reduction, while CCA is used to perform a supervised one.
Note: in the plot, “unlabeled samples” does not mean that we don’t know the labels (as in semi-supervised learning)
but that the samples simply do not have a label.

print(__doc__)
import numpy as np
from sklearn.datasets import make_multilabel_classification

from sklearn.multiclass import OneVsRestClassifier
from sklearn.cross_decomposition import CCA
def plot_hyperplane(clf, min_x, max_x, linestyle, label):

# get the separating hyperplane
w = clf.coef_[0]
a = -w[0] / w[1]
xx = np.linspace(min_x - 5, max_x + 5) # make sure the line is long enough
yy = a * xx - (clf.intercept_[0]) / w[1]
plt.plot(xx, yy, linestyle, label=label)
def plot_subfigure(X, Y, subplot, title, transform):

if transform == "pca":
X = PCA(n_components=2).fit_transform(X)
elif transform == "cca":
X = CCA(n_components=2).fit(X, Y).transform(X)

else:
raise ValueError
min_x = np.min(X[:, 0])

max_x = np.max(X[:, 0])
min_y = np.min(X[:, 1])

max_y = np.max(X[:, 1])
classif = OneVsRestClassifier(SVC(kernel='linear'))
classif.fit(X, Y)
plt.subplot(2, 2, subplot)
plt.title(title)
zero_class = np.where(Y[:, 0])

one_class = np.where(Y[:, 1])
plt.scatter(X[:, 0], X[:, 1], s=40, c='gray', edgecolors=(0, 0, 0))
plt.scatter(X[zero_class, 0], X[zero_class, 1], s=160, edgecolors='b',
facecolors='none', linewidths=2, label='Class 1')
plt.scatter(X[one_class, 0], X[one_class, 1], s=80, edgecolors='orange',
facecolors='none', linewidths=2, label='Class 2')
plot_hyperplane(classif.estimators_[0], min_x, max_x, 'k--',

'Boundary\nfor class 1')
plot_hyperplane(classif.estimators_[1], min_x, max_x, 'k-.',
'Boundary\nfor class 2')
plt.xticks(())
plt.yticks(())
plt.xlim(min_x - .5 * max_x, max_x + .5 * max_x)

plt.ylim(min_y - .5 * max_y, max_y + .5 * max_y)
if subplot == 2:
plt.xlabel('First principal component')
plt.ylabel('Second principal component')
plt.legend(loc="upper left")
plt.figure(figsize=(8, 6))
X, Y = make_multilabel_classification(n_classes=2, n_labels=1,
allow_unlabeled=True,
random_state=1)
plot_subfigure(X, Y, 1, "With unlabeled samples + CCA", "cca")

plot_subfigure(X, Y, 2, "With unlabeled samples + PCA", "pca")
X, Y = make_multilabel_classification(n_classes=2, n_labels=1,
allow_unlabeled=False,
random_state=1)
plot_subfigure(X, Y, 3, "Without unlabeled samples + CCA", "cca")

plot_subfigure(X, Y, 4, "Without unlabeled samples + PCA", "pca")
plt.subplots_adjust(.04, .02, .97, .94, .09, .2)

plt.show()

5.1.4 Comparing anomaly detection algorithms for outlier detection on toy datasets
This example shows characteristics of different anomaly detection algorithms on 2D datasets. Datasets contain one or
two modes (regions of high density) to illustrate the ability of algorithms to cope with multimodal data.
For each dataset, 15% of samples are generated as random uniform noise. This proportion is the value given to the nu
parameter of the OneClassSVM and the contamination parameter of the other outlier detection algorithms. Decision
boundaries between inliers and outliers are displayed in black except for Local Outlier Factor (LOF) as it has no
predict method to be applied on new data when it is used for outlier detection.
The sklearn.svm.OneClassSVM is known to be sensitive to outliers and thus does not perform very well for
outlier detection. This estimator is best suited for novelty detection when the training set is not contaminated by
outliers. That said, outlier detection in high-dimension, or without any assumptions on the distribution of the inlying
data is very challenging, and a One-class SVM might give useful results in these situations depending on the value of
its hyperparameters.
sklearn.covariance.EllipticEnvelope assumes the data is Gaussian and learns an ellipse. It thus de-
grades when the data is not unimodal. Notice however that this estimator is robust to outliers.
sklearn.ensemble.IsolationForest and sklearn.neighbors.LocalOutlierFactor seem
to perform reasonably well for multi-modal data sets. The advantage of sklearn.neighbors.
LocalOutlierFactor over the other estimators is shown for the third data set, where the two modes have different
densities. This advantage is explained by the local aspect of LOF, meaning that it only compares the score of abnor-
mality of one sample with the scores of its neighbors.
Finally, for the last data set, it is hard to say that one sample is more abnormal than another sample as they are
uniformly distributed in a hypercube. Except for the sklearn.svm.OneClassSVM which overfits a little, all
estimators present decent solutions for this situation. In such a case, it would be wise to look more closely at the
scores of abnormality of the samples as a good estimator should assign similar scores to all the samples.
While these examples give some intuition about the algorithms, this intuition might not apply to very high dimensional
data.
Finally, note that parameters of the models have been here handpicked but that in practice they need to be adjusted. In
the absence of labelled data, the problem is completely unsupervised so model selection can be a challenge.

# Author: Alexandre Gramfort <alexandre.gramfort@inria.fr>

# Albert Thomas <albert.thomas@telecom-paristech.fr>
# License: BSD 3 clause
import time
import numpy as np
import matplotlib

from sklearn import svm

from sklearn.datasets import make_moons, make_blobs
from sklearn.covariance import EllipticEnvelope
from sklearn.ensemble import IsolationForest
from sklearn.neighbors import LocalOutlierFactor
print(__doc__)
matplotlib.rcParams['contour.negative_linestyle'] = 'solid'
# Example settings
n_samples = 300
outliers_fraction = 0.15
n_outliers = int(outliers_fraction * n_samples)
n_inliers = n_samples - n_outliers
# define outlier/anomaly detection methods to be compared

anomaly_algorithms = [
("Robust covariance", EllipticEnvelope(contamination=outliers_fraction)),
("One-Class SVM", svm.OneClassSVM(nu=outliers_fraction, kernel="rbf",
gamma=0.1)),
("Isolation Forest", IsolationForest(behaviour='new',
contamination=outliers_fraction,
random_state=42)),
("Local Outlier Factor", LocalOutlierFactor(
n_neighbors=35, contamination=outliers_fraction))]
# Define datasets
blobs_params = dict(random_state=0, n_samples=n_inliers, n_features=2)
datasets = [
make_blobs(centers=[[0, 0], [0, 0]], cluster_std=0.5,
**blobs_params)[0],
make_blobs(centers=[[2, 2], [-2, -2]], cluster_std=[0.5, 0.5],
**blobs_params)[0],
make_blobs(centers=[[2, 2], [-2, -2]], cluster_std=[1.5, .3],
**blobs_params)[0],
4. * (make_moons(n_samples=n_samples, noise=.05, random_state=0)[0] -
np.array([0.5, 0.25])),
14. * (np.random.RandomState(42).rand(n_samples, 2) - 0.5)]
# Compare given classifiers under given settings

xx, yy = np.meshgrid(np.linspace(-7, 7, 150),
np.linspace(-7, 7, 150))
plt.figure(figsize=(len(anomaly_algorithms) * 2 + 3, 12.5))
plt.subplots_adjust(left=.02, right=.98, bottom=.001, top=.96, wspace=.05,
hspace=.01)
plot_num = 1
rng = np.random.RandomState(42)
for i_dataset, X in enumerate(datasets):

# Add outliers
X = np.concatenate([X, rng.uniform(low=-6, high=6,
size=(n_outliers, 2))], axis=0)
for name, algorithm in anomaly_algorithms:

t0 = time.time()
algorithm.fit(X)
t1 = time.time()
plt.subplot(len(datasets), len(anomaly_algorithms), plot_num)
if i_dataset == 0:
plt.title(name, size=18)
# fit the data and tag outliers

if name == "Local Outlier Factor":
y_pred = algorithm.fit_predict(X)
else:
y_pred = algorithm.fit(X).predict(X)
# plot the levels lines and the points

if name != "Local Outlier Factor": # LOF does not implement predict
Z = algorithm.predict(np.c_[xx.ravel(), yy.ravel()])
Z = Z.reshape(xx.shape)
plt.contour(xx, yy, Z, levels=[0], linewidths=2, colors='black')
colors = np.array(['#377eb8', '#ff7f00'])

plt.scatter(X[:, 0], X[:, 1], s=10, color=colors[(y_pred + 1) // 2])
plt.xlim(-7, 7)
plt.ylim(-7, 7)
plt.xticks(())
plt.yticks(())
plt.text(.99, .01, ('%.2fs' % (t1 - t0)).lstrip('0'),
transform=plt.gca().transAxes, size=15,
horizontalalignment='right')
plot_num += 1
plt.show()
5.1.5 Imputing missing values before building an estimator
This example shows that imputing the missing values can give better results than discarding the samples containing
any missing value. Imputing does not always improve the predictions, so please check via cross-validation. Sometimes
dropping rows or using marker values is more effective.
Missing values can be replaced by the mean, the median or the most frequent value using the basic sklearn.
impute.SimpleImputer. The median is a more robust estimator for data with high magnitude variables which
could dominate results (otherwise known as a ‘long tail’).
In addition of using an imputing method, we can also keep an indication of the missing information using sklearn.
impute.MissingIndicator which might carry some information.

import numpy as np
from sklearn.datasets import load_diabetes

from sklearn.datasets import load_boston
from sklearn.ensemble import RandomForestRegressor
from sklearn.pipeline import make_pipeline, make_union
from sklearn.impute import SimpleImputer, MissingIndicator
def get_results(dataset):
X_full, y_full = dataset.data, dataset.target
n_samples = X_full.shape[0]
n_features = X_full.shape[1]
# Estimate the score on the entire dataset, with no missing values

estimator = RandomForestRegressor(random_state=0, n_estimators=100)
full_scores = cross_val_score(estimator, X_full, y_full,
scoring='neg_mean_squared_error', cv=5)
# Add missing values in 75% of the lines

missing_rate = 0.75
n_missing_samples = int(np.floor(n_samples * missing_rate))
missing_samples = np.hstack((np.zeros(n_samples - n_missing_samples,
dtype=np.bool),
np.ones(n_missing_samples,
dtype=np.bool)))
rng.shuffle(missing_samples)
missing_features = rng.randint(0, n_features, n_missing_samples)
# Estimate the score after replacing missing values by 0

X_missing = X_full.copy()
X_missing[np.where(missing_samples)[0], missing_features] = 0

y_missing = y_full.copy()
estimator = RandomForestRegressor(random_state=0, n_estimators=100)
zero_impute_scores = cross_val_score(estimator, X_missing, y_missing,
scoring='neg_mean_squared_error',
cv=5)
# Estimate the score after imputation (mean strategy) of the missing values
X_missing = X_full.copy()
X_missing[np.where(missing_samples)[0], missing_features] = 0
y_missing = y_full.copy()
estimator = make_pipeline(
make_union(SimpleImputer(missing_values=0, strategy="mean"),
MissingIndicator(missing_values=0)),
RandomForestRegressor(random_state=0, n_estimators=100))
mean_impute_scores = cross_val_score(estimator, X_missing, y_missing,
scoring='neg_mean_squared_error',
cv=5)
return ((full_scores.mean(), full_scores.std()),

(zero_impute_scores.mean(), zero_impute_scores.std()),
(mean_impute_scores.mean(), mean_impute_scores.std()))
results_diabetes = np.array(get_results(load_diabetes()))
mses_diabetes = results_diabetes[:, 0] * -1
stds_diabetes = results_diabetes[:, 1]
results_boston = np.array(get_results(load_boston()))
mses_boston = results_boston[:, 0] * -1
stds_boston = results_boston[:, 1]
n_bars = len(mses_diabetes)
xval = np.arange(n_bars)
x_labels = ['Full data',

'Zero imputation',
'Mean Imputation']
colors = ['r', 'g', 'b', 'orange']
# plot diabetes results

ax1 = plt.subplot(121)
for j in xval:
ax1.barh(j, mses_diabetes[j], xerr=stds_diabetes[j],
color=colors[j], alpha=0.6, align='center')
ax1.set_title('Imputation Techniques with Diabetes Data')

ax1.set_xlim(left=np.min(mses_diabetes) * 0.9,
right=np.max(mses_diabetes) * 1.1)
ax1.set_yticks(xval)
ax1.set_xlabel('MSE')
ax1.invert_yaxis()
ax1.set_yticklabels(x_labels)
# plot boston results

ax2 = plt.subplot(122)
for j in xval:

ax2.barh(j, mses_boston[j], xerr=stds_boston[j],

color=colors[j], alpha=0.6, align='center')
ax2.set_title('Imputation Techniques with Boston Data')

ax2.set_yticks(xval)
ax2.set_xlabel('MSE')
ax2.invert_yaxis()
ax2.set_yticklabels([''] * n_bars)
plt.show()
5.1.6 The Johnson-Lindenstrauss bound for embedding with random projections
The Johnson-Lindenstrauss lemma states that any high dimensional dataset can be randomly projected into a lower
dimensional Euclidean space while controlling the distortion in the pairwise distances.
Theoretical bounds
The distortion introduced by a random projection p is asserted by the fact that p is defining an eps-embedding with
good probability as defined by:
(1 − 𝑒𝑝𝑠)‖𝑢 − 𝑣‖2 < ‖𝑝(𝑢) − 𝑝(𝑣)‖2 < (1 + 𝑒𝑝𝑠)‖𝑢 − 𝑣‖2
Where u and v are any rows taken from a dataset of shape [n_samples, n_features] and p is a projection by a random
Gaussian N(0, 1) matrix with shape [n_components, n_features] (or a sparse Achlioptas matrix).
The minimum number of components to guarantees the eps-embedding is given by:
𝑛_𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠 >= 4𝑙𝑜𝑔(𝑛_𝑠𝑎𝑚𝑝𝑙𝑒𝑠)/(𝑒𝑝𝑠2 /2 − 𝑒𝑝𝑠3 /3)
The first plot shows that with an increasing number of samples n_samples, the minimal number of dimensions
n_components increased logarithmically in order to guarantee an eps-embedding.
The second plot shows that an increase of the admissible distortion eps allows to reduce drastically the minimal
number of dimensions n_components for a given number of samples n_samples
Empirical validation
We validate the above bounds on the digits dataset or on the 20 newsgroups text document (TF-IDF word frequencies)
dataset:
• for the digits dataset, some 8x8 gray level pixels data for 500 handwritten digits pictures are randomly projected
to spaces for various larger number of dimensions n_components.
• for the 20 newsgroups dataset some 500 documents with 100k features in total are projected using a
sparse random matrix to smaller euclidean spaces with various values for the target number of dimensions
n_components.

The default dataset is the digits dataset. To run the example on the twenty newsgroups dataset, pass the –twenty-
newsgroups command line argument to this script.
For each value of n_components, we plot:
• 2D distribution of sample pairs with pairwise distances in original and projected spaces as x and y axis respec-
tively.
• 1D histogram of the ratio of those distances (projected / original).
We can see that for low values of n_components the distribution is wide with many distorted pairs and a skewed
distribution (due to the hard limit of zero ratio on the left as distances are always positives) while for larger values of
n_components the distortion is controlled and the distances are well preserved by the random projection.
Remarks
According to the JL lemma, projecting 500 samples without too much distortion will require at least several thousands
dimensions, irrespective of the number of features of the original dataset.
Hence using random projections on the digits dataset which only has 64 features in the input space does not make
sense: it does not allow for dimensionality reduction in this case.
On the twenty newsgroups on the other hand the dimensionality can be decreased from 56436 down to 10000 while
reasonably preserving pairwise distances.







•
Out:
Embedding 500 samples with dim 64 using various random projections

Projected 500 samples from 64 to 300 in 0.007s
Random matrix with size: 0.029MB
Mean distances rate: 1.02 (0.08)
print(__doc__)
import sys
import numpy as np
import matplotlib
from distutils.version import LooseVersion
from sklearn.random_projection import johnson_lindenstrauss_min_dim

from sklearn.random_projection import SparseRandomProjection

from sklearn.datasets import fetch_20newsgroups_vectorized
from sklearn.datasets import load_digits
from sklearn.metrics.pairwise import euclidean_distances
# `normed` is being deprecated in favor of `density` in histograms

if LooseVersion(matplotlib.__version__) >= '2.1':
density_param = {'density': True}
else:
density_param = {'normed': True}
# Part 1: plot the theoretical dependency between n_components_min and

# n_samples
# range of admissible distortions

eps_range = np.linspace(0.1, 0.99, 5)
colors = plt.cm.Blues(np.linspace(0.3, 1.0, len(eps_range)))
# range of number of samples (observation) to embed

n_samples_range = np.logspace(1, 9, 9)
plt.figure()
for eps, color in zip(eps_range, colors):
min_n_components = johnson_lindenstrauss_min_dim(n_samples_range, eps=eps)
plt.loglog(n_samples_range, min_n_components, color=color)
plt.legend(["eps = %0.1f" % eps for eps in eps_range], loc="lower right")

plt.xlabel("Number of observations to eps-embed")
plt.ylabel("Minimum number of dimensions")
plt.title("Johnson-Lindenstrauss bounds:\nn_samples vs n_components")
# range of admissible distortions

eps_range = np.linspace(0.01, 0.99, 100)
# range of number of samples (observation) to embed

n_samples_range = np.logspace(2, 6, 5)
colors = plt.cm.Blues(np.linspace(0.3, 1.0, len(n_samples_range)))
plt.figure()
for n_samples, color in zip(n_samples_range, colors):
min_n_components = johnson_lindenstrauss_min_dim(n_samples, eps=eps_range)
plt.semilogy(eps_range, min_n_components, color=color)
plt.legend(["n_samples = %d" % n for n in n_samples_range], loc="upper right")

plt.xlabel("Distortion eps")
plt.ylabel("Minimum number of dimensions")
plt.title("Johnson-Lindenstrauss bounds:\nn_components vs eps")
# Part 2: perform sparse random projection of some digits images which are
# quite low dimensional and dense or documents of the 20 newsgroups dataset
# which is both high dimensional and sparse
if '--twenty-newsgroups' in sys.argv:
# Need an internet connection hence not enabled by default
data = fetch_20newsgroups_vectorized().data[:500]
else:
data = load_digits().data[:500]

n_samples, n_features = data.shape

print("Embedding %d samples with dim %d using various random projections"
% (n_samples, n_features))
n_components_range = np.array([300, 1000, 10000])

dists = euclidean_distances(data, squared=True).ravel()
# select only non-identical samples pairs

nonzero = dists != 0
dists = dists[nonzero]
for n_components in n_components_range:

t0 = time()
rp = SparseRandomProjection(n_components=n_components)
projected_data = rp.fit_transform(data)
print("Projected %d samples from %d to %d in %0.3fs"
% (n_samples, n_features, n_components, time() - t0))
if hasattr(rp, 'components_'):
n_bytes = rp.components_.data.nbytes
n_bytes += rp.components_.indices.nbytes
print("Random matrix with size: %0.3fMB" % (n_bytes / 1e6))
projected_dists = euclidean_distances(
projected_data, squared=True).ravel()[nonzero]
plt.figure()
plt.hexbin(dists, projected_dists, gridsize=100, cmap=plt.cm.PuBu)
plt.xlabel("Pairwise squared distances in original space")
plt.ylabel("Pairwise squared distances in projected space")
plt.title("Pairwise distances distribution for n_components=%d" %
n_components)
cb = plt.colorbar()
cb.set_label('Sample pairs counts')
rates = projected_dists / dists

print("Mean distances rate: %0.2f (%0.2f)"
% (np.mean(rates), np.std(rates)))
plt.figure()
plt.hist(rates, bins=50, range=(0., 2.), edgecolor='k', **density_param)
plt.xlabel("Squared distances rate: projected / original")
plt.ylabel("Distribution of samples pairs")
plt.title("Histogram of pairwise distance rates for n_components=%d" %
n_components)
# TODO: compute the expected value of eps and add them to the previous plot
# as vertical lines / region
plt.show()

5.1.7 Comparison of kernel ridge regression and SVR
Both kernel ridge regression (KRR) and SVR learn a non-linear function by employing the kernel trick, i.e., they
learn a linear function in the space induced by the respective kernel which corresponds to a non-linear function in the
original space. They differ in the loss functions (ridge versus epsilon-insensitive loss). In contrast to SVR, fitting a
KRR can be done in closed-form and is typically faster for medium-sized datasets. On the other hand, the learned
model is non-sparse and thus slower than SVR at prediction-time.
This example illustrates both methods on an artificial dataset, which consists of a sinusoidal target function and strong
noise added to every fifth datapoint. The first figure compares the learned model of KRR and SVR when both com-
plexity/regularization and bandwidth of the RBF kernel are optimized using grid-search. The learned functions are
very similar; however, fitting KRR is approx. seven times faster than fitting SVR (both with grid-search). However,
prediction of 100000 target values is more than tree times faster with SVR since it has learned a sparse model using
only approx. 1/3 of the 100 training datapoints as support vectors.
The next figure compares the time for fitting and prediction of KRR and SVR for different sizes of the training set.
Fitting KRR is faster than SVR for medium- sized training sets (less than 1000 samples); however, for larger training
sets SVR scales better. With regard to prediction time, SVR is faster than KRR for all sizes of the training set because
of the learned sparse solution. Note that the degree of sparsity and thus the prediction time depends on the parameters
epsilon and C of the SVR.


•
Out:
SVR complexity and bandwidth selected and model fitted in 0.449 s

KRR complexity and bandwidth selected and model fitted in 0.221 s
Support vector ratio: 0.320
SVR prediction for 100000 inputs in 0.086 s
KRR prediction for 100000 inputs in 0.235 s
# Authors: Jan Hendrik Metzen <jhm@informatik.uni-bremen.de>

from __future__ import division

import time
import numpy as np
from sklearn.svm import SVR

from sklearn.model_selection import learning_curve
from sklearn.kernel_ridge import KernelRidge

# #############################################################################
# Generate sample data
X = 5 * rng.rand(10000, 1)
y = np.sin(X).ravel()
# Add noise to targets

y[::5] += 3 * (0.5 - rng.rand(X.shape[0] // 5))
X_plot = np.linspace(0, 5, 100000)[:, None]
# #############################################################################
# Fit regression model
train_size = 100
svr = GridSearchCV(SVR(kernel='rbf', gamma=0.1), cv=5,
param_grid={"C": [1e0, 1e1, 1e2, 1e3],
"gamma": np.logspace(-2, 2, 5)})
kr = GridSearchCV(KernelRidge(kernel='rbf', gamma=0.1), cv=5,

param_grid={"alpha": [1e0, 0.1, 1e-2, 1e-3],
"gamma": np.logspace(-2, 2, 5)})
t0 = time.time()
svr.fit(X[:train_size], y[:train_size])
svr_fit = time.time() - t0
print("SVR complexity and bandwidth selected and model fitted in %.3f s"
% svr_fit)
t0 = time.time()
kr.fit(X[:train_size], y[:train_size])
kr_fit = time.time() - t0
print("KRR complexity and bandwidth selected and model fitted in %.3f s"
% kr_fit)
sv_ratio = svr.best_estimator_.support_.shape[0] / train_size

print("Support vector ratio: %.3f" % sv_ratio)
t0 = time.time()
y_svr = svr.predict(X_plot)
svr_predict = time.time() - t0
print("SVR prediction for %d inputs in %.3f s"
% (X_plot.shape[0], svr_predict))
t0 = time.time()
y_kr = kr.predict(X_plot)
kr_predict = time.time() - t0
print("KRR prediction for %d inputs in %.3f s"
% (X_plot.shape[0], kr_predict))
# #############################################################################
# Look at the results
sv_ind = svr.best_estimator_.support_
plt.scatter(X[sv_ind], y[sv_ind], c='r', s=50, label='SVR support vectors',
zorder=2, edgecolors=(0, 0, 0))
plt.scatter(X[:100], y[:100], c='k', label='data', zorder=1,

edgecolors=(0, 0, 0))
plt.plot(X_plot, y_svr, c='r',
label='SVR (fit: %.3fs, predict: %.3fs)' % (svr_fit, svr_predict))
plt.plot(X_plot, y_kr, c='g',
label='KRR (fit: %.3fs, predict: %.3fs)' % (kr_fit, kr_predict))
plt.xlabel('data')
plt.ylabel('target')
plt.title('SVR versus Kernel Ridge')
plt.legend()
# Visualize training and prediction time

plt.figure()

X = 5 * rng.rand(10000, 1)
y[::5] += 3 * (0.5 - rng.rand(X.shape[0] // 5))
sizes = np.logspace(1, 4, 7).astype(np.int)
for name, estimator in {"KRR": KernelRidge(kernel='rbf', alpha=0.1,
gamma=10),
"SVR": SVR(kernel='rbf', C=1e1, gamma=10)}.items():
train_time = []
test_time = []
for train_test_size in sizes:
t0 = time.time()
estimator.fit(X[:train_test_size], y[:train_test_size])
train_time.append(time.time() - t0)
t0 = time.time()
estimator.predict(X_plot[:1000])
test_time.append(time.time() - t0)
plt.plot(sizes, train_time, 'o-', color="r" if name == "SVR" else "g",

label="%s (train)" % name)
plt.plot(sizes, test_time, 'o--', color="r" if name == "SVR" else "g",
label="%s (test)" % name)
plt.xscale("log")
plt.yscale("log")
plt.xlabel("Train size")
plt.ylabel("Time (seconds)")
plt.title('Execution Time')
plt.legend(loc="best")
# Visualize learning curves

plt.figure()
svr = SVR(kernel='rbf', C=1e1, gamma=0.1)

kr = KernelRidge(kernel='rbf', alpha=0.1, gamma=0.1)
train_sizes, train_scores_svr, test_scores_svr = \
learning_curve(svr, X[:100], y[:100], train_sizes=np.linspace(0.1, 1, 10),
scoring="neg_mean_squared_error", cv=10)
train_sizes_abs, train_scores_kr, test_scores_kr = \
learning_curve(kr, X[:100], y[:100], train_sizes=np.linspace(0.1, 1, 10),
plt.plot(train_sizes, -test_scores_svr.mean(1), 'o-', color="r",

label="SVR")

plt.plot(train_sizes, -test_scores_kr.mean(1), 'o-', color="g",

label="KRR")
plt.xlabel("Train size")
plt.ylabel("Mean Squared Error")
plt.title('Learning curves')
plt.show()
5.1.8 Explicit feature map approximation for RBF kernels
An example illustrating the approximation of the feature map of an RBF kernel.

It shows how to use RBFSampler and Nystroem to approximate the feature map of an RBF kernel for classification
with an SVM on the digits dataset. Results using a linear SVM in the original space, a linear SVM using the approx-
imate mappings and using a kernelized SVM are compared. Timings and accuracy for varying amounts of Monte
Carlo samplings (in the case of RBFSampler, which uses random Fourier features) and different sized subsets of the
training set (for Nystroem) for the approximate mapping are shown.
Please note that the dataset here is not large enough to show the benefits of kernel approximation, as the exact SVM is
still reasonably fast.
Sampling more dimensions clearly leads to better classification results, but comes at a greater cost. This means there
is a tradeoff between runtime and accuracy, given by the parameter n_components. Note that solving the Linear
SVM and also the approximate kernel SVM could be greatly accelerated by using stochastic gradient descent via
sklearn.linear_model.SGDClassifier. This is not easily possible for the case of the kernelized SVM.
The second plot visualized the decision surfaces of the RBF kernel SVM and the linear SVM with approximate kernel
maps. The plot shows decision surfaces of the classifiers projected onto the first two principal components of the data.
This visualization should be taken with a grain of salt since it is just an interesting slice through the decision surface
in 64 dimensions. In particular note that a datapoint (represented as a dot) does not necessarily be classified into the
region it is lying in, since it will not lie on the plane that the first two principal components span.
The usage of RBFSampler and Nystroem is described in detail in Kernel Approximation.


print(__doc__)
# Author: Gael Varoquaux <gael dot varoquaux at normalesup dot org>

# Andreas Mueller <amueller@ais.uni-bonn.de>
# Standard scientific Python imports

import numpy as np
# Import datasets, classifiers and performance metrics

from sklearn import datasets, svm, pipeline
from sklearn.kernel_approximation import (RBFSampler,
Nystroem)
# The digits dataset

digits = datasets.load_digits(n_class=9)
# To apply an classifier on this data, we need to flatten the image, to

# turn the data in a (samples, feature) matrix:
n_samples = len(digits.data)
data = digits.data / 16.
data -= data.mean(axis=0)
# We learn the digits on the first half of the digits

data_train, targets_train = (data[:n_samples // 2],
digits.target[:n_samples // 2])
# Now predict the value of the digit on the second half:

data_test, targets_test = (data[n_samples // 2:],
digits.target[n_samples // 2:])
# data_test = scaler.transform(data_test)
# Create a classifier: a support vector classifier

kernel_svm = svm.SVC(gamma=.2)
linear_svm = svm.LinearSVC()
# create pipeline from kernel approximation

# and linear svm
feature_map_fourier = RBFSampler(gamma=.2, random_state=1)
feature_map_nystroem = Nystroem(gamma=.2, random_state=1)
fourier_approx_svm = pipeline.Pipeline([("feature_map", feature_map_fourier),
("svm", svm.LinearSVC())])
nystroem_approx_svm = pipeline.Pipeline([("feature_map", feature_map_nystroem),

("svm", svm.LinearSVC())])
# fit and predict using linear and kernel svm:
kernel_svm_time = time()
kernel_svm.fit(data_train, targets_train)
kernel_svm_score = kernel_svm.score(data_test, targets_test)
kernel_svm_time = time() - kernel_svm_time
linear_svm_time = time()

linear_svm.fit(data_train, targets_train)
linear_svm_score = linear_svm.score(data_test, targets_test)
linear_svm_time = time() - linear_svm_time
sample_sizes = 30 * np.arange(1, 10)

fourier_scores = []
nystroem_scores = []
fourier_times = []
nystroem_times = []
for D in sample_sizes:
fourier_approx_svm.set_params(feature_map__n_components=D)
nystroem_approx_svm.set_params(feature_map__n_components=D)
start = time()
nystroem_approx_svm.fit(data_train, targets_train)
nystroem_times.append(time() - start)
start = time()
fourier_approx_svm.fit(data_train, targets_train)
fourier_times.append(time() - start)
fourier_score = fourier_approx_svm.score(data_test, targets_test)

nystroem_score = nystroem_approx_svm.score(data_test, targets_test)
nystroem_scores.append(nystroem_score)
fourier_scores.append(fourier_score)
# plot the results:

accuracy = plt.subplot(211)
# second y axis for timeings
timescale = plt.subplot(212)
accuracy.plot(sample_sizes, nystroem_scores, label="Nystroem approx. kernel")

timescale.plot(sample_sizes, nystroem_times, '--',
label='Nystroem approx. kernel')
accuracy.plot(sample_sizes, fourier_scores, label="Fourier approx. kernel")

timescale.plot(sample_sizes, fourier_times, '--',
label='Fourier approx. kernel')
# horizontal lines for exact rbf and linear kernels:

accuracy.plot([sample_sizes[0], sample_sizes[-1]],
[linear_svm_score, linear_svm_score], label="linear svm")
timescale.plot([sample_sizes[0], sample_sizes[-1]],
[linear_svm_time, linear_svm_time], '--', label='linear svm')
accuracy.plot([sample_sizes[0], sample_sizes[-1]],
[kernel_svm_score, kernel_svm_score], label="rbf svm")
timescale.plot([sample_sizes[0], sample_sizes[-1]],
[kernel_svm_time, kernel_svm_time], '--', label='rbf svm')
# vertical line for dataset dimensionality = 64

accuracy.plot([64, 64], [0.7, 1], label="n_features")
# legends and labels

accuracy.set_title("Classification accuracy")
timescale.set_title("Training times")
accuracy.set_xlim(sample_sizes[0], sample_sizes[-1])

accuracy.set_xticks(())
accuracy.set_ylim(np.min(fourier_scores), 1)
timescale.set_xlabel("Sampling steps = transformed feature dimension")
accuracy.set_ylabel("Classification accuracy")
timescale.set_ylabel("Training time in seconds")
accuracy.legend(loc='best')
timescale.legend(loc='best')
# visualize the decision surface, projected down to the first

# two principal components of the dataset
pca = PCA(n_components=8).fit(data_train)
X = pca.transform(data_train)
# Generate grid along first two principal components

multiples = np.arange(-2, 2, 0.1)
# steps along first component
first = multiples[:, np.newaxis] * pca.components_[0, :]
# steps along second component
second = multiples[:, np.newaxis] * pca.components_[1, :]
# combine
grid = first[np.newaxis, :, :] + second[:, np.newaxis, :]
flat_grid = grid.reshape(-1, data.shape[1])
# title for the plots

titles = ['SVC with rbf kernel',
'SVC (linear kernel)\n with Fourier rbf feature map\n'
'n_components=100',
'SVC (linear kernel)\n with Nystroem rbf feature map\n'
'n_components=100']
plt.tight_layout()
# predict and plot

for i, clf in enumerate((kernel_svm, nystroem_approx_svm,
fourier_approx_svm)):
# Plot the decision boundary. For that, we will assign a color to each
# point in the mesh [x_min, x_max]x[y_min, y_max].
plt.subplot(1, 3, i + 1)
Z = clf.predict(flat_grid)
# Put the result into a color plot

Z = Z.reshape(grid.shape[:-1])
plt.contourf(multiples, multiples, Z, cmap=plt.cm.Paired)
plt.axis('off')
# Plot also the training points

plt.scatter(X[:, 0], X[:, 1], c=targets_train, cmap=plt.cm.Paired,
plt.title(titles[i])
plt.tight_layout()
plt.show()

5.2 Examples based on real world datasets
Applications to real world problems with some medium sized datasets or interactive user interface.
5.2.1 Outlier detection on a real data set
This example illustrates the need for robust covariance estimation on a real data set. It is useful both for outlier
detection and for a better understanding of the data structure.
We selected two sets of two variables from the Boston housing data set as an illustration of what kind of analysis can
be done with several outlier detection tools. For the purpose of visualization, we are working with two-dimensional
examples, but one should be aware that things are not so trivial in high-dimension, as it will be pointed out.
In both examples below, the main result is that the empirical covariance estimate, as a non-robust one, is highly
influenced by the heterogeneous structure of the observations. Although the robust covariance estimate is able to
focus on the main mode of the data distribution, it sticks to the assumption that the data should be Gaussian distributed,
yielding some biased estimation of the data structure, but yet accurate to some extent. The One-Class SVM does not
assume any parametric form of the data distribution and can therefore model the complex shape of the data much
better.
First example
The first example illustrates how robust covariance estimation can help concentrating on a relevant cluster when an-
other one exists. Here, many observations are confounded into one and break down the empirical covariance estima-
tion. Of course, some screening tools would have pointed out the presence of two clusters (Support Vector Machines,
Gaussian Mixture Models, univariate outlier detection, . . . ). But had it been a high-dimensional example, none of
these could be applied that easily.
Second example
The second example shows the ability of the Minimum Covariance Determinant robust estimator of covariance to
concentrate on the main mode of the data distribution: the location seems to be well estimated, although the covariance
is hard to estimate due to the banana-shaped distribution. Anyway, we can get rid of some outlying observations. The
One-Class SVM is able to capture the real data structure, but the difficulty is to adjust its kernel bandwidth parameter
so as to obtain a good compromise between the shape of the data scatter matrix and the risk of over-fitting the data.

5.2. Examples based on real world datasets 679

print(__doc__)
# Author: Virgile Fritsch <virgile.fritsch@inria.fr>

import numpy as np
from sklearn.covariance import EllipticEnvelope
from sklearn.svm import OneClassSVM
import matplotlib.font_manager
# Get data
X1 = load_boston()['data'][:, [8, 10]] # two clusters
X2 = load_boston()['data'][:, [5, 12]] # "banana"-shaped
# Define "classifiers" to be used

classifiers = {
"Empirical Covariance": EllipticEnvelope(support_fraction=1.,
contamination=0.261),
"Robust Covariance (Minimum Covariance Determinant)":
EllipticEnvelope(contamination=0.261),
"OCSVM": OneClassSVM(nu=0.261, gamma=0.05)}
colors = ['m', 'g', 'b']
legend1 = {}
legend2 = {}

# Learn a frontier for outlier detection with several classifiers

xx1, yy1 = np.meshgrid(np.linspace(-8, 28, 500), np.linspace(3, 40, 500))
xx2, yy2 = np.meshgrid(np.linspace(3, 10, 500), np.linspace(-5, 45, 500))
for i, (clf_name, clf) in enumerate(classifiers.items()):
plt.figure(1)
clf.fit(X1)
Z1 = clf.decision_function(np.c_[xx1.ravel(), yy1.ravel()])
Z1 = Z1.reshape(xx1.shape)
legend1[clf_name] = plt.contour(
xx1, yy1, Z1, levels=[0], linewidths=2, colors=colors[i])
plt.figure(2)
clf.fit(X2)
Z2 = clf.decision_function(np.c_[xx2.ravel(), yy2.ravel()])
Z2 = Z2.reshape(xx2.shape)
legend2[clf_name] = plt.contour(
xx2, yy2, Z2, levels=[0], linewidths=2, colors=colors[i])
legend1_values_list = list(legend1.values())
legend1_keys_list = list(legend1.keys())
# Plot the results (= shape of the data points cloud)

plt.figure(1) # two clusters
plt.title("Outlier detection on a real data set (boston housing)")
plt.scatter(X1[:, 0], X1[:, 1], color='black')
bbox_args = dict(boxstyle="round", fc="0.8")
arrow_args = dict(arrowstyle="->")
plt.annotate("several confounded points", xy=(24, 19),
xycoords="data", textcoords="data",
xytext=(13, 10), bbox=bbox_args, arrowprops=arrow_args)
plt.xlim((xx1.min(), xx1.max()))
plt.ylim((yy1.min(), yy1.max()))
plt.legend((legend1_values_list[0].collections[0],
legend1_values_list[1].collections[0],
legend1_values_list[2].collections[0]),
(legend1_keys_list[0], legend1_keys_list[1], legend1_keys_list[2]),
loc="upper center",
prop=matplotlib.font_manager.FontProperties(size=12))
plt.ylabel("accessibility to radial highways")
plt.xlabel("pupil-teacher ratio by town")
legend2_values_list = list(legend2.values())
legend2_keys_list = list(legend2.keys())
plt.figure(2) # "banana" shape

plt.title("Outlier detection on a real data set (boston housing)")
plt.scatter(X2[:, 0], X2[:, 1], color='black')
plt.xlim((xx2.min(), xx2.max()))
plt.ylim((yy2.min(), yy2.max()))
plt.legend((legend2_values_list[0].collections[0],
legend2_values_list[1].collections[0],
legend2_values_list[2].collections[0]),
(legend2_keys_list[0], legend2_keys_list[1], legend2_keys_list[2]),
loc="upper center",
plt.ylabel("% lower status of the population")
plt.xlabel("average number of rooms per dwelling")
plt.show()

5.2.2 Compressive sensing: tomography reconstruction with L1 prior (Lasso)
This example shows the reconstruction of an image from a set of parallel projections, acquired along different angles.
Such a dataset is acquired in computed tomography (CT).
Without any prior information on the sample, the number of projections required to reconstruct the image is of the
order of the linear size l of the image (in pixels). For simplicity we consider here a sparse image, where only pixels
on the boundary of objects have a non-zero value. Such data could correspond for example to a cellular material.
Note however that most images are sparse in a different basis, such as the Haar wavelets. Only l/7 projections are
acquired, therefore it is necessary to use prior information available on the sample (its sparsity): this is an example of
compressive sensing.
The tomography projection operation is a linear transformation. In addition to the data-fidelity term corresponding
to a linear regression, we penalize the L1 norm of the image to account for its sparsity. The resulting optimization
problem is called the Lasso. We use the class sklearn.linear_model.Lasso, that uses the coordinate descent
algorithm. Importantly, this implementation is more computationally efficient on a sparse matrix, than the projection
operator used here.
The reconstruction with L1 penalization gives a result with zero error (all pixels are successfully labeled with 0 or 1),
even if noise was added to the projections. In comparison, an L2 penalization (sklearn.linear_model.Ridge)
produces a large number of labeling errors for the pixels. Important artifacts are observed on the reconstructed image,
contrary to the L1 penalization. Note in particular the circular artifact separating the pixels in the corners, that have
contributed to fewer projections than the central disk.
print(__doc__)
# Author: Emmanuelle Gouillart <emmanuelle.gouillart@nsup.org>

import numpy as np
from scipy import sparse

from scipy import ndimage

from sklearn.linear_model import Ridge
def _weights(x, dx=1, orig=0):

x = np.ravel(x)
floor_x = np.floor((x - orig) / dx).astype(np.int64)
alpha = (x - orig - floor_x * dx) / dx
return np.hstack((floor_x, floor_x + 1)), np.hstack((1 - alpha, alpha))
def _generate_center_coordinates(l_x):
X, Y = np.mgrid[:l_x, :l_x].astype(np.float64)
center = l_x / 2.
X += 0.5 - center
Y += 0.5 - center
return X, Y
def build_projection_operator(l_x, n_dir):

""" Compute the tomography design matrix.
Parameters
----------
l_x : int
linear size of image array
n_dir : int
number of angles at which projections are acquired.
Returns
-------
p : sparse matrix of shape (n_dir l_x, l_x**2)
"""
X, Y = _generate_center_coordinates(l_x)
angles = np.linspace(0, np.pi, n_dir, endpoint=False)
data_inds, weights, camera_inds = [], [], []
data_unravel_indices = np.arange(l_x ** 2)
data_unravel_indices = np.hstack((data_unravel_indices,
data_unravel_indices))
for i, angle in enumerate(angles):
Xrot = np.cos(angle) * X - np.sin(angle) * Y
inds, w = _weights(Xrot, dx=1, orig=X.min())
mask = np.logical_and(inds >= 0, inds < l_x)
weights += list(w[mask])
camera_inds += list(inds[mask] + i * l_x)
data_inds += list(data_unravel_indices[mask])
proj_operator = sparse.coo_matrix((weights, (camera_inds, data_inds)))
return proj_operator
def generate_synthetic_data():
""" Synthetic binary data """
rs = np.random.RandomState(0)
n_pts = 36

x, y = np.ogrid[0:l, 0:l]
mask_outer = (x - l / 2.) ** 2 + (y - l / 2.) ** 2 < (l / 2.) ** 2
mask = np.zeros((l, l))
points = l * rs.rand(2, n_pts)
mask[(points[0]).astype(np.int), (points[1]).astype(np.int)] = 1
mask = ndimage.gaussian_filter(mask, sigma=l / n_pts)
res = np.logical_and(mask > mask.mean(), mask_outer)
return np.logical_xor(res, ndimage.binary_erosion(res))
# Generate synthetic images, and projections

l = 128
proj_operator = build_projection_operator(l, l // 7)
data = generate_synthetic_data()
proj = proj_operator * data.ravel()[:, np.newaxis]
proj += 0.15 * np.random.randn(*proj.shape)
# Reconstruction with L2 (Ridge) penalization

rgr_ridge = Ridge(alpha=0.2)
rgr_ridge.fit(proj_operator, proj.ravel())
rec_l2 = rgr_ridge.coef_.reshape(l, l)
# Reconstruction with L1 (Lasso) penalization

# the best value of alpha was determined using cross validation
# with LassoCV
rgr_lasso = Lasso(alpha=0.001)
rgr_lasso.fit(proj_operator, proj.ravel())
rec_l1 = rgr_lasso.coef_.reshape(l, l)
plt.figure(figsize=(8, 3.3))
plt.subplot(131)
plt.imshow(data, cmap=plt.cm.gray, interpolation='nearest')
plt.axis('off')
plt.title('original image')
plt.subplot(132)
plt.imshow(rec_l2, cmap=plt.cm.gray, interpolation='nearest')
plt.title('L2 penalization')
plt.axis('off')
plt.subplot(133)
plt.imshow(rec_l1, cmap=plt.cm.gray, interpolation='nearest')
plt.title('L1 penalization')
plt.axis('off')
plt.subplots_adjust(hspace=0.01, wspace=0.01, top=1, bottom=0, left=0,

right=1)
plt.show()

5.2.3 Topic extraction with Non-negative Matrix Factorization and Latent Dirichlet
Allocation
This is an example of applying sklearn.decomposition.NMF and sklearn.decomposition.

LatentDirichletAllocation on a corpus of documents and extract additive models of the topic structure
of the corpus. The output is a list of topics, each represented as a list of terms (weights are not shown).
Non-negative Matrix Factorization is applied with two different objective functions: the Frobenius norm, and the
generalized Kullback-Leibler divergence. The latter is equivalent to Probabilistic Latent Semantic Indexing.
The default parameters (n_samples / n_features / n_components) should make the example runnable in a couple of
tens of seconds. You can try to increase the dimensions of the problem, but be aware that the time complexity is
polynomial in NMF. In LDA, the time complexity is proportional to (n_samples * iterations).
Out:
Loading dataset...
done in 9.591s.
Extracting tf-idf features for NMF...
done in 0.304s.
Extracting tf features for LDA...
done in 0.322s.
Fitting the NMF model (Frobenius norm) with tf-idf features, n_samples=2000 and n_
˓→features=1000...
done in 3.724s.
Topics in NMF model (Frobenius norm):

Topic #0: just people don think like know time good make way really say right ve want
˓→did ll new use years
Topic #1: windows use dos using window program os drivers application help software
˓→pc running ms screen files version card code work
Topic #2: god jesus bible faith christian christ christians does heaven sin believe
˓→lord life church mary atheism belief human love religion
Topic #3: thanks know does mail advance hi info interested email anybody looking card
˓→help like appreciated information send list video need
Topic #4: car cars tires miles 00 new engine insurance price condition oil power
˓→speed good 000 brake year models used bought
Topic #5: edu soon com send university internet mit ftp mail cc pub article
˓→information hope program mac email home contact blood
Topic #6: file problem files format win sound ftp pub read save site help image
˓→available create copy running memory self version
Topic #7: game team games year win play season players nhl runs goal hockey toronto
˓→division flyers player defense leafs bad teams
Topic #8: drive drives hard disk floppy software card mac computer power scsi
˓→controller apple mb 00 pc rom sale problem internal
Topic #9: key chip clipper keys encryption government public use secure enforcement
˓→phone nsa communications law encrypted security clinton used legal standard
Fitting the NMF model (generalized Kullback-Leibler divergence) with tf-idf features,
˓→n_samples=2000 and n_features=1000...
done in 1.860s.
Topics in NMF model (generalized Kullback-Leibler divergence):

Topic #0: people just like time don say really know way things make think right said
˓→did want ve probably work years
Topic #1: windows thanks using help need hi work know use looking mail software does
˓→used pc video available running info advance
Topic #2: god does true read know say believe subject says religion mean question
˓→point jesus people book christian mind understand matter

Topic #3: thanks know like interested mail just want new send edu list does bike
˓→thing email reply post wondering hear heard
Topic #4: time new 10 year sale old offer 20 16 15 great 30 weeks good test model
˓→condition 11 14 power
Topic #5: use number com government new university data states information talk phone
˓→right including security provide control following long used research
Topic #6: edu try file soon remember problem com program hope mike space article
˓→wrong library short include win little couldn sun
Topic #7: year world team game play won win games season maybe case second does did
˓→series playing nhl fact said points
Topic #8: think don drive need hard make people mac read going pretty try sure order
˓→means trying apple case bit drives
Topic #9: just good use way got like ll doesn want sure don doing thought does wrong
˓→right better make stuff speed
Fitting LDA models with tf features, n_samples=2000 and n_features=1000...

done in 3.591s.
Topics in LDA model:

Topic #0: edu com mail send graphics ftp pub available contact university list faq ca
˓→information cs 1993 program sun uk mit
Topic #1: don like just know think ve way use right good going make sure ll point got
˓→need really time doesn
Topic #2: christian think atheism faith pittsburgh new bible radio games alt lot just
˓→religion like book read play time subject believe
Topic #3: drive disk windows thanks use card drives hard version pc software file
˓→using scsi help does new dos controller 16
Topic #4: hiv health aids disease april medical care research 1993 light information
˓→study national service test led 10 page new drug
Topic #5: god people does just good don jesus say israel way life know true fact time
˓→law want believe make think
Topic #6: 55 10 11 18 15 team game 19 period play 23 12 13 flyers 20 25 22 17 24 16

Topic #7: car year just cars new engine like bike good oil insurance better tires 000
˓→thing speed model brake driving performance
Topic #8: people said did just didn know time like went think children came come don
˓→took years say dead told started
Topic #9: key space law government public use encryption earth section security moon
˓→probe enforcement keys states lunar military crime surface technology
# Author: Olivier Grisel <olivier.grisel@ensta.org>

# Lars Buitinck
# Chyi-Kwei Yau <chyikwei.yau@gmail.com>

from sklearn.feature_extraction.text import TfidfVectorizer, CountVectorizer

from sklearn.decomposition import NMF, LatentDirichletAllocation
from sklearn.datasets import fetch_20newsgroups
n_samples = 2000
n_features = 1000

n_components = 10
n_top_words = 20
def print_top_words(model, feature_names, n_top_words):

for topic_idx, topic in enumerate(model.components_):
message = "Topic #%d: " % topic_idx
message += " ".join([feature_names[i]
for i in topic.argsort()[:-n_top_words - 1:-1]])
print(message)
print()
# Load the 20 newsgroups dataset and vectorize it. We use a few heuristics
# to filter out useless terms early on: the posts are stripped of headers,
# footers and quoted replies, and common English words, words occurring in
# only one document or in at least 95% of the documents are removed.
print("Loading dataset...")
t0 = time()
dataset = fetch_20newsgroups(shuffle=True, random_state=1,
remove=('headers', 'footers', 'quotes'))
data_samples = dataset.data[:n_samples]
print("done in %0.3fs." % (time() - t0))
# Use tf-idf features for NMF.

print("Extracting tf-idf features for NMF...")
tfidf_vectorizer = TfidfVectorizer(max_df=0.95, min_df=2,
max_features=n_features,
stop_words='english')
t0 = time()
tfidf = tfidf_vectorizer.fit_transform(data_samples)
# Use tf (raw term count) features for LDA.

print("Extracting tf features for LDA...")
tf_vectorizer = CountVectorizer(max_df=0.95, min_df=2,
max_features=n_features,
t0 = time()
tf = tf_vectorizer.fit_transform(data_samples)
print()
# Fit the NMF model

print("Fitting the NMF model (Frobenius norm) with tf-idf features, "
"n_samples=%d and n_features=%d..."
t0 = time()
nmf = NMF(n_components=n_components, random_state=1,
alpha=.1, l1_ratio=.5).fit(tfidf)
print("\nTopics in NMF model (Frobenius norm):")

tfidf_feature_names = tfidf_vectorizer.get_feature_names()
print_top_words(nmf, tfidf_feature_names, n_top_words)
# Fit the NMF model

print("Fitting the NMF model (generalized Kullback-Leibler divergence) with "

"tf-idf features, n_samples=%d and n_features=%d..."
t0 = time()
nmf = NMF(n_components=n_components, random_state=1,
beta_loss='kullback-leibler', solver='mu', max_iter=1000, alpha=.1,
l1_ratio=.5).fit(tfidf)
print("\nTopics in NMF model (generalized Kullback-Leibler divergence):")

tfidf_feature_names = tfidf_vectorizer.get_feature_names()
print_top_words(nmf, tfidf_feature_names, n_top_words)
print("Fitting LDA models with tf features, "

"n_samples=%d and n_features=%d..."
lda = LatentDirichletAllocation(n_components=n_components, max_iter=5,
learning_method='online',
learning_offset=50.,
random_state=0)
t0 = time()
lda.fit(tf)
print("\nTopics in LDA model:")

tf_feature_names = tf_vectorizer.get_feature_names()
print_top_words(lda, tf_feature_names, n_top_words)
5.2.4 Faces recognition example using eigenfaces and SVMs
The dataset used in this example is a preprocessed excerpt of the “Labeled Faces in the Wild”, aka LFW:
Ariel Sharon 0.67 0.92 0.77 13

Colin Powell 0.75 0.78 0.76 60
George W Bush 0.86 0.86 0.86 146
Hugo Chavez 0.67 0.67 0.67 15
Tony Blair 0.81 0.69 0.75 36
avg / total 0.80 0.80 0.80 322


•
Out:
Total dataset size:
n_samples: 1288
n_features: 1850
n_classes: 7
Extracting the top 150 eigenfaces from 966 faces
done in 0.988s
Projecting the input data on the eigenfaces orthonormal basis
done in 0.007s
Fitting the classifier to the training set
done in 52.895s
Best estimator found by grid search:
SVC(C=1000.0, cache_size=200, class_weight='balanced', coef0=0.0,
Predicting people's names on the test set

done in 0.061s
Ariel Sharon 0.86 0.46 0.60 13

Colin Powell 0.80 0.87 0.83 60
George W Bush 0.82 0.98 0.89 146
Hugo Chavez 1.00 0.53 0.70 15
Tony Blair 0.97 0.78 0.86 36
micro avg 0.85 0.85 0.85 322

macro avg 0.90 0.72 0.78 322
weighted avg 0.86 0.85 0.84 322
[[ 6 2 0 5 0 0 0]
[ 1 52 0 7 0 0 0]
[ 0 2 16 9 0 0 0]
[ 0 3 0 143 0 0 0]
[ 0 1 0 3 20 0 1]
[ 0 4 0 2 1 8 0]
[ 0 1 1 6 0 0 28]]

import logging

from sklearn.datasets import fetch_lfw_people
print(__doc__)

logging.basicConfig(level=logging.INFO, format='%(asctime)s %(message)s')
# #############################################################################
# Download the data, if not already on disk and load it as numpy arrays
lfw_people = fetch_lfw_people(min_faces_per_person=70, resize=0.4)
# introspect the images arrays to find the shapes (for plotting)

n_samples, h, w = lfw_people.images.shape
# for machine learning we use the 2 data directly (as relative pixel

# positions info is ignored by this model)

X = lfw_people.data
# the label to predict is the id of the person

y = lfw_people.target
target_names = lfw_people.target_names
n_classes = target_names.shape[0]
print("Total dataset size:")

print("n_samples: %d" % n_samples)
print("n_features: %d" % n_features)
print("n_classes: %d" % n_classes)
# #############################################################################
# Split into a training set and a test set using a stratified k fold
# split into a training and testing set

# #############################################################################
# Compute a PCA (eigenfaces) on the face dataset (treated as unlabeled
# dataset): unsupervised feature extraction / dimensionality reduction
n_components = 150
print("Extracting the top %d eigenfaces from %d faces"

% (n_components, X_train.shape[0]))
t0 = time()
pca = PCA(n_components=n_components, svd_solver='randomized',
whiten=True).fit(X_train)
eigenfaces = pca.components_.reshape((n_components, h, w))
print("Projecting the input data on the eigenfaces orthonormal basis")

t0 = time()
X_train_pca = pca.transform(X_train)
X_test_pca = pca.transform(X_test)
# #############################################################################
# Train a SVM classification model
print("Fitting the classifier to the training set")

t0 = time()
param_grid = {'C': [1e3, 5e3, 1e4, 5e4, 1e5],
'gamma': [0.0001, 0.0005, 0.001, 0.005, 0.01, 0.1], }
clf = GridSearchCV(SVC(kernel='rbf', class_weight='balanced'),
param_grid, cv=5)
clf = clf.fit(X_train_pca, y_train)
print("Best estimator found by grid search:")
print(clf.best_estimator_)

# #############################################################################
# Quantitative evaluation of the model quality on the test set
print("Predicting people's names on the test set")

t0 = time()
y_pred = clf.predict(X_test_pca)
print(classification_report(y_test, y_pred, target_names=target_names))

print(confusion_matrix(y_test, y_pred, labels=range(n_classes)))
# #############################################################################
# Qualitative evaluation of the predictions using matplotlib
def plot_gallery(images, titles, h, w, n_row=3, n_col=4):

"""Helper function to plot a gallery of portraits"""
plt.figure(figsize=(1.8 * n_col, 2.4 * n_row))
plt.subplots_adjust(bottom=0, left=.01, right=.99, top=.90, hspace=.35)
for i in range(n_row * n_col):
plt.imshow(images[i].reshape((h, w)), cmap=plt.cm.gray)
plt.title(titles[i], size=12)
plt.xticks(())
plt.yticks(())
# plot the result of the prediction on a portion of the test set
def title(y_pred, y_test, target_names, i):

pred_name = target_names[y_pred[i]].rsplit(' ', 1)[-1]
true_name = target_names[y_test[i]].rsplit(' ', 1)[-1]
return 'predicted: %s\ntrue: %s' % (pred_name, true_name)
prediction_titles = [title(y_pred, y_test, target_names, i)

for i in range(y_pred.shape[0])]
plot_gallery(X_test, prediction_titles, h, w)
# plot the gallery of the most significative eigenfaces
eigenface_titles = ["eigenface %d" % i for i in range(eigenfaces.shape[0])]

plot_gallery(eigenfaces, eigenface_titles, h, w)
plt.show()
5.2.5 Model Complexity Influence
Demonstrate how model complexity influences both prediction accuracy and computational performance.

The dataset is the Boston Housing dataset (resp. 20 Newsgroups) for regression (resp. classification).
For each class of models we make the model complexity vary through the choice of relevant model parameters and
measure the influence on both computational performance (latency) and predictive power (MSE or Hamming Loss).

•
Out:
Benchmarking SGDClassifier(alpha=0.001, average=False, class_weight=None,
l1_ratio=0.25, learning_rate='optimal', loss='modified_huber',
max_iter=None, n_iter=None, n_iter_no_change=5, n_jobs=None,
penalty='elasticnet', power_t=0.5, random_state=None, shuffle=True,
tol=0.001, validation_fraction=0.1, verbose=0, warm_start=False)
Complexity: 4495 | Hamming Loss (Misclassification Ratio): 0.2536 | Pred. Time: 0.
˓→021621s

˓→017138s

˓→014366s

˓→012016s

Benchmarking NuSVR(C=1000.0, cache_size=200, coef0=0.0, degree=3, gamma=3.0517578125e-

˓→05,
kernel='rbf', max_iter=-1, nu=0.1, shrinking=True, tol=0.001,

verbose=False)
Complexity: 69 | MSE: 31.8139 | Pred. Time: 0.000283s

˓→05,

verbose=False)

˓→05,

verbose=False)

˓→05,

verbose=False)

˓→05,

verbose=False)
Benchmarking GradientBoostingRegressor(alpha=0.9, criterion='friedman_mse', init=None,

learning_rate=0.1, loss='ls', max_depth=3, max_features=None,
max_leaf_nodes=None, min_impurity_decrease=0.0,
min_impurity_split=None, min_samples_leaf=1,
min_samples_split=2, min_weight_fraction_leaf=0.0,
n_estimators=10, n_iter_no_change=None, presort='auto',
random_state=None, subsample=1.0, tol=0.0001,





print(__doc__)
# Author: Eustache Diemert <eustache@diemert.fr>

import time
import numpy as np
from mpl_toolkits.axes_grid1.parasite_axes import host_subplot
from mpl_toolkits.axisartist.axislines import Axes
from scipy.sparse.csr import csr_matrix

from sklearn.utils import shuffle
from sklearn.metrics import mean_squared_error
from sklearn.svm.classes import NuSVR
from sklearn.ensemble.gradient_boosting import GradientBoostingRegressor
from sklearn.linear_model.stochastic_gradient import SGDClassifier
from sklearn.metrics import hamming_loss
# #############################################################################
# Routines
# Initialize random generator

np.random.seed(0)
def generate_data(case, sparse=False):

"""Generate regression/classification data."""
bunch = None

if case == 'regression':
bunch = datasets.load_boston()
elif case == 'classification':
bunch = datasets.fetch_20newsgroups_vectorized(subset='all')
X, y = shuffle(bunch.data, bunch.target)
offset = int(X.shape[0] * 0.8)
X_train, y_train = X[:offset], y[:offset]
X_test, y_test = X[offset:], y[offset:]
if sparse:
X_train = csr_matrix(X_train)
X_test = csr_matrix(X_test)
else:
X_train = np.array(X_train)
X_test = np.array(X_test)
y_test = np.array(y_test)
y_train = np.array(y_train)
data = {'X_train': X_train, 'X_test': X_test, 'y_train': y_train,
'y_test': y_test}
return data
def benchmark_influence(conf):
"""
Benchmark influence of :changing_param: on both MSE and latency.
"""
prediction_times = []
prediction_powers = []
complexities = []
for param_value in conf['changing_param_values']:
conf['tuned_params'][conf['changing_param']] = param_value
estimator = conf['estimator'](**conf['tuned_params'])
print("Benchmarking %s" % estimator)
estimator.fit(conf['data']['X_train'], conf['data']['y_train'])
conf['postfit_hook'](estimator)
complexity = conf['complexity_computer'](estimator)
complexities.append(complexity)
start_time = time.time()
for _ in range(conf['n_samples']):
y_pred = estimator.predict(conf['data']['X_test'])
elapsed_time = (time.time() - start_time) / float(conf['n_samples'])
prediction_times.append(elapsed_time)
pred_score = conf['prediction_performance_computer'](
conf['data']['y_test'], y_pred)
prediction_powers.append(pred_score)
print("Complexity: %d | %s: %.4f | Pred. Time: %fs\n" % (
complexity, conf['prediction_performance_label'], pred_score,
elapsed_time))
return prediction_powers, prediction_times, complexities
def plot_influence(conf, mse_values, prediction_times, complexities):

"""
Plot influence of model complexity on both accuracy and latency.
"""
host = host_subplot(111, axes_class=Axes)
plt.subplots_adjust(right=0.75)
par1 = host.twinx()

host.set_xlabel('Model Complexity (%s)' % conf['complexity_label'])

y1_label = conf['prediction_performance_label']
y2_label = "Time (s)"
host.set_ylabel(y1_label)
par1.set_ylabel(y2_label)
p1, = host.plot(complexities, mse_values, 'b-', label="prediction error")
p2, = par1.plot(complexities, prediction_times, 'r-',
label="latency")
host.legend(loc='upper right')
host.axis["left"].label.set_color(p1.get_color())
par1.axis["right"].label.set_color(p2.get_color())
plt.title('Influence of Model Complexity - %s' % conf['estimator'].__name__)
plt.show()
def _count_nonzero_coefficients(estimator):
a = estimator.coef_.toarray()
return np.count_nonzero(a)
# #############################################################################
# Main code
regression_data = generate_data('regression')
classification_data = generate_data('classification', sparse=True)
configurations = [
{'estimator': SGDClassifier,
'tuned_params': {'penalty': 'elasticnet', 'alpha': 0.001, 'loss':
'modified_huber', 'fit_intercept': True, 'tol': 1e-3},
'changing_param': 'l1_ratio',
'changing_param_values': [0.25, 0.5, 0.75, 0.9],
'complexity_label': 'non_zero coefficients',
'complexity_computer': _count_nonzero_coefficients,
'prediction_performance_computer': hamming_loss,
'prediction_performance_label': 'Hamming Loss (Misclassification Ratio)',
'postfit_hook': lambda x: x.sparsify(),
'data': classification_data,
'n_samples': 30},
{'estimator': NuSVR,
'tuned_params': {'C': 1e3, 'gamma': 2 ** -15},
'changing_param': 'nu',
'changing_param_values': [0.1, 0.25, 0.5, 0.75, 0.9],
'complexity_label': 'n_support_vectors',
'complexity_computer': lambda x: len(x.support_vectors_),
'data': regression_data,
'postfit_hook': lambda x: x,
'prediction_performance_computer': mean_squared_error,
'prediction_performance_label': 'MSE',
'n_samples': 30},
{'estimator': GradientBoostingRegressor,
'tuned_params': {'loss': 'ls'},
'changing_param': 'n_estimators',
'changing_param_values': [10, 50, 100, 200, 500],
'complexity_label': 'n_trees',
'complexity_computer': lambda x: x.n_estimators,
'data': regression_data,
'postfit_hook': lambda x: x,
'prediction_performance_computer': mean_squared_error,
'prediction_performance_label': 'MSE',
'n_samples': 30},

]
for conf in configurations:
prediction_performances, prediction_times, complexities = \
benchmark_influence(conf)
plot_influence(conf, prediction_performances, prediction_times,
complexities)
5.2.6 Species distribution modeling
Modeling species’ geographic distributions is an important problem in conservation biology. In this example we model
the geographic distribution of two south american mammals given past observations and 14 environmental variables.
Since we have only positive examples (there are no unsuccessful observations), we cast this problem as a density
estimation problem and use the OneClassSVM provided by the package sklearn.svm as our modeling tool. The dataset
is provided by Phillips et. al. (2006). If available, the example uses basemap to plot the coast lines and national
boundaries of South America.
The two species are:
• “Bradypus variegatus” , the Brown-throated Sloth.
• “Microryzomys minutus” , also known as the Forest Small Rice Rat, a rodent that lives in Peru, Colombia,
Ecuador, Peru, and Venezuela.
References
• “Maximum entropy modeling of species geographic distributions” S. J. Phillips, R. P. Anderson, R. E. Schapire

- Ecological Modelling, 190:231-259, 2006.

Out:
________________________________________________________________________________
Modeling distribution of species 'bradypus variegatus'
- fit OneClassSVM ... done.
- plot coastlines from coverage
- predict species distribution
Area under the ROC curve : 0.868443

________________________________________________________________________________
Modeling distribution of species 'microryzomys minutus'
- fit OneClassSVM ... done.
- predict species distribution
Area under the ROC curve : 0.993919
time elapsed: 20.02s
# Authors: Peter Prettenhofer <peter.prettenhofer@gmail.com>

# Jake Vanderplas <vanderplas@astro.washington.edu>

#
import numpy as np
from sklearn.datasets.base import Bunch

from sklearn.datasets import fetch_species_distributions
from sklearn.datasets.species_distributions import construct_grids
from sklearn import svm, metrics
# if basemap is available, we'll use it.

# otherwise, we'll improvise later...
try:
from mpl_toolkits.basemap import Basemap
basemap = True
except ImportError:
basemap = False
print(__doc__)
def create_species_bunch(species_name, train, test, coverages, xgrid, ygrid):

"""Create a bunch with information about a particular organism
This will use the test/train record arrays to extract the

data specific to the given species name.
"""
bunch = Bunch(name=' '.join(species_name.split("_")[:2]))
species_name = species_name.encode('ascii')
points = dict(test=test, train=train)
for label, pts in points.items():

# choose points associated with the desired species
pts = pts[pts['species'] == species_name]
bunch['pts_%s' % label] = pts
# determine coverage values for each of the training & testing points
ix = np.searchsorted(xgrid, pts['dd long'])
iy = np.searchsorted(ygrid, pts['dd lat'])
bunch['cov_%s' % label] = coverages[:, -iy, ix].T
return bunch
def plot_species_distribution(species=("bradypus_variegatus_0",
"microryzomys_minutus_0")):
"""
Plot the species distribution.
"""
if len(species) > 2:
print("Note: when more than two species are provided,"
" only the first two will be used")

t0 = time()
# Load the compressed data

data = fetch_species_distributions()
# Set up the data grid

xgrid, ygrid = construct_grids(data)
# The grid in x,y coordinates

X, Y = np.meshgrid(xgrid, ygrid[::-1])
# create a bunch for each species

BV_bunch = create_species_bunch(species[0],
data.train, data.test,
data.coverages, xgrid, ygrid)
MM_bunch = create_species_bunch(species[1],
data.train, data.test,
data.coverages, xgrid, ygrid)
# background points (grid coordinates) for evaluation

np.random.seed(13)
background_points = np.c_[np.random.randint(low=0, high=data.Ny,
size=10000),
np.random.randint(low=0, high=data.Nx,
size=10000)].T
# We'll make use of the fact that coverages[6] has measurements at all
# land points. This will help us decide between land and water.
land_reference = data.coverages[6]
# Fit, predict, and plot for each species.

for i, species in enumerate([BV_bunch, MM_bunch]):
print("_" * 80)
print("Modeling distribution of species '%s'" % species.name)
# Standardize features
mean = species.cov_train.mean(axis=0)
std = species.cov_train.std(axis=0)
train_cover_std = (species.cov_train - mean) / std
# Fit OneClassSVM
print(" - fit OneClassSVM ... ", end='')
clf = svm.OneClassSVM(nu=0.1, kernel="rbf", gamma=0.5)
clf.fit(train_cover_std)
print("done.")
# Plot map of South America

if basemap:
print(" - plot coastlines using basemap")
m = Basemap(projection='cyl', llcrnrlat=Y.min(),
urcrnrlat=Y.max(), llcrnrlon=X.min(),
urcrnrlon=X.max(), resolution='c')
m.drawcoastlines()
m.drawcountries()
else:
print(" - plot coastlines from coverage")
plt.contour(X, Y, land_reference,

levels=[-9998], colors="k",
linestyles="solid")
plt.xticks([])
plt.yticks([])
print(" - predict species distribution")
# Predict species distribution using the training data

Z = np.ones((data.Ny, data.Nx), dtype=np.float64)
# We'll predict only for the land points.

idx = np.where(land_reference > -9999)
coverages_land = data.coverages[:, idx[0], idx[1]].T
pred = clf.decision_function((coverages_land - mean) / std)

Z *= pred.min()
Z[idx[0], idx[1]] = pred
levels = np.linspace(Z.min(), Z.max(), 25)

Z[land_reference == -9999] = -9999
# plot contours of the prediction

plt.contourf(X, Y, Z, levels=levels, cmap=plt.cm.Reds)
plt.colorbar(format='%.2f')
# scatter training/testing points

plt.scatter(species.pts_train['dd long'], species.pts_train['dd lat'],
s=2 ** 2, c='black',
marker='^', label='train')
plt.scatter(species.pts_test['dd long'], species.pts_test['dd lat'],
s=2 ** 2, c='black',
marker='x', label='test')
plt.legend()
plt.title(species.name)
plt.axis('equal')
# Compute AUC with regards to background points

pred_background = Z[background_points[0], background_points[1]]
pred_test = clf.decision_function((species.cov_test - mean) / std)
scores = np.r_[pred_test, pred_background]
y = np.r_[np.ones(pred_test.shape), np.zeros(pred_background.shape)]
fpr, tpr, thresholds = metrics.roc_curve(y, scores)
roc_auc = metrics.auc(fpr, tpr)
plt.text(-35, -70, "AUC: %.3f" % roc_auc, ha="right")
print("\n Area under the ROC curve : %f" % roc_auc)
print("\ntime elapsed: %.2fs" % (time() - t0))
plot_species_distribution()
plt.show()

5.2.7 Visualizing the stock market structure
This example employs several unsupervised learning techniques to extract the stock market structure from variations
in historical quotes.
The quantity that we use is the daily variation in quote price: quotes that are linked tend to cofluctuate during a day.
Learning a graph structure
We use sparse inverse covariance estimation to find which quotes are correlated conditionally on the others. Specifi-
cally, sparse inverse covariance gives us a graph, that is a list of connection. For each symbol, the symbols that it is
connected too are those useful to explain its fluctuations.
Clustering
We use clustering to group together quotes that behave similarly. Here, amongst the various clustering techniques
available in the scikit-learn, we use Affinity Propagation as it does not enforce equal-size clusters, and it can choose
automatically the number of clusters from the data.
Note that this gives us a different indication than the graph, as the graph reflects conditional relations between variables,
while the clustering reflects marginal properties: variables clustered together can be considered as having a similar
impact at the level of the full stock market.
Embedding in 2D space
For visualization purposes, we need to lay out the different symbols on a 2D canvas. For this we use Manifold learning
techniques to retrieve 2D embedding.
Visualization
The output of the 3 models are combined in a 2D graph where nodes represents the stocks and edges the:
• cluster labels are used to define the color of the nodes
• the sparse covariance model is used to display the strength of the edges
• the 2D embedding is used to position the nodes in the plan
This example has a fair amount of visualization-related code, as visualization is crucial here to display the graph. One
of the challenge is to position the labels minimizing overlap. For this we use an heuristic based on the direction of the
nearest neighbor along each axis.

Out:
Cluster 1: Apple, Amazon, Yahoo

Cluster 2: Comcast, Cablevision, Time Warner
Cluster 3: ConocoPhillips, Chevron, Total, Valero Energy, Exxon
Cluster 4: Cisco, Dell, HP, IBM, Microsoft, SAP, Texas Instruments
Cluster 5: Boeing, General Dynamics, Northrop Grumman, Raytheon
Cluster 6: AIG, American express, Bank of America, Caterpillar, CVS, DuPont de
˓→Nemours, Ford, General Electrics, Goldman Sachs, Home Depot, JPMorgan Chase,
˓→Marriott, 3M, Ryder, Wells Fargo, Wal-Mart
Cluster 7: McDonald's
Cluster 8: GlaxoSmithKline, Novartis, Pfizer, Sanofi-Aventis, Unilever
Cluster 9: Kellogg, Coca Cola, Pepsi
Cluster 10: Colgate-Palmolive, Kimberly-Clark, Procter Gamble
Cluster 11: Canon, Honda, Navistar, Sony, Toyota, Xerox

# Author: Gael Varoquaux gael.varoquaux@normalesup.org

import sys
import numpy as np
import pandas as pd
from sklearn import cluster, covariance, manifold
print(__doc__)
# #############################################################################
# Retrieve the data from Internet
# The data is from 2003 - 2008. This is reasonably calm: (not too long ago so
# that we get high-tech firms, and before the 2008 crash). This kind of
# historical data can be obtained for from APIs like the quandl.com and
# alphavantage.co ones.
symbol_dict = {
'TOT': 'Total',
'XOM': 'Exxon',
'CVX': 'Chevron',
'COP': 'ConocoPhillips',
'VLO': 'Valero Energy',
'MSFT': 'Microsoft',
'IBM': 'IBM',
'TWX': 'Time Warner',
'CMCSA': 'Comcast',
'CVC': 'Cablevision',
'YHOO': 'Yahoo',
'DELL': 'Dell',
'HPQ': 'HP',
'AMZN': 'Amazon',
'TM': 'Toyota',
'CAJ': 'Canon',
'SNE': 'Sony',
'F': 'Ford',
'HMC': 'Honda',
'NAV': 'Navistar',
'NOC': 'Northrop Grumman',
'BA': 'Boeing',
'KO': 'Coca Cola',
'MMM': '3M',
'MCD': 'McDonald\'s',
'PEP': 'Pepsi',
'K': 'Kellogg',
'UN': 'Unilever',
'MAR': 'Marriott',
'PG': 'Procter Gamble',
'CL': 'Colgate-Palmolive',
'GE': 'General Electrics',
'WFC': 'Wells Fargo',

'JPM': 'JPMorgan Chase',

'AIG': 'AIG',
'AXP': 'American express',
'BAC': 'Bank of America',
'GS': 'Goldman Sachs',
'AAPL': 'Apple',
'SAP': 'SAP',
'CSCO': 'Cisco',
'TXN': 'Texas Instruments',
'XRX': 'Xerox',
'WMT': 'Wal-Mart',
'HD': 'Home Depot',
'GSK': 'GlaxoSmithKline',
'PFE': 'Pfizer',
'SNY': 'Sanofi-Aventis',
'NVS': 'Novartis',
'KMB': 'Kimberly-Clark',
'R': 'Ryder',
'GD': 'General Dynamics',
'RTN': 'Raytheon',
'CVS': 'CVS',
'CAT': 'Caterpillar',
'DD': 'DuPont de Nemours'}
symbols, names = np.array(sorted(symbol_dict.items())).T
quotes = []
for symbol in symbols:

print('Fetching quote history for %r' % symbol, file=sys.stderr)
url = ('https://raw.githubusercontent.com/scikit-learn/examples-data/'
'master/financial-data/{}.csv')
quotes.append(pd.read_csv(url.format(symbol)))
close_prices = np.vstack([q['close'] for q in quotes])

open_prices = np.vstack([q['open'] for q in quotes])
# The daily variations of the quotes are what carry most information
variation = close_prices - open_prices
# #############################################################################
# Learn a graphical structure from the correlations
edge_model = covariance.GraphicalLassoCV(cv=5)
# standardize the time series: using correlations rather than covariance

# is more efficient for structure recovery
X = variation.copy().T
X /= X.std(axis=0)
edge_model.fit(X)
# #############################################################################
# Cluster using affinity propagation
_, labels = cluster.affinity_propagation(edge_model.covariance_)
n_labels = labels.max()

for i in range(n_labels + 1):

print('Cluster %i: %s' % ((i + 1), ', '.join(names[labels == i])))
# #############################################################################
# Find a low-dimension embedding for visualization: find the best position of
# the nodes (the stocks) on a 2D plane
# We use a dense eigen_solver to achieve reproducibility (arpack is

# initiated with random vectors that we don't control). In addition, we
# use a large number of neighbors to capture the large-scale structure.
node_position_model = manifold.LocallyLinearEmbedding(
n_components=2, eigen_solver='dense', n_neighbors=6)
embedding = node_position_model.fit_transform(X.T).T
# #############################################################################
# Visualization
plt.figure(1, facecolor='w', figsize=(10, 8))
plt.clf()
ax = plt.axes([0., 0., 1., 1.])
plt.axis('off')
# Display a graph of the partial correlations

partial_correlations = edge_model.precision_.copy()
d = 1 / np.sqrt(np.diag(partial_correlations))
partial_correlations *= d
partial_correlations *= d[:, np.newaxis]
non_zero = (np.abs(np.triu(partial_correlations, k=1)) > 0.02)
# Plot the nodes using the coordinates of our embedding

plt.scatter(embedding[0], embedding[1], s=100 * d ** 2, c=labels,
cmap=plt.cm.nipy_spectral)
# Plot the edges

start_idx, end_idx = np.where(non_zero)
# a sequence of (*line0*, *line1*, *line2*), where::
# linen = (x0, y0), (x1, y1), ... (xm, ym)
segments = [[embedding[:, start], embedding[:, stop]]
for start, stop in zip(start_idx, end_idx)]
values = np.abs(partial_correlations[non_zero])
lc = LineCollection(segments,
zorder=0, cmap=plt.cm.hot_r,
norm=plt.Normalize(0, .7 * values.max()))
lc.set_array(values)
lc.set_linewidths(15 * values)
ax.add_collection(lc)
# Add a label to each node. The challenge here is that we want to

# position the labels to avoid overlap with other labels
for index, (name, label, (x, y)) in enumerate(
zip(names, labels, embedding.T)):
dx = x - embedding[0]
dx[index] = 1
dy = y - embedding[1]
dy[index] = 1
this_dx = dx[np.argmin(np.abs(dy))]
this_dy = dy[np.argmin(np.abs(dx))]

if this_dx > 0:
horizontalalignment = 'left'
x = x + .002
else:
horizontalalignment = 'right'
x = x - .002
if this_dy > 0:
verticalalignment = 'bottom'
y = y + .002
else:
verticalalignment = 'top'
y = y - .002
plt.text(x, y, name, size=10,
horizontalalignment=horizontalalignment,
verticalalignment=verticalalignment,
bbox=dict(facecolor='w',
edgecolor=plt.cm.nipy_spectral(label / float(n_labels)),
alpha=.6))
plt.xlim(embedding[0].min() - .15 * embedding[0].ptp(),

embedding[0].max() + .10 * embedding[0].ptp(),)
plt.ylim(embedding[1].min() - .03 * embedding[1].ptp(),
embedding[1].max() + .03 * embedding[1].ptp())
plt.show()
5.2.8 Wikipedia principal eigenvector
A classical way to assert the relative importance of vertices in a graph is to compute the principal eigenvector of the
adjacency matrix so as to assign to each vertex the values of the components of the first eigenvector as a centrality
score:
https://en.wikipedia.org/wiki/Eigenvector_centrality
On the graph of webpages and links those values are called the PageRank scores by Google.
The goal of this example is to analyze the graph of links inside wikipedia articles to rank articles by relative importance
according to this eigenvector centrality.
The traditional way to compute the principal eigenvector is to use the power iteration method:
https://en.wikipedia.org/wiki/Power_iteration
Here the computation is achieved thanks to Martinsson’s Randomized SVD algorithm implemented in scikit-learn.
The graph data is fetched from the DBpedia dumps. DBpedia is an extraction of the latent structured data of the
Wikipedia content.


from bz2 import BZ2File

import os
from datetime import datetime
from pprint import pprint
import numpy as np
from joblib import Memory
from sklearn.decomposition import randomized_svd

from sklearn.externals.six.moves.urllib.request import urlopen
from sklearn.externals.six import iteritems
print(__doc__)
# #############################################################################
# Where to download the data, if not already on disk
redirects_url = "http://downloads.dbpedia.org/3.5.1/en/redirects_en.nt.bz2"
redirects_filename = redirects_url.rsplit("/", 1)[1]
page_links_url = "http://downloads.dbpedia.org/3.5.1/en/page_links_en.nt.bz2"
page_links_filename = page_links_url.rsplit("/", 1)[1]
resources = [
(redirects_url, redirects_filename),
(page_links_url, page_links_filename),
]
for url, filename in resources:

if not os.path.exists(filename):
print("Downloading data from '%s', please wait..." % url)
opener = urlopen(url)
open(filename, 'wb').write(opener.read())
print()
# #############################################################################
# Loading the redirect files
memory = Memory(cachedir=".")
def index(redirects, index_map, k):

"""Find the index of an article name after redirect resolution"""
k = redirects.get(k, k)
return index_map.setdefault(k, len(index_map))
DBPEDIA_RESOURCE_PREFIX_LEN = len("http://dbpedia.org/resource/")
SHORTNAME_SLICE = slice(DBPEDIA_RESOURCE_PREFIX_LEN + 1, -1)
def short_name(nt_uri):
"""Remove the < and > URI markers and the common URI prefix"""

return nt_uri[SHORTNAME_SLICE]
def get_redirects(redirects_filename):
"""Parse the redirections and build a transitively closed map out of it"""
redirects = {}
print("Parsing the NT redirect file")
for l, line in enumerate(BZ2File(redirects_filename)):
split = line.split()
if len(split) != 4:
print("ignoring malformed line: " + line)
continue
redirects[short_name(split[0])] = short_name(split[2])
if l % 1000000 == 0:
print("[%s] line: %08d" % (datetime.now().isoformat(), l))
# compute the transitive closure

print("Computing the transitive closure of the redirect relation")
for l, source in enumerate(redirects.keys()):
transitive_target = None
target = redirects[source]
seen = set([source])
while True:
transitive_target = target
target = redirects.get(target)
if target is None or target in seen:
break
seen.add(target)
redirects[source] = transitive_target
if l % 1000000 == 0:
return redirects
# disabling joblib as the pickling of large dicts seems much too slow
#@memory.cache
def get_adjacency_matrix(redirects_filename, page_links_filename, limit=None):
"""Extract the adjacency graph as a scipy sparse matrix
Redirects are resolved first.
Returns X, the scipy sparse adjacency matrix, redirects as python

dict from article names to article names and index_map a python dict
from article names to python int (article indexes).
"""
print("Computing the redirect map")

redirects = get_redirects(redirects_filename)
print("Computing the integer index map")

index_map = dict()
links = list()
for l, line in enumerate(BZ2File(page_links_filename)):
split = line.split()
if len(split) != 4:
print("ignoring malformed line: " + line)
continue

i = index(redirects, index_map, short_name(split[0]))

j = index(redirects, index_map, short_name(split[2]))
links.append((i, j))
if l % 1000000 == 0:
if limit is not None and l >= limit - 1:

break
print("Computing the adjacency matrix")

X = sparse.lil_matrix((len(index_map), len(index_map)), dtype=np.float32)
for i, j in links:
X[i, j] = 1.0
del links
print("Converting to CSR representation")
X = X.tocsr()
print("CSR conversion done")
return X, redirects, index_map
# stop after 5M links to make it possible to work in RAM

X, redirects, index_map = get_adjacency_matrix(
redirects_filename, page_links_filename, limit=5000000)
names = dict((i, name) for name, i in iteritems(index_map))
print("Computing the principal singular vectors using randomized_svd")

t0 = time()
U, s, V = randomized_svd(X, 5, n_iter=3)
# print the names of the wikipedia related strongest components of the

# principal singular vector which should be similar to the highest eigenvector
print("Top wikipedia pages according to principal singular vectors")
pprint([names[i] for i in np.abs(U.T[0]).argsort()[-10:]])
pprint([names[i] for i in np.abs(V[0]).argsort()[-10:]])
def centrality_scores(X, alpha=0.85, max_iter=100, tol=1e-10):

"""Power iteration computation of the principal eigenvector
This method is also known as Google PageRank and the implementation

is based on the one from the NetworkX project (BSD licensed too)
with copyrights by:
Aric Hagberg <hagberg@lanl.gov>

Dan Schult <dschult@colgate.edu>
Pieter Swart <swart@lanl.gov>
"""
n = X.shape[0]
X = X.copy()
incoming_counts = np.asarray(X.sum(axis=1)).ravel()
print("Normalizing the graph")

for i in incoming_counts.nonzero()[0]:
X.data[X.indptr[i]:X.indptr[i + 1]] *= 1.0 / incoming_counts[i]
dangle = np.asarray(np.where(np.isclose(X.sum(axis=1), 0),
1.0 / n, 0)).ravel()

scores = np.full(n, 1. / n, dtype=np.float32) # initial guess

for i in range(max_iter):
print("power iteration #%d" % i)
prev_scores = scores
scores = (alpha * (scores * X + np.dot(dangle, prev_scores))
+ (1 - alpha) * prev_scores.sum() / n)
# check convergence: normalized l_inf norm
scores_max = np.abs(scores).max()
if scores_max == 0.0:
scores_max = 1.0
err = np.abs(scores - prev_scores).max() / scores_max
print("error: %0.6f" % err)
if err < n * tol:
return scores
return scores
print("Computing principal eigenvector score using a power iteration method")

t0 = time()
scores = centrality_scores(X, max_iter=100, tol=1e-10)
pprint([names[i] for i in np.abs(scores).argsort()[-10:]])
5.2.9 Libsvm GUI
A simple graphical frontend for Libsvm mainly intended for didactic purposes. You can create data points by point
and click and visualize the decision region induced by different kernels and parameter settings.
To create positive examples click the left mouse button; to create negative examples click the right button.
If all examples are from the same class, it uses a one-class SVM.
from __future__ import division, print_function
print(__doc__)
# Author: Peter Prettenhoer <peter.prettenhofer@gmail.com>

#
import matplotlib
matplotlib.use('TkAgg')
from matplotlib.backends.backend_tkagg import FigureCanvasTkAgg

from matplotlib.backends.backend_tkagg import NavigationToolbar2TkAgg
from matplotlib.figure import Figure
from matplotlib.contour import ContourSet
try:
import tkinter as Tk
except ImportError:

# Backward compat for Python 2

import Tkinter as Tk
import sys
import numpy as np

from sklearn.datasets import dump_svmlight_file
from sklearn.externals.six.moves import xrange
y_min, y_max = -50, 50

x_min, x_max = -50, 50
class Model(object):
"""The Model which hold the data. It implements the
observable in the observer pattern and notifies the
registered observers on change event.
"""
def __init__(self):
self.observers = []
self.surface = None
self.data = []
self.cls = None
self.surface_type = 0
def changed(self, event):

"""Notify the observers. """
for observer in self.observers:
observer.update(event, self)
def add_observer(self, observer):

"""Register an observer. """
self.observers.append(observer)
def set_surface(self, surface):

self.surface = surface
def dump_svmlight_file(self, file):

data = np.array(self.data)
X = data[:, 0:2]
y = data[:, 2]
dump_svmlight_file(X, y, file)
class Controller(object):
def __init__(self, model):
self.model = model
self.kernel = Tk.IntVar()
self.surface_type = Tk.IntVar()
# Whether or not a model has been fitted
self.fitted = False
def fit(self):
print("fit the model")
train = np.array(self.model.data)
X = train[:, 0:2]

y = train[:, 2]
C = float(self.complexity.get())
gamma = float(self.gamma.get())
coef0 = float(self.coef0.get())
degree = int(self.degree.get())
kernel_map = {0: "linear", 1: "rbf", 2: "poly"}
if len(np.unique(y)) == 1:
clf = svm.OneClassSVM(kernel=kernel_map[self.kernel.get()],
gamma=gamma, coef0=coef0, degree=degree)
clf.fit(X)
else:
clf = svm.SVC(kernel=kernel_map[self.kernel.get()], C=C,
gamma=gamma, coef0=coef0, degree=degree)
clf.fit(X, y)
if hasattr(clf, 'score'):
print("Accuracy:", clf.score(X, y) * 100)
X1, X2, Z = self.decision_surface(clf)
self.model.clf = clf
self.model.set_surface((X1, X2, Z))
self.model.surface_type = self.surface_type.get()
self.fitted = True
self.model.changed("surface")
def decision_surface(self, cls):

delta = 1
x = np.arange(x_min, x_max + delta, delta)
y = np.arange(y_min, y_max + delta, delta)
X1, X2 = np.meshgrid(x, y)
Z = cls.decision_function(np.c_[X1.ravel(), X2.ravel()])
Z = Z.reshape(X1.shape)
return X1, X2, Z
def clear_data(self):
self.model.data = []
self.fitted = False
self.model.changed("clear")
def add_example(self, x, y, label):

self.model.data.append((x, y, label))
self.model.changed("example_added")
# update decision surface if already fitted.

self.refit()
def refit(self):
"""Refit the model if already fitted. """
if self.fitted:
self.fit()
class View(object):
"""Test docstring. """
def __init__(self, root, controller):
f = Figure()
ax = f.add_subplot(111)
ax.set_xticks([])
ax.set_yticks([])

ax.set_xlim((x_min, x_max))
ax.set_ylim((y_min, y_max))
canvas = FigureCanvasTkAgg(f, master=root)
canvas.show()
canvas.get_tk_widget().pack(side=Tk.TOP, fill=Tk.BOTH, expand=1)
canvas._tkcanvas.pack(side=Tk.TOP, fill=Tk.BOTH, expand=1)
canvas.mpl_connect('button_press_event', self.onclick)
toolbar = NavigationToolbar2TkAgg(canvas, root)
toolbar.update()
self.controllbar = ControllBar(root, controller)
self.f = f
self.ax = ax
self.canvas = canvas
self.controller = controller
self.contours = []
self.c_labels = None
self.plot_kernels()
def plot_kernels(self):
self.ax.text(-50, -60, "Linear: $u^T v$")
self.ax.text(-20, -60, r"RBF: $\exp (-\gamma \| u-v \|^2)$")
self.ax.text(10, -60, r"Poly: $(\gamma \, u^T v + r)^d$")
def onclick(self, event):

if event.xdata and event.ydata:
if event.button == 1:
self.controller.add_example(event.xdata, event.ydata, 1)
elif event.button == 3:
self.controller.add_example(event.xdata, event.ydata, -1)
def update_example(self, model, idx):

x, y, l = model.data[idx]
if l == 1:
color = 'w'
elif l == -1:
color = 'k'
self.ax.plot([x], [y], "%so" % color, scalex=0.0, scaley=0.0)
def update(self, event, model):

if event == "examples_loaded":
for i in xrange(len(model.data)):
self.update_example(model, i)
if event == "example_added":
self.update_example(model, -1)
if event == "clear":
self.ax.clear()
self.ax.set_xticks([])
self.ax.set_yticks([])
self.contours = []
self.c_labels = None
self.plot_kernels()
if event == "surface":
self.remove_surface()
self.plot_support_vectors(model.clf.support_vectors_)
self.plot_decision_surface(model.surface, model.surface_type)

self.canvas.draw()
def remove_surface(self):
"""Remove old decision surface."""
if len(self.contours) > 0:
for contour in self.contours:
if isinstance(contour, ContourSet):
for lineset in contour.collections:
lineset.remove()
else:
contour.remove()
self.contours = []
def plot_support_vectors(self, support_vectors):

"""Plot the support vectors by placing circles over the
corresponding data points and adds the circle collection
to the contours list."""
cs = self.ax.scatter(support_vectors[:, 0], support_vectors[:, 1],
s=80, edgecolors="k", facecolors="none")
self.contours.append(cs)
def plot_decision_surface(self, surface, type):

X1, X2, Z = surface
if type == 0:
levels = [-1.0, 0.0, 1.0]
linestyles = ['dashed', 'solid', 'dashed']
colors = 'k'
self.contours.append(self.ax.contour(X1, X2, Z, levels,
colors=colors,
linestyles=linestyles))
elif type == 1:
self.contours.append(self.ax.contourf(X1, X2, Z, 10,
cmap=matplotlib.cm.bone,
origin='lower', alpha=0.85))
self.contours.append(self.ax.contour(X1, X2, Z, [0.0], colors='k',
linestyles=['solid']))
else:
raise ValueError("surface type unknown")
class ControllBar(object):
def __init__(self, root, controller):
fm = Tk.Frame(root)
kernel_group = Tk.Frame(fm)
Tk.Radiobutton(kernel_group, text="Linear", variable=controller.kernel,
value=0, command=controller.refit).pack(anchor=Tk.W)
Tk.Radiobutton(kernel_group, text="RBF", variable=controller.kernel,
Tk.Radiobutton(kernel_group, text="Poly", variable=controller.kernel,
kernel_group.pack(side=Tk.LEFT)
valbox = Tk.Frame(fm)
controller.complexity = Tk.StringVar()
controller.complexity.set("1.0")
c = Tk.Frame(valbox)
Tk.Label(c, text="C:", anchor="e", width=7).pack(side=Tk.LEFT)

Tk.Entry(c, width=6, textvariable=controller.complexity).pack(

side=Tk.LEFT)
c.pack()
controller.gamma = Tk.StringVar()
controller.gamma.set("0.01")
g = Tk.Frame(valbox)
Tk.Label(g, text="gamma:", anchor="e", width=7).pack(side=Tk.LEFT)
Tk.Entry(g, width=6, textvariable=controller.gamma).pack(side=Tk.LEFT)
g.pack()
controller.degree = Tk.StringVar()
controller.degree.set("3")
d = Tk.Frame(valbox)
Tk.Label(d, text="degree:", anchor="e", width=7).pack(side=Tk.LEFT)
Tk.Entry(d, width=6, textvariable=controller.degree).pack(side=Tk.LEFT)
d.pack()
controller.coef0 = Tk.StringVar()
controller.coef0.set("0")
r = Tk.Frame(valbox)
Tk.Label(r, text="coef0:", anchor="e", width=7).pack(side=Tk.LEFT)
Tk.Entry(r, width=6, textvariable=controller.coef0).pack(side=Tk.LEFT)
r.pack()
valbox.pack(side=Tk.LEFT)
cmap_group = Tk.Frame(fm)
Tk.Radiobutton(cmap_group, text="Hyperplanes",
variable=controller.surface_type, value=0,
command=controller.refit).pack(anchor=Tk.W)
Tk.Radiobutton(cmap_group, text="Surface",
variable=controller.surface_type, value=1,
command=controller.refit).pack(anchor=Tk.W)
cmap_group.pack(side=Tk.LEFT)
train_button = Tk.Button(fm, text='Fit', width=5,

command=controller.fit)
train_button.pack()
fm.pack(side=Tk.LEFT)
Tk.Button(fm, text='Clear', width=5,
command=controller.clear_data).pack(side=Tk.LEFT)
def get_parser():
from optparse import OptionParser
op = OptionParser()
op.add_option("--output",
action="store", type="str", dest="output",
help="Path where to dump data.")
return op
def main(argv):
op = get_parser()
opts, args = op.parse_args(argv[1:])
root = Tk.Tk()
model = Model()

controller = Controller(model)
root.wm_title("Scikit-learn Libsvm GUI")
view = View(root, controller)
model.add_observer(view)
Tk.mainloop()
if opts.output:
model.dump_svmlight_file(opts.output)
if __name__ == "__main__":
main(sys.argv)
5.2.10 Prediction Latency
This is an example showing the prediction latency of various scikit-learn estimators.

The goal is to measure the latency one can expect when doing predictions either in bulk or atomic (i.e. one by one)
mode.
The plots represent the distribution of the prediction latency as a boxplot.


•
Out:
Benchmarking SGDRegressor(alpha=0.01, average=False, early_stopping=False, epsilon=0.

˓→1,
eta0=0.01, fit_intercept=True, l1_ratio=0.25,

learning_rate='invscaling', loss='squared_loss', max_iter=None,
n_iter=None, n_iter_no_change=5, penalty='elasticnet', power_t=0.25,
random_state=None, shuffle=True, tol=0.0001,
Benchmarking RandomForestRegressor(bootstrap=True, criterion='mse', max_depth=None,
max_features='auto', max_leaf_nodes=None,
oob_score=False, random_state=None, verbose=0, warm_start=False)
Benchmarking SVR(C=1.0, cache_size=200, coef0=0.0, degree=3, epsilon=0.1,
gamma='auto_deprecated', kernel='rbf', max_iter=-1, shrinking=True,
benchmarking with 100 features
example run in 9.77s
# Authors: Eustache Diemert <eustache@diemert.fr>


from collections import defaultdict

import time
import gc
import numpy as np

from sklearn.datasets.samples_generator import make_regression
from sklearn.ensemble.forest import RandomForestRegressor
from sklearn.linear_model.ridge import Ridge
from sklearn.linear_model.stochastic_gradient import SGDRegressor
from sklearn.svm.classes import SVR
def _not_in_sphinx():
# Hack to detect whether we are running by the sphinx builder
return '__file__' in globals()
def atomic_benchmark_estimator(estimator, X_test, verbose=False):

"""Measure runtime prediction of each instance."""
n_instances = X_test.shape[0]
runtimes = np.zeros(n_instances, dtype=np.float)
for i in range(n_instances):
instance = X_test[[i], :]
start = time.time()
estimator.predict(instance)
runtimes[i] = time.time() - start
if verbose:
print("atomic_benchmark runtimes:", min(runtimes), np.percentile(
runtimes, 50), max(runtimes))
return runtimes
def bulk_benchmark_estimator(estimator, X_test, n_bulk_repeats, verbose):

"""Measure runtime prediction of the whole input."""
n_instances = X_test.shape[0]
runtimes = np.zeros(n_bulk_repeats, dtype=np.float)
for i in range(n_bulk_repeats):
start = time.time()
estimator.predict(X_test)
runtimes[i] = time.time() - start
runtimes = np.array(list(map(lambda x: x / float(n_instances), runtimes)))
if verbose:
print("bulk_benchmark runtimes:", min(runtimes), np.percentile(
runtimes, 50), max(runtimes))
return runtimes
def benchmark_estimator(estimator, X_test, n_bulk_repeats=30, verbose=False):

"""
Measure runtimes of prediction in both atomic and bulk mode.
Parameters
----------
estimator : already trained estimator supporting `predict()`

X_test : test input

n_bulk_repeats : how many times to repeat when evaluating bulk mode
Returns
-------
atomic_runtimes, bulk_runtimes : a pair of `np.array` which contain the
runtimes in seconds.
"""
atomic_runtimes = atomic_benchmark_estimator(estimator, X_test, verbose)
bulk_runtimes = bulk_benchmark_estimator(estimator, X_test, n_bulk_repeats,
verbose)
return atomic_runtimes, bulk_runtimes
def generate_dataset(n_train, n_test, n_features, noise=0.1, verbose=False):

"""Generate a regression dataset with the given parameters."""
if verbose:
print("generating dataset...")
X, y, coef = make_regression(n_samples=n_train + n_test,

n_features=n_features, noise=noise, coef=True)
random_seed = 13
X, y, train_size=n_train, test_size=n_test, random_state=random_seed)
X_train, y_train = shuffle(X_train, y_train, random_state=random_seed)
X_scaler = StandardScaler()
X_train = X_scaler.fit_transform(X_train)
X_test = X_scaler.transform(X_test)
y_scaler = StandardScaler()
y_train = y_scaler.fit_transform(y_train[:, None])[:, 0]
y_test = y_scaler.transform(y_test[:, None])[:, 0]
gc.collect()
if verbose:
print("ok")
return X_train, y_train, X_test, y_test
def boxplot_runtimes(runtimes, pred_type, configuration):

"""
Plot a new `Figure` with boxplots of prediction runtimes.
Parameters
----------
runtimes : list of `np.array` of latencies in micro-seconds
cls_names : list of estimator class names that generated the runtimes
pred_type : 'bulk' or 'atomic'
"""
fig, ax1 = plt.subplots(figsize=(10, 6))

bp = plt.boxplot(runtimes, )
cls_infos = ['%s\n(%d %s)' % (estimator_conf['name'],

estimator_conf['complexity_computer'](
estimator_conf['instance']),
estimator_conf['complexity_label']) for
estimator_conf in configuration['estimators']]
plt.setp(ax1, xticklabels=cls_infos)
plt.setp(bp['boxes'], color='black')
plt.setp(bp['whiskers'], color='black')
plt.setp(bp['fliers'], color='red', marker='+')
ax1.yaxis.grid(True, linestyle='-', which='major', color='lightgrey',

alpha=0.5)
ax1.set_axisbelow(True)
ax1.set_title('Prediction Time per Instance - %s, %d feats.' % (
pred_type.capitalize(),
configuration['n_features']))
ax1.set_ylabel('Prediction Time (us)')
plt.show()
def benchmark(configuration):
"""Run the whole benchmark."""
X_train, y_train, X_test, y_test = generate_dataset(
configuration['n_train'], configuration['n_test'],
configuration['n_features'])
stats = {}
for estimator_conf in configuration['estimators']:
print("Benchmarking", estimator_conf['instance'])
estimator_conf['instance'].fit(X_train, y_train)
gc.collect()
a, b = benchmark_estimator(estimator_conf['instance'], X_test)
stats[estimator_conf['name']] = {'atomic': a, 'bulk': b}
cls_names = [estimator_conf['name'] for estimator_conf in configuration[

'estimators']]
runtimes = [1e6 * stats[clf_name]['atomic'] for clf_name in cls_names]
boxplot_runtimes(runtimes, 'atomic', configuration)
runtimes = [1e6 * stats[clf_name]['bulk'] for clf_name in cls_names]
boxplot_runtimes(runtimes, 'bulk (%d)' % configuration['n_test'],
configuration)
def n_feature_influence(estimators, n_train, n_test, n_features, percentile):

"""
Estimate influence of the number of features on prediction time.
Parameters
----------
estimators : dict of (name (str), estimator) to benchmark

n_train : nber of training instances (int)
n_test : nber of testing instances (int)
n_features : list of feature-space dimensionality to test (int)
percentile : percentile at which to measure the speed (int [0-100])
Returns:

--------
percentiles : dict(estimator_name,
dict(n_features, percentile_perf_in_us))
"""
percentiles = defaultdict(defaultdict)
for n in n_features:
print("benchmarking with %d features" % n)
X_train, y_train, X_test, y_test = generate_dataset(n_train, n_test, n)
for cls_name, estimator in estimators.items():
gc.collect()
runtimes = bulk_benchmark_estimator(estimator, X_test, 30, False)
percentiles[cls_name][n] = 1e6 * np.percentile(runtimes,
percentile)
return percentiles
def plot_n_features_influence(percentiles, percentile):

fig, ax1 = plt.subplots(figsize=(10, 6))
colors = ['r', 'g', 'b']
for i, cls_name in enumerate(percentiles.keys()):
x = np.array(sorted([n for n in percentiles[cls_name].keys()]))
y = np.array([percentiles[cls_name][n] for n in x])
plt.plot(x, y, color=colors[i], )
ax1.yaxis.grid(True, linestyle='-', which='major', color='lightgrey',
alpha=0.5)
ax1.set_axisbelow(True)
ax1.set_title('Evolution of Prediction Time with #Features')
ax1.set_xlabel('#Features')
ax1.set_ylabel('Prediction Time at %d%%-ile (us)' % percentile)
plt.show()
def benchmark_throughputs(configuration, duration_secs=0.1):

"""benchmark throughput for different estimators."""
X_train, y_train, X_test, y_test = generate_dataset(
configuration['n_train'], configuration['n_test'],
configuration['n_features'])
throughputs = dict()
for estimator_config in configuration['estimators']:
estimator_config['instance'].fit(X_train, y_train)
n_predictions = 0
while (time.time() - start_time) < duration_secs:
estimator_config['instance'].predict(X_test[[0]])
n_predictions += 1
throughputs[estimator_config['name']] = n_predictions / duration_secs
return throughputs
def plot_benchmark_throughput(throughputs, configuration):

fig, ax = plt.subplots(figsize=(10, 6))
colors = ['r', 'g', 'b']
cls_infos = ['%s\n(%d %s)' % (estimator_conf['name'],
estimator_conf['complexity_computer'](
estimator_conf['instance']),

estimator_conf['complexity_label']) for
estimator_conf in configuration['estimators']]
cls_values = [throughputs[estimator_conf['name']] for estimator_conf in
configuration['estimators']]
plt.bar(range(len(throughputs)), cls_values, width=0.5, color=colors)
ax.set_xticks(np.linspace(0.25, len(throughputs) - 0.75, len(throughputs)))
ax.set_xticklabels(cls_infos, fontsize=10)
ymax = max(cls_values) * 1.2
ax.set_ylim((0, ymax))
ax.set_ylabel('Throughput (predictions/sec)')
ax.set_title('Prediction Throughput for different estimators (%d '
'features)' % configuration['n_features'])
plt.show()
# #############################################################################
# Main code
# #############################################################################
# Benchmark bulk/atomic prediction speed for various regressors
configuration = {
'n_train': int(1e3),
'n_test': int(1e2),
'n_features': int(1e2),
'estimators': [
{'name': 'Linear Model',
'instance': SGDRegressor(penalty='elasticnet', alpha=0.01,
l1_ratio=0.25, fit_intercept=True,
tol=1e-4),
'complexity_label': 'non-zero coefficients',
'complexity_computer': lambda clf: np.count_nonzero(clf.coef_)},
{'name': 'RandomForest',
'instance': RandomForestRegressor(n_estimators=100),
'complexity_label': 'estimators',
'complexity_computer': lambda clf: clf.n_estimators},
{'name': 'SVR',
'instance': SVR(kernel='rbf'),
'complexity_label': 'support vectors',
'complexity_computer': lambda clf: len(clf.support_vectors_)},
]
}
benchmark(configuration)
# benchmark n_features influence on prediction speed

percentile = 90
percentiles = n_feature_influence({'ridge': Ridge()},
configuration['n_train'],
configuration['n_test'],
[100, 250, 500], percentile)
plot_n_features_influence(percentiles, percentile)
# benchmark throughput
throughputs = benchmark_throughputs(configuration)
plot_benchmark_throughput(throughputs, configuration)
stop_time = time.time()

print("example run in %.2fs" % (stop_time - start_time))
5.2.11 Out-of-core classification of text documents
This is an example showing how scikit-learn can be used for classification using an out-of-core approach: learning
from data that doesn’t fit into main memory. We make use of an online classifier, i.e., one that supports the partial_fit
method, that will be fed with batches of examples. To guarantee that the features space remains the same over time
we leverage a HashingVectorizer that will project each example into the same feature space. This is especially useful
in the case of text classification where new features (words) may appear in each batch.
The dataset used in this example is Reuters-21578 as provided by the UCI ML repository. It will be automatically
downloaded and uncompressed on first run.
The plot represents the learning curve of the classifier: the evolution of classification accuracy over the course of the
mini-batches. Accuracy is measured on the first 1000 samples, held out as a validation set.
To limit the memory consumption, we queue examples up to a fixed amount before feeding them to the learner.
# Authors: Eustache Diemert <eustache@diemert.fr>

# @FedericoV <https://github.com/FedericoV/>

from glob import glob
import itertools
import os.path
import re
import tarfile
import time
import sys
import numpy as np
from matplotlib import rcParams
from sklearn.externals.six.moves import html_parser

from sklearn.externals.six.moves.urllib.request import urlretrieve
from sklearn.datasets import get_data_home
from sklearn.feature_extraction.text import HashingVectorizer
from sklearn.linear_model import PassiveAggressiveClassifier
from sklearn.linear_model import Perceptron
from sklearn.naive_bayes import MultinomialNB
def _not_in_sphinx():
# Hack to detect whether we are running by the sphinx builder
return '__file__' in globals()

Reuters Dataset related routines
class ReutersParser(html_parser.HTMLParser):
"""Utility class to parse a SGML file and yield documents one at a time."""
def __init__(self, encoding='latin-1'):

html_parser.HTMLParser.__init__(self)
self._reset()
self.encoding = encoding
def handle_starttag(self, tag, attrs):

method = 'start_' + tag
getattr(self, method, lambda x: None)(attrs)
def handle_endtag(self, tag):

method = 'end_' + tag
getattr(self, method, lambda: None)()
def _reset(self):
self.in_title = 0
self.in_body = 0
self.in_topics = 0
self.in_topic_d = 0
self.title = ""
self.body = ""
self.topics = []
self.topic_d = ""
def parse(self, fd):

self.docs = []
for chunk in fd:
self.feed(chunk.decode(self.encoding))
for doc in self.docs:
yield doc
self.docs = []
self.close()
def handle_data(self, data):

if self.in_body:
self.body += data
elif self.in_title:
self.title += data
elif self.in_topic_d:
self.topic_d += data
def start_reuters(self, attributes):

pass
def end_reuters(self):
self.body = re.sub(r'\s+', r' ', self.body)
self.docs.append({'title': self.title,
'body': self.body,
'topics': self.topics})
self._reset()
def start_title(self, attributes):

self.in_title = 1

def end_title(self):
self.in_title = 0
def start_body(self, attributes):

self.in_body = 1
def end_body(self):
self.in_body = 0
def start_topics(self, attributes):

self.in_topics = 1
def end_topics(self):
self.in_topics = 0
def start_d(self, attributes):

self.in_topic_d = 1
def end_d(self):
self.in_topic_d = 0
self.topics.append(self.topic_d)
self.topic_d = ""
def stream_reuters_documents(data_path=None):
"""Iterate over documents of the Reuters dataset.
The Reuters archive will automatically be downloaded and uncompressed if

the `data_path` directory does not exist.
Documents are represented as dictionaries with 'body' (str),

'title' (str), 'topics' (list(str)) keys.
"""
DOWNLOAD_URL = ('http://archive.ics.uci.edu/ml/machine-learning-databases/'
'reuters21578-mld/reuters21578.tar.gz')
ARCHIVE_FILENAME = 'reuters21578.tar.gz'
if data_path is None:
data_path = os.path.join(get_data_home(), "reuters")
if not os.path.exists(data_path):
"""Download the dataset."""
print("downloading dataset (once and for all) into %s" %
data_path)
os.mkdir(data_path)
def progress(blocknum, bs, size):

total_sz_mb = '%.2f MB' % (size / 1e6)
current_sz_mb = '%.2f MB' % ((blocknum * bs) / 1e6)
if _not_in_sphinx():
sys.stdout.write(
'\rdownloaded %s / %s' % (current_sz_mb, total_sz_mb))
archive_path = os.path.join(data_path, ARCHIVE_FILENAME)

urlretrieve(DOWNLOAD_URL, filename=archive_path,
reporthook=progress)
if _not_in_sphinx():

sys.stdout.write('\r')
print("untarring Reuters dataset...")
tarfile.open(archive_path, 'r:gz').extractall(data_path)
print("done.")
parser = ReutersParser()
for filename in glob(os.path.join(data_path, "*.sgm")):
for doc in parser.parse(open(filename, 'rb')):
yield doc
Main
Create the vectorizer and limit the number of features to a reasonable maximum
vectorizer = HashingVectorizer(decode_error='ignore', n_features=2 ** 18,
alternate_sign=False)
# Iterator over parsed Reuters SGML files.

data_stream = stream_reuters_documents()
# We learn a binary classification between the "acq" class and all the others.
# "acq" was chosen as it is more or less evenly distributed in the Reuters
# files. For other datasets, one should take care of creating a test set with
# a realistic portion of positive instances.
all_classes = np.array([0, 1])
positive_class = 'acq'
# Here are some classifiers that support the `partial_fit` method

partial_fit_classifiers = {
'SGD': SGDClassifier(max_iter=5),
'Perceptron': Perceptron(tol=1e-3),
'NB Multinomial': MultinomialNB(alpha=0.01),
'Passive-Aggressive': PassiveAggressiveClassifier(tol=1e-3),
}
def get_minibatch(doc_iter, size, pos_class=positive_class):

"""Extract a minibatch of examples, return a tuple X_text, y.
Note: size is before excluding invalid docs with no topics assigned.
"""
data = [(u'{title}\n\n{body}'.format(**doc), pos_class in doc['topics'])
for doc in itertools.islice(doc_iter, size)
if doc['topics']]
if not len(data):
return np.asarray([], dtype=int), np.asarray([], dtype=int)
X_text, y = zip(*data)
return X_text, np.asarray(y, dtype=int)
def iter_minibatches(doc_iter, minibatch_size):

"""Generator of minibatches."""
X_text, y = get_minibatch(doc_iter, minibatch_size)
while len(X_text):
yield X_text, y

X_text, y = get_minibatch(doc_iter, minibatch_size)
# test data statistics

test_stats = {'n_test': 0, 'n_test_pos': 0}
# First we hold out a number of examples to estimate accuracy

n_test_documents = 1000
tick = time.time()
X_test_text, y_test = get_minibatch(data_stream, 1000)
parsing_time = time.time() - tick
tick = time.time()
X_test = vectorizer.transform(X_test_text)
vectorizing_time = time.time() - tick
test_stats['n_test'] += len(y_test)
test_stats['n_test_pos'] += sum(y_test)
print("Test set is %d documents (%d positive)" % (len(y_test), sum(y_test)))
def progress(cls_name, stats):

"""Report progress information, return a string."""
duration = time.time() - stats['t0']
s = "%20s classifier : \t" % cls_name
s += "%(n_train)6d train docs (%(n_train_pos)6d positive) " % stats
s += "%(n_test)6d test docs (%(n_test_pos)6d positive) " % test_stats
s += "accuracy: %(accuracy).3f " % stats
s += "in %.2fs (%5d docs/s)" % (duration, stats['n_train'] / duration)
return s
cls_stats = {}
for cls_name in partial_fit_classifiers:

stats = {'n_train': 0, 'n_train_pos': 0,
'accuracy': 0.0, 'accuracy_history': [(0, 0)], 't0': time.time(),
'runtime_history': [(0, 0)], 'total_fit_time': 0.0}
cls_stats[cls_name] = stats
get_minibatch(data_stream, n_test_documents)
# Discard test set
# We will feed the classifier with mini-batches of 1000 documents; this means
# we have at most 1000 docs in memory at any time. The smaller the document
# batch, the bigger the relative overhead of the partial fit methods.
minibatch_size = 1000
# Create the data_stream that parses Reuters SGML files and iterates on
# documents as a stream.
minibatch_iterators = iter_minibatches(data_stream, minibatch_size)
total_vect_time = 0.0
# Main loop : iterate on mini-batches of examples

for i, (X_train_text, y_train) in enumerate(minibatch_iterators):
tick = time.time()
X_train = vectorizer.transform(X_train_text)
total_vect_time += time.time() - tick

for cls_name, cls in partial_fit_classifiers.items():

tick = time.time()
# update estimator with examples in the current mini-batch
cls.partial_fit(X_train, y_train, classes=all_classes)
# accumulate test accuracy stats

cls_stats[cls_name]['total_fit_time'] += time.time() - tick
cls_stats[cls_name]['n_train'] += X_train.shape[0]
cls_stats[cls_name]['n_train_pos'] += sum(y_train)
tick = time.time()
cls_stats[cls_name]['accuracy'] = cls.score(X_test, y_test)
cls_stats[cls_name]['prediction_time'] = time.time() - tick
acc_history = (cls_stats[cls_name]['accuracy'],
cls_stats[cls_name]['n_train'])
cls_stats[cls_name]['accuracy_history'].append(acc_history)
run_history = (cls_stats[cls_name]['accuracy'],
total_vect_time + cls_stats[cls_name]['total_fit_time'])
cls_stats[cls_name]['runtime_history'].append(run_history)
if i % 3 == 0:
print(progress(cls_name, cls_stats[cls_name]))
if i % 3 == 0:
print('\n')
Out:
downloading dataset (once and for all) into /home/circleci/scikit_learn_data/reuters

untarring Reuters dataset...
done.
Test set is 988 documents (122 positive)
SGD classifier : 928 train docs ( 142 positive) 988
˓→test docs ( 122 positive) accuracy: 0.919 in 0.78s ( 1193 docs/s)
Perceptron classifier : 928 train docs ( 142 positive) 988
NB Multinomial classifier : 928 train docs ( 142 positive) 988
Passive-Aggressive classifier : 928 train docs ( 142 positive) 988







Plot results
def plot_accuracy(x, y, x_legend):

"""Plot accuracy as a function of x."""
x = np.array(x)
y = np.array(y)
plt.title('Classification accuracy as a function of %s' % x_legend)
plt.xlabel('%s' % x_legend)
plt.ylabel('Accuracy')
plt.grid(True)
plt.plot(x, y)
rcParams['legend.fontsize'] = 10
cls_names = list(sorted(cls_stats.keys()))

# Plot accuracy evolution

plt.figure()
for _, stats in sorted(cls_stats.items()):
# Plot accuracy evolution with #examples
accuracy, n_examples = zip(*stats['accuracy_history'])
plot_accuracy(n_examples, accuracy, "training examples (#)")
ax = plt.gca()
ax.set_ylim((0.8, 1))
plt.legend(cls_names, loc='best')
plt.figure()
for _, stats in sorted(cls_stats.items()):
# Plot accuracy evolution with runtime
accuracy, runtime = zip(*stats['runtime_history'])
plot_accuracy(runtime, accuracy, 'runtime (s)')
ax = plt.gca()
ax.set_ylim((0.8, 1))
plt.legend(cls_names, loc='best')
# Plot fitting times

plt.figure()
fig = plt.gcf()
cls_runtime = []
for cls_name, stats in sorted(cls_stats.items()):
cls_runtime.append(stats['total_fit_time'])
cls_runtime.append(total_vect_time)
cls_names.append('Vectorization')
bar_colors = ['b', 'g', 'r', 'c', 'm', 'y']
ax = plt.subplot(111)
rectangles = plt.bar(range(len(cls_names)), cls_runtime, width=0.5,
color=bar_colors)
ax.set_xticks(np.linspace(0, len(cls_names) - 1, len(cls_names)))

ax.set_xticklabels(cls_names, fontsize=10)
ymax = max(cls_runtime) * 1.2
ax.set_ylabel('runtime (s)')
ax.set_title('Training Times')
def autolabel(rectangles):
"""attach some text vi autolabel on rectangles."""
for rect in rectangles:
height = rect.get_height()
ax.text(rect.get_x() + rect.get_width() / 2.,
1.05 * height, '%.4f' % height,
ha='center', va='bottom')
plt.setp(plt.xticks()[1], rotation=30)
autolabel(rectangles)
plt.tight_layout()
plt.show()
# Plot prediction times

plt.figure()
cls_runtime = []
cls_names = list(sorted(cls_stats.keys()))
for cls_name, stats in sorted(cls_stats.items()):
cls_runtime.append(stats['prediction_time'])
cls_runtime.append(parsing_time)
cls_names.append('Read/Parse\n+Feat.Extr.')
cls_runtime.append(vectorizing_time)
cls_names.append('Hashing\n+Vect.')
rectangles = plt.bar(range(len(cls_names)), cls_runtime, width=0.5,
color=bar_colors)
ax.set_xticks(np.linspace(0, len(cls_names) - 1, len(cls_names)))

ax.set_xticklabels(cls_names, fontsize=8)
plt.setp(plt.xticks()[1], rotation=30)
ymax = max(cls_runtime) * 1.2
ax.set_ylabel('runtime (s)')
ax.set_title('Prediction Times (%d instances)' % n_test_documents)
autolabel(rectangles)
plt.tight_layout()
plt.show()



•
5.3 Biclustering
Examples concerning the sklearn.cluster.bicluster module.
5.3.1 A demo of the Spectral Co-Clustering algorithm
This example demonstrates how to generate a dataset and bicluster it using the Spectral Co-Clustering algorithm.
The dataset is generated using the make_biclusters function, which creates a matrix of small values and im-
plants bicluster with large values. The rows and columns are then shuffled and passed to the Spectral Co-Clustering
algorithm. Rearranging the shuffled matrix to make biclusters contiguous shows how accurately the algorithm found
the biclusters.
5.3. Biclustering 739



•
Out:
consensus score: 1.000
print(__doc__)
# Author: Kemal Eren <kemal@kemaleren.com>

import numpy as np
from matplotlib import pyplot as plt
from sklearn.datasets import make_biclusters

from sklearn.datasets import samples_generator as sg
from sklearn.cluster.bicluster import SpectralCoclustering
from sklearn.metrics import consensus_score
data, rows, columns = make_biclusters(

shape=(300, 300), n_clusters=5, noise=5,
shuffle=False, random_state=0)

plt.matshow(data, cmap=plt.cm.Blues)
plt.title("Original dataset")
data, row_idx, col_idx = sg._shuffle(data, random_state=0)

plt.title("Shuffled dataset")
model = SpectralCoclustering(n_clusters=5, random_state=0)

model.fit(data)
score = consensus_score(model.biclusters_,
(rows[:, row_idx], columns[:, col_idx]))
print("consensus score: {:.3f}".format(score))
fit_data = data[np.argsort(model.row_labels_)]
fit_data = fit_data[:, np.argsort(model.column_labels_)]
plt.matshow(fit_data, cmap=plt.cm.Blues)
plt.title("After biclustering; rearranged to show biclusters")
plt.show()
5.3.2 A demo of the Spectral Biclustering algorithm
This example demonstrates how to generate a checkerboard dataset and bicluster it using the Spectral Biclustering
algorithm.
The data is generated with the make_checkerboard function, then shuffled and passed to the Spectral Biclustering
algorithm. The rows and columns of the shuffled matrix are rearranged to show the biclusters found by the algorithm.
The outer product of the row and column label vectors shows a representation of the checkerboard structure.




•
Out:
consensus score: 1.0
print(__doc__)
# Author: Kemal Eren <kemal@kemaleren.com>

import numpy as np
from sklearn.datasets import make_checkerboard

from sklearn.datasets import samples_generator as sg
from sklearn.cluster.bicluster import SpectralBiclustering
from sklearn.metrics import consensus_score
n_clusters = (4, 3)
data, rows, columns = make_checkerboard(
shape=(300, 300), n_clusters=n_clusters, noise=10,
shuffle=False, random_state=0)

plt.title("Original dataset")
data, row_idx, col_idx = sg._shuffle(data, random_state=0)

plt.title("Shuffled dataset")
model = SpectralBiclustering(n_clusters=n_clusters, method='log',

random_state=0)
model.fit(data)
score = consensus_score(model.biclusters_,
(rows[:, row_idx], columns[:, col_idx]))
print("consensus score: {:.1f}".format(score))
fit_data = data[np.argsort(model.row_labels_)]
fit_data = fit_data[:, np.argsort(model.column_labels_)]
plt.matshow(fit_data, cmap=plt.cm.Blues)
plt.title("After biclustering; rearranged to show biclusters")
plt.matshow(np.outer(np.sort(model.row_labels_) + 1,
np.sort(model.column_labels_) + 1),
cmap=plt.cm.Blues)
plt.title("Checkerboard structure of rearranged data")
plt.show()
5.3.3 Biclustering documents with the Spectral Co-clustering algorithm
This example demonstrates the Spectral Co-clustering algorithm on the twenty newsgroups dataset. The ‘comp.os.ms-
windows.misc’ category is excluded because it contains many posts containing nothing but data.
The TF-IDF vectorized posts form a word frequency matrix, which is then biclustered using Dhillon’s Spectral Co-
Clustering algorithm. The resulting document-word biclusters indicate subsets words used more often in those subsets
documents.
For a few of the best biclusters, its most common document categories and its ten most important words get printed.
The best biclusters are determined by their normalized cut. The best words are determined by comparing their sums
inside and outside the bicluster.
For comparison, the documents are also clustered using MiniBatchKMeans. The document clusters derived from the
biclusters achieve a better V-measure than clusters found by MiniBatchKMeans.
Out:
Vectorizing...
Coclustering...
Done in 4.67s. V-measure: 0.4435
MiniBatchKMeans...
Done in 8.57s. V-measure: 0.3344

Best biclusters:
----------------
bicluster 0 : 1957 documents, 4363 words
categories : 23% talk.politics.guns, 18% talk.politics.misc, 17% sci.med
words : gun, guns, geb, banks, gordon, clinton, pitt, cdt, surrender, veal

categories : 27% soc.religion.christian, 25% talk.politics.mideast, 24% alt.atheism
words : god, jesus, christians, sin, objective, kent, belief, christ, faith,
˓→moral

categories : 18% comp.sys.mac.hardware, 17% comp.sys.ibm.pc.hardware, 15% comp.
˓→graphics
words : voltage, board, dsp, stereo, receiver, packages, shipping, circuit,

˓→package, compression

categories : 27% rec.motorcycles, 23% rec.autos, 13% misc.forsale
words : bike, car, dod, engine, motorcycle, ride, honda, bikes, helmet, bmw

categories : 81% talk.politics.mideast, 10% alt.atheism, 8% soc.religion.christian
words : turkish, armenia, armenian, armenians, turks, petch, sera, zuma, argic,
˓→ gvg47

import operator
import numpy as np
from sklearn.cluster.bicluster import SpectralCoclustering

from sklearn.cluster import MiniBatchKMeans
from sklearn.externals.six import iteritems
from sklearn.datasets.twenty_newsgroups import fetch_20newsgroups
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.cluster import v_measure_score
print(__doc__)
def number_normalizer(tokens):
""" Map all numeric tokens to a placeholder.
For many applications, tokens that begin with a number are not directly
useful, but the fact that such a token exists can be relevant. By applying
this form of dimensionality reduction, some methods may perform better.
"""
return ("#NUMBER" if token[0].isdigit() else token for token in tokens)

class NumberNormalizingVectorizer(TfidfVectorizer):
def build_tokenizer(self):
tokenize = super(NumberNormalizingVectorizer, self).build_tokenizer()
return lambda doc: list(number_normalizer(tokenize(doc)))
# exclude 'comp.os.ms-windows.misc'
categories = ['alt.atheism', 'comp.graphics',
'comp.sys.ibm.pc.hardware', 'comp.sys.mac.hardware',
'comp.windows.x', 'misc.forsale', 'rec.autos',
'rec.motorcycles', 'rec.sport.baseball',
'rec.sport.hockey', 'sci.crypt', 'sci.electronics',
'sci.med', 'sci.space', 'soc.religion.christian',
'talk.politics.guns', 'talk.politics.mideast',
'talk.politics.misc', 'talk.religion.misc']
newsgroups = fetch_20newsgroups(categories=categories)
y_true = newsgroups.target
vectorizer = NumberNormalizingVectorizer(stop_words='english', min_df=5)

cocluster = SpectralCoclustering(n_clusters=len(categories),
svd_method='arpack', random_state=0)
kmeans = MiniBatchKMeans(n_clusters=len(categories), batch_size=20000,
random_state=0)
print("Vectorizing...")
X = vectorizer.fit_transform(newsgroups.data)
print("Coclustering...")
start_time = time()
cocluster.fit(X)
y_cocluster = cocluster.row_labels_
print("Done in {:.2f}s. V-measure: {:.4f}".format(
time() - start_time,
v_measure_score(y_cocluster, y_true)))
print("MiniBatchKMeans...")
start_time = time()
y_kmeans = kmeans.fit_predict(X)
print("Done in {:.2f}s. V-measure: {:.4f}".format(
time() - start_time,
v_measure_score(y_kmeans, y_true)))
feature_names = vectorizer.get_feature_names()
document_names = list(newsgroups.target_names[i] for i in newsgroups.target)
def bicluster_ncut(i):
rows, cols = cocluster.get_indices(i)
if not (np.any(rows) and np.any(cols)):
import sys
return sys.float_info.max
row_complement = np.nonzero(np.logical_not(cocluster.rows_[i]))[0]
col_complement = np.nonzero(np.logical_not(cocluster.columns_[i]))[0]
# Note: the following is identical to X[rows[:, np.newaxis],
# cols].sum() but much faster in scipy <= 0.16
weight = X[rows][:, cols].sum()

cut = (X[row_complement][:, cols].sum() +

X[rows][:, col_complement].sum())
return cut / weight
def most_common(d):
"""Items of a defaultdict(int) with the highest values.
Like Counter.most_common in Python >=2.7.

"""
return sorted(iteritems(d), key=operator.itemgetter(1), reverse=True)
bicluster_ncuts = list(bicluster_ncut(i)
for i in range(len(newsgroups.target_names)))
best_idx = np.argsort(bicluster_ncuts)[:5]
print()
print("Best biclusters:")
print("----------------")
for idx, cluster in enumerate(best_idx):
n_rows, n_cols = cocluster.get_shape(cluster)
cluster_docs, cluster_words = cocluster.get_indices(cluster)
if not len(cluster_docs) or not len(cluster_words):
continue
# categories
counter = defaultdict(int)
for i in cluster_docs:
counter[document_names[i]] += 1
cat_string = ", ".join("{:.0f}% {}".format(float(c) / n_rows * 100, name)
for name, c in most_common(counter)[:3])
# words
out_of_cluster_docs = cocluster.row_labels_ != cluster
out_of_cluster_docs = np.where(out_of_cluster_docs)[0]
word_col = X[:, cluster_words]
word_scores = np.array(word_col[cluster_docs, :].sum(axis=0) -
word_col[out_of_cluster_docs, :].sum(axis=0))
word_scores = word_scores.ravel()
important_words = list(feature_names[cluster_words[i]]
for i in word_scores.argsort()[:-11:-1])
print("bicluster {} : {} documents, {} words".format(

idx, n_rows, n_cols))
print("categories : {}".format(cat_string))
print("words : {}\n".format(', '.join(important_words)))
5.4 Calibration
Examples illustrating the calibration of predicted probabilities of classifiers.
5.4. Calibration 751

5.4.1 Comparison of Calibration of Classifiers
Well calibrated classifiers are probabilistic classifiers for which the output of the predict_proba method can be directly
interpreted as a confidence level. For instance a well calibrated (binary) classifier should classify the samples such that
among the samples to which it gave a predict_proba value close to 0.8, approx. 80% actually belong to the positive
class.
LogisticRegression returns well calibrated predictions as it directly optimizes log-loss. In contrast, the other methods
return biased probabilities, with different biases per method:
• GaussianNaiveBayes tends to push probabilities to 0 or 1 (note the counts in the histograms). This is mainly
because it makes the assumption that features are conditionally independent given the class, which is not the
case in this dataset which contains 2 redundant features.
• RandomForestClassifier shows the opposite behavior: the histograms show peaks at approx. 0.2 and 0.9 proba-
bility, while probabilities close to 0 or 1 are very rare. An explanation for this is given by Niculescu-Mizil and
Caruana1 : “Methods such as bagging and random forests that average predictions from a base set of models can
have difficulty making predictions near 0 and 1 because variance in the underlying base models will bias predic-
tions that should be near zero or one away from these values. Because predictions are restricted to the interval
[0,1], errors caused by variance tend to be one- sided near zero and one. For example, if a model should predict
p = 0 for a case, the only way bagging can achieve this is if all bagged trees predict zero. If we add noise to the
trees that bagging is averaging over, this noise will cause some trees to predict values larger than 0 for this case,
thus moving the average prediction of the bagged ensemble away from 0. We observe this effect most strongly
with random forests because the base-level trees trained with random forests have relatively high variance due
to feature subsetting.” As a result, the calibration curve shows a characteristic sigmoid shape, indicating that the
classifier could trust its “intuition” more and return probabilities closer to 0 or 1 typically.
• Support Vector Classification (SVC) shows an even more sigmoid curve as the RandomForestClassifier, which
is typical for maximum-margin methods (compare Niculescu-Mizil and Caruana1 ), which focus on hard samples
that are close to the decision boundary (the support vectors).
References:
1 Predicting Good Probabilities with Supervised Learning, A. Niculescu-Mizil & R. Caruana, ICML 2005

print(__doc__)
# Author: Jan Hendrik Metzen <jhm@informatik.uni-bremen.de>

# License: BSD Style.
import numpy as np
np.random.seed(0)

from sklearn.naive_bayes import GaussianNB
from sklearn.linear_model import LogisticRegression
from sklearn.ensemble import RandomForestClassifier
from sklearn.svm import LinearSVC

from sklearn.calibration import calibration_curve
X, y = datasets.make_classification(n_samples=100000, n_features=20,
n_informative=2, n_redundant=2)
train_samples = 100 # Samples used for training the models
X_train = X[:train_samples]
X_test = X[train_samples:]
y_train = y[:train_samples]
y_test = y[train_samples:]
# Create classifiers
lr = LogisticRegression(solver='lbfgs')
gnb = GaussianNB()
svc = LinearSVC(C=1.0)
rfc = RandomForestClassifier(n_estimators=100)
# #############################################################################
# Plot calibration plots
ax1 = plt.subplot2grid((3, 1), (0, 0), rowspan=2)
ax2 = plt.subplot2grid((3, 1), (2, 0))
ax1.plot([0, 1], [0, 1], "k:", label="Perfectly calibrated")

for clf, name in [(lr, 'Logistic'),
(gnb, 'Naive Bayes'),
(svc, 'Support Vector Classification'),
(rfc, 'Random Forest')]:
clf.fit(X_train, y_train)
if hasattr(clf, "predict_proba"):
prob_pos = clf.predict_proba(X_test)[:, 1]
else: # use decision function
prob_pos = clf.decision_function(X_test)
prob_pos = \
(prob_pos - prob_pos.min()) / (prob_pos.max() - prob_pos.min())
fraction_of_positives, mean_predicted_value = \
calibration_curve(y_test, prob_pos, n_bins=10)
ax1.plot(mean_predicted_value, fraction_of_positives, "s-",

label="%s" % (name, ))
ax2.hist(prob_pos, range=(0, 1), bins=10, label=name,

histtype="step", lw=2)
ax1.set_ylabel("Fraction of positives")
ax1.set_ylim([-0.05, 1.05])
ax1.legend(loc="lower right")
ax1.set_title('Calibration plots (reliability curve)')
ax2.set_xlabel("Mean predicted value")

ax2.set_ylabel("Count")
ax2.legend(loc="upper center", ncol=2)
plt.tight_layout()
plt.show()

5.4.2 Probability Calibration curves
When performing classification one often wants to predict not only the class label, but also the associated probability.
This probability gives some kind of confidence on the prediction. This example demonstrates how to display how well
calibrated the predicted probabilities are and how to calibrate an uncalibrated classifier.
The experiment is performed on an artificial dataset for binary classification with 100,000 samples (1,000 of them are
used for model fitting) with 20 features. Of the 20 features, only 2 are informative and 10 are redundant. The first
figure shows the estimated probabilities obtained with logistic regression, Gaussian naive Bayes, and Gaussian naive
Bayes with both isotonic calibration and sigmoid calibration. The calibration performance is evaluated with Brier
score, reported in the legend (the smaller the better). One can observe here that logistic regression is well calibrated
while raw Gaussian naive Bayes performs very badly. This is because of the redundant features which violate the
assumption of feature-independence and result in an overly confident classifier, which is indicated by the typical
transposed-sigmoid curve.
Calibration of the probabilities of Gaussian naive Bayes with isotonic regression can fix this issue as can be seen from
the nearly diagonal calibration curve. Sigmoid calibration also improves the brier score slightly, albeit not as strongly
as the non-parametric isotonic regression. This can be attributed to the fact that we have plenty of calibration data such
that the greater flexibility of the non-parametric model can be exploited.
The second figure shows the calibration curve of a linear support-vector classifier (LinearSVC). LinearSVC shows
the opposite behavior as Gaussian naive Bayes: the calibration curve has a sigmoid curve, which is typical for an
under-confident classifier. In the case of LinearSVC, this is caused by the margin property of the hinge loss, which
lets the model focus on hard samples that are close to the decision boundary (the support vectors).
Both kinds of calibration can fix this issue and yield nearly identical results. This shows that sigmoid calibration can
deal with situations where the calibration curve of the base classifier is sigmoid (e.g., for LinearSVC) but not where it
is transposed-sigmoid (e.g., Gaussian naive Bayes).


•
Out:
Logistic:
Brier: 0.099
Precision: 0.872
Recall: 0.851
F1: 0.862
Naive Bayes:
Brier: 0.118
Precision: 0.857
Recall: 0.876
F1: 0.867
Naive Bayes + Isotonic:

Brier: 0.098
Precision: 0.883

Recall: 0.836
F1: 0.859
Naive Bayes + Sigmoid:

Brier: 0.109
Precision: 0.861
Recall: 0.871
F1: 0.866
Logistic:
Brier: 0.099
Precision: 0.872
Recall: 0.851
F1: 0.862
SVC:
Brier: 0.163
Precision: 0.872
Recall: 0.852
F1: 0.862
SVC + Isotonic:
Brier: 0.100
Precision: 0.853
Recall: 0.878
F1: 0.865
SVC + Sigmoid:
Brier: 0.099
Precision: 0.874
Recall: 0.849
F1: 0.861
print(__doc__)
# Author: Alexandre Gramfort <alexandre.gramfort@telecom-paristech.fr>

# Jan Hendrik Metzen <jhm@informatik.uni-bremen.de>

from sklearn.metrics import (brier_score_loss, precision_score, recall_score,
f1_score)
from sklearn.calibration import CalibratedClassifierCV, calibration_curve
# Create dataset of classification task with many redundant and few

# informative features

X, y = datasets.make_classification(n_samples=100000, n_features=20,
n_informative=2, n_redundant=10,
random_state=42)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.99,

random_state=42)
def plot_calibration_curve(est, name, fig_index):

"""Plot calibration curve for est w/o and with calibration. """
# Calibrated with isotonic calibration
isotonic = CalibratedClassifierCV(est, cv=2, method='isotonic')
# Calibrated with sigmoid calibration

sigmoid = CalibratedClassifierCV(est, cv=2, method='sigmoid')
# Logistic regression with no calibration as baseline

lr = LogisticRegression(C=1., solver='lbfgs')
fig = plt.figure(fig_index, figsize=(10, 10))

ax1 = plt.subplot2grid((3, 1), (0, 0), rowspan=2)
ax2 = plt.subplot2grid((3, 1), (2, 0))
ax1.plot([0, 1], [0, 1], "k:", label="Perfectly calibrated")

for clf, name in [(lr, 'Logistic'),
(est, name),
(isotonic, name + ' + Isotonic'),
(sigmoid, name + ' + Sigmoid')]:
y_pred = clf.predict(X_test)
if hasattr(clf, "predict_proba"):
prob_pos = clf.predict_proba(X_test)[:, 1]
else: # use decision function
prob_pos = clf.decision_function(X_test)
prob_pos = \
(prob_pos - prob_pos.min()) / (prob_pos.max() - prob_pos.min())
clf_score = brier_score_loss(y_test, prob_pos, pos_label=y.max())

print("%s:" % name)
print("\tBrier: %1.3f" % (clf_score))
print("\tPrecision: %1.3f" % precision_score(y_test, y_pred))
print("\tRecall: %1.3f" % recall_score(y_test, y_pred))
print("\tF1: %1.3f\n" % f1_score(y_test, y_pred))
fraction_of_positives, mean_predicted_value = \
calibration_curve(y_test, prob_pos, n_bins=10)
ax1.plot(mean_predicted_value, fraction_of_positives, "s-",

label="%s (%1.3f)" % (name, clf_score))
ax2.hist(prob_pos, range=(0, 1), bins=10, label=name,

histtype="step", lw=2)
ax1.set_ylabel("Fraction of positives")
ax1.set_ylim([-0.05, 1.05])
ax1.legend(loc="lower right")
ax1.set_title('Calibration plots (reliability curve)')

ax2.set_xlabel("Mean predicted value")

ax2.set_ylabel("Count")
ax2.legend(loc="upper center", ncol=2)
plt.tight_layout()
# Plot calibration curve for Gaussian Naive Bayes

plot_calibration_curve(GaussianNB(), "Naive Bayes", 1)
# Plot calibration curve for Linear SVC

plot_calibration_curve(LinearSVC(), "SVC", 2)
plt.show()
5.4.3 Probability calibration of classifiers
When performing classification you often want to predict not only the class label, but also the associated probability.
This probability gives you some kind of confidence on the prediction. However, not all classifiers provide well-
calibrated probabilities, some being over-confident while others being under-confident. Thus, a separate calibration
of predicted probabilities is often desirable as a postprocessing. This example illustrates two different methods for
this calibration and evaluates the quality of the returned probabilities using Brier’s score (see https://en.wikipedia.org/
wiki/Brier_score).
Compared are the estimated probability using a Gaussian naive Bayes classifier without calibration, with a sigmoid
calibration, and with a non-parametric isotonic calibration. One can observe that only the non-parametric model is
able to provide a probability calibration that returns probabilities close to the expected 0.5 for most of the samples
belonging to the middle cluster with heterogeneous labels. This results in a significantly improved Brier score.


•
Out:
Brier scores: (the smaller the better)

No calibration: 0.104
With isotonic calibration: 0.084
With sigmoid calibration: 0.109
print(__doc__)
# Author: Mathieu Blondel <mathieu@mblondel.org>

# Alexandre Gramfort <alexandre.gramfort@telecom-paristech.fr>
# Balazs Kegl <balazs.kegl@gmail.com>
import numpy as np
from matplotlib import cm
from sklearn.datasets import make_blobs

from sklearn.metrics import brier_score_loss

from sklearn.calibration import CalibratedClassifierCV

n_samples = 50000
n_bins = 3 # use 3 bins for calibration_curve as we have 3 clusters here
# Generate 3 blobs with 2 classes where the second blob contains

# half positive samples and half negative samples. Probability in this
# blob is therefore 0.5.
centers = [(-5, -5), (0, 0), (5, 5)]
X, y = make_blobs(n_samples=n_samples, n_features=2, cluster_std=1.0,
centers=centers, shuffle=False, random_state=42)
y[:n_samples // 2] = 0
y[n_samples // 2:] = 1
sample_weight = np.random.RandomState(42).rand(y.shape[0])
# split train, test for calibration

X_train, X_test, y_train, y_test, sw_train, sw_test = \
train_test_split(X, y, sample_weight, test_size=0.9, random_state=42)
# Gaussian Naive-Bayes with no calibration

clf = GaussianNB()
clf.fit(X_train, y_train) # GaussianNB itself does not support sample-weights
prob_pos_clf = clf.predict_proba(X_test)[:, 1]
# Gaussian Naive-Bayes with isotonic calibration

clf_isotonic = CalibratedClassifierCV(clf, cv=2, method='isotonic')
clf_isotonic.fit(X_train, y_train, sw_train)
prob_pos_isotonic = clf_isotonic.predict_proba(X_test)[:, 1]
# Gaussian Naive-Bayes with sigmoid calibration

clf_sigmoid = CalibratedClassifierCV(clf, cv=2, method='sigmoid')
clf_sigmoid.fit(X_train, y_train, sw_train)
prob_pos_sigmoid = clf_sigmoid.predict_proba(X_test)[:, 1]
print("Brier scores: (the smaller the better)")
clf_score = brier_score_loss(y_test, prob_pos_clf, sw_test)

print("No calibration: %1.3f" % clf_score)
clf_isotonic_score = brier_score_loss(y_test, prob_pos_isotonic, sw_test)

print("With isotonic calibration: %1.3f" % clf_isotonic_score)
clf_sigmoid_score = brier_score_loss(y_test, prob_pos_sigmoid, sw_test)

print("With sigmoid calibration: %1.3f" % clf_sigmoid_score)
# #############################################################################
# Plot the data and the predicted probabilities
plt.figure()
y_unique = np.unique(y)
colors = cm.rainbow(np.linspace(0.0, 1.0, y_unique.size))
for this_y, color in zip(y_unique, colors):
this_X = X_train[y_train == this_y]
this_sw = sw_train[y_train == this_y]
plt.scatter(this_X[:, 0], this_X[:, 1], s=this_sw * 50, c=color,
alpha=0.5, edgecolor='k',

label="Class %s" % this_y)

plt.title("Data")
plt.figure()
order = np.lexsort((prob_pos_clf, ))
plt.plot(prob_pos_clf[order], 'r', label='No calibration (%1.3f)' % clf_score)
plt.plot(prob_pos_isotonic[order], 'g', linewidth=3,
label='Isotonic calibration (%1.3f)' % clf_isotonic_score)
plt.plot(prob_pos_sigmoid[order], 'b', linewidth=3,
label='Sigmoid calibration (%1.3f)' % clf_sigmoid_score)
plt.plot(np.linspace(0, y_test.size, 51)[1::2],
y_test[order].reshape(25, -1).mean(1),
'k', linewidth=3, label=r'Empirical')
plt.ylim([-0.05, 1.05])
plt.xlabel("Instances sorted according to predicted probability "
"(uncalibrated GNB)")
plt.ylabel("P(y=1)")
plt.title("Gaussian naive Bayes probabilities")
plt.show()
5.4.4 Probability Calibration for 3-class classification
This example illustrates how sigmoid calibration changes predicted probabilities for a 3-class classification problem.
Illustrated is the standard 2-simplex, where the three corners correspond to the three classes. Arrows point from the
probability vectors predicted by an uncalibrated classifier to the probability vectors predicted by the same classifier
after sigmoid calibration on a hold-out validation set. Colors indicate the true class of an instance (red: class 1, green:
class 2, blue: class 3).
The base classifier is a random forest classifier with 25 base estimators (trees). If this classifier is trained on all 800
training datapoints, it is overly confident in its predictions and thus incurs a large log-loss. Calibrating an identical
classifier, which was trained on 600 datapoints, with method=’sigmoid’ on the remaining 200 datapoints reduces the
confidence of the predictions, i.e., moves the probability vectors from the edges of the simplex towards the center.
This calibration results in a lower log-loss. Note that an alternative would have been to increase the number of base
estimators which would have resulted in a similar decrease in log-loss.


•
Out:
Log-loss of
* uncalibrated classifier trained on 800 datapoints: 1.280
* classifier trained on 600 datapoints and calibrated on 200 datapoint: 0.534
print(__doc__)
# Author: Jan Hendrik Metzen <jhm@informatik.uni-bremen.de>

import numpy as np

from sklearn.calibration import CalibratedClassifierCV
from sklearn.metrics import log_loss
np.random.seed(0)

# Generate data
X, y = make_blobs(n_samples=1000, n_features=2, random_state=42,
cluster_std=5.0)
X_train, y_train = X[:600], y[:600]
X_valid, y_valid = X[600:800], y[600:800]
X_train_valid, y_train_valid = X[:800], y[:800]
X_test, y_test = X[800:], y[800:]
# Train uncalibrated random forest classifier on whole train and validation

# data and evaluate on test data
clf = RandomForestClassifier(n_estimators=25)
clf.fit(X_train_valid, y_train_valid)
clf_probs = clf.predict_proba(X_test)
score = log_loss(y_test, clf_probs)
# Train random forest classifier, calibrate on validation data and evaluate

# on test data
clf_probs = clf.predict_proba(X_test)
sig_clf = CalibratedClassifierCV(clf, method="sigmoid", cv="prefit")
sig_clf.fit(X_valid, y_valid)
sig_clf_probs = sig_clf.predict_proba(X_test)
sig_score = log_loss(y_test, sig_clf_probs)
# Plot changes in predicted probabilities via arrows

plt.figure(0)
colors = ["r", "g", "b"]
for i in range(clf_probs.shape[0]):
plt.arrow(clf_probs[i, 0], clf_probs[i, 1],
sig_clf_probs[i, 0] - clf_probs[i, 0],
sig_clf_probs[i, 1] - clf_probs[i, 1],
color=colors[y_test[i]], head_width=1e-2)
# Plot perfect predictions

plt.plot([1.0], [0.0], 'ro', ms=20, label="Class 1")
plt.plot([0.0], [1.0], 'go', ms=20, label="Class 2")
plt.plot([0.0], [0.0], 'bo', ms=20, label="Class 3")
# Plot boundaries of unit simplex

plt.plot([0.0, 1.0, 0.0, 0.0], [0.0, 0.0, 1.0, 0.0], 'k', label="Simplex")
# Annotate points on the simplex

plt.annotate(r'($\frac{1}{3}$, $\frac{1}{3}$, $\frac{1}{3}$)',
xy=(1.0/3, 1.0/3), xytext=(1.0/3, .23), xycoords='data',
arrowprops=dict(facecolor='black', shrink=0.05),
horizontalalignment='center', verticalalignment='center')
plt.plot([1.0/3], [1.0/3], 'ko', ms=5)
plt.annotate(r'($\frac{1}{2}$, $0$, $\frac{1}{2}$)',
xy=(.5, .0), xytext=(.5, .1), xycoords='data',
plt.annotate(r'($0$, $\frac{1}{2}$, $\frac{1}{2}$)',
plt.annotate(r'($\frac{1}{2}$, $\frac{1}{2}$, $0$)',


plt.annotate(r'($0$, $0$, $1$)',
xy=(0, 0), xytext=(.1, .1), xycoords='data',
plt.annotate(r'($1$, $0$, $0$)',
xy=(1, 0), xytext=(1, .1), xycoords='data',
plt.annotate(r'($0$, $1$, $0$)',
xy=(0, 1), xytext=(.1, 1), xycoords='data',
# Add grid
plt.grid("off")
for x in [0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0]:
plt.plot([0, x], [x, 0], 'k', alpha=0.2)
plt.plot([0, 0 + (1-x)/2], [x, x + (1-x)/2], 'k', alpha=0.2)
plt.plot([x, x + (1-x)/2], [0, 0 + (1-x)/2], 'k', alpha=0.2)
plt.title("Change of predicted probabilities after sigmoid calibration")

plt.xlabel("Probability class 1")
plt.ylabel("Probability class 2")
plt.xlim(-0.05, 1.05)
plt.ylim(-0.05, 1.05)
print("Log-loss of")
print(" * uncalibrated classifier trained on 800 datapoints: %.3f "
% score)
print(" * classifier trained on 600 datapoints and calibrated on "
"200 datapoint: %.3f" % sig_score)
# Illustrate calibrator
plt.figure(1)
# generate grid over 2-simplex
p1d = np.linspace(0, 1, 20)
p0, p1 = np.meshgrid(p1d, p1d)
p2 = 1 - p0 - p1
p = np.c_[p0.ravel(), p1.ravel(), p2.ravel()]
p = p[p[:, 2] >= 0]
calibrated_classifier = sig_clf.calibrated_classifiers_[0]
prediction = np.vstack([calibrator.predict(this_p)
for calibrator, this_p in
zip(calibrated_classifier.calibrators_, p.T)]).T
prediction /= prediction.sum(axis=1)[:, None]
# Plot modifications of calibrator

for i in range(prediction.shape[0]):
plt.arrow(p[i, 0], p[i, 1],
prediction[i, 0] - p[i, 0], prediction[i, 1] - p[i, 1],
head_width=1e-2, color=colors[np.argmax(p[i])])
# Plot boundaries of unit simplex
plt.plot([0.0, 1.0, 0.0, 0.0], [0.0, 0.0, 1.0, 0.0], 'k', label="Simplex")

plt.grid("off")
for x in [0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0]:
plt.plot([0, x], [x, 0], 'k', alpha=0.2)
plt.plot([0, 0 + (1-x)/2], [x, x + (1-x)/2], 'k', alpha=0.2)
plt.plot([x, x + (1-x)/2], [0, 0 + (1-x)/2], 'k', alpha=0.2)
plt.title("Illustration of sigmoid calibrator")

plt.xlabel("Probability class 1")
plt.ylabel("Probability class 2")
plt.xlim(-0.05, 1.05)
plt.ylim(-0.05, 1.05)
plt.show()
5.5 Classification
General examples about classification algorithms.
5.5.1 Recognizing hand-written digits
An example showing how the scikit-learn can be used to recognize images of hand-written digits.
This example is commented in the tutorial section of the user manual.
5.5. Classification 769

Out:
Classification report for classifier SVC(C=1.0, cache_size=200, class_weight=None,

˓→coef0=0.0,

tol=0.001, verbose=False):
0 1.00 0.99 0.99 88

1 0.99 0.97 0.98 91
2 0.99 0.99 0.99 86
3 0.98 0.87 0.92 91
4 0.99 0.96 0.97 92
5 0.95 0.97 0.96 91
6 0.99 0.99 0.99 91
7 0.96 0.99 0.97 89
8 0.94 1.00 0.97 88
9 0.93 0.98 0.95 92
micro avg 0.97 0.97 0.97 899

macro avg 0.97 0.97 0.97 899
weighted avg 0.97 0.97 0.97 899
Confusion matrix:

[[87 0 0 0 1 0 0 0 0 0]
[ 0 88 1 0 0 0 0 0 1 1]
[ 0 0 85 1 0 0 0 0 0 0]
[ 0 0 0 79 0 3 0 4 5 0]
[ 0 0 0 0 88 0 0 0 0 4]
[ 0 0 0 0 0 88 1 0 0 2]
[ 0 1 0 0 0 0 90 0 0 0]
[ 0 0 0 0 0 1 0 88 0 0]
[ 0 0 0 0 0 0 0 0 88 0]
[ 0 0 0 1 0 1 0 0 0 90]]
print(__doc__)
# Author: Gael Varoquaux <gael dot varoquaux at normalesup dot org>

# Standard scientific Python imports

# Import datasets, classifiers and performance metrics

from sklearn import datasets, svm, metrics
# The digits dataset

# The data that we are interested in is made of 8x8 images of digits, let's
# have a look at the first 4 images, stored in the ìmages` attribute of the
# dataset. If we were working from image files, we could load them using
# matplotlib.pyplot.imread. Note that each image must have the same size. For these
# images, we know which digit they represent: it is given in the 'target' of
# the dataset.
images_and_labels = list(zip(digits.images, digits.target))
for index, (image, label) in enumerate(images_and_labels[:4]):
plt.subplot(2, 4, index + 1)
plt.axis('off')
plt.imshow(image, cmap=plt.cm.gray_r, interpolation='nearest')
plt.title('Training: %i' % label)
# To apply a classifier on this data, we need to flatten the image, to

n_samples = len(digits.images)
data = digits.images.reshape((n_samples, -1))
# Create a classifier: a support vector classifier

classifier = svm.SVC(gamma=0.001)
# We learn the digits on the first half of the digits

classifier.fit(data[:n_samples // 2], digits.target[:n_samples // 2])
# Now predict the value of the digit on the second half:

expected = digits.target[n_samples // 2:]
predicted = classifier.predict(data[n_samples // 2:])

print("Classification report for classifier %s:\n%s\n"

% (classifier, metrics.classification_report(expected, predicted)))
print("Confusion matrix:\n%s" % metrics.confusion_matrix(expected, predicted))
images_and_predictions = list(zip(digits.images[n_samples // 2:], predicted))

for index, (image, prediction) in enumerate(images_and_predictions[:4]):
plt.axis('off')
plt.imshow(image, cmap=plt.cm.gray_r, interpolation='nearest')
plt.title('Prediction: %i' % prediction)
plt.show()
5.5.2 Normal and Shrinkage Linear Discriminant Analysis for classification
Shows how shrinkage improves classification.

import numpy as np

from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
n_train = 20 # samples for training

n_test = 200 # samples for testing
n_averages = 50 # how often to repeat classification
n_features_max = 75 # maximum number of features
step = 4 # step size for the calculation
def generate_data(n_samples, n_features):

"""Generate random blob-ish data with noisy features.
This returns an array of input data with shape `(n_samples, n_features)`

and an array of `n_samples` target labels.
Only one feature contains discriminative information, the other features

contain only noise.
"""
X, y = make_blobs(n_samples=n_samples, n_features=1, centers=[[-2], [2]])
# add non-discriminative features

if n_features > 1:
X = np.hstack([X, np.random.randn(n_samples, n_features - 1)])
return X, y
acc_clf1, acc_clf2 = [], []

n_features_range = range(1, n_features_max + 1, step)
for n_features in n_features_range:
score_clf1, score_clf2 = 0, 0
for _ in range(n_averages):
X, y = generate_data(n_train, n_features)
clf1 = LinearDiscriminantAnalysis(solver='lsqr', shrinkage='auto').fit(X, y)

clf2 = LinearDiscriminantAnalysis(solver='lsqr', shrinkage=None).fit(X, y)
X, y = generate_data(n_test, n_features)
score_clf1 += clf1.score(X, y)
score_clf2 += clf2.score(X, y)
acc_clf1.append(score_clf1 / n_averages)
acc_clf2.append(score_clf2 / n_averages)
features_samples_ratio = np.array(n_features_range) / n_train
plt.plot(features_samples_ratio, acc_clf1, linewidth=2,

label="Linear Discriminant Analysis with shrinkage", color='navy')
plt.plot(features_samples_ratio, acc_clf2, linewidth=2,
label="Linear Discriminant Analysis", color='gold')
plt.xlabel('n_features / n_samples')
plt.ylabel('Classification accuracy')
plt.legend(loc=1, prop={'size': 12})

plt.suptitle('Linear Discriminant Analysis vs. \

shrinkage Linear Discriminant Analysis (1 discriminative feature)')
plt.show()
5.5.3 Plot classification probability
Plot the classification probability for different classifiers. We use a 3 class dataset, and we classify it with a Sup-
port Vector classifier, L1 and L2 penalized logistic regression with either a One-Vs-Rest or multinomial setting, and
Gaussian process classification.
Linear SVC is not a probabilistic classifier by default but it has a built-in calibration option enabled in this example
(probability=True).
The logistic regression with One-Vs-Rest is not a multiclass classifier out of the box. As a result it has more trouble
in separating class 2 and 3 than the other estimators.


Out:
Accuracy (train) for L1 logistic: 82.7%

Accuracy (train) for L2 logistic (Multinomial): 82.7%
Accuracy (train) for L2 logistic (OvR): 79.3%
Accuracy (train) for Linear SVC: 82.0%
Accuracy (train) for GPC: 82.7%
print(__doc__)


import numpy as np
from sklearn.metrics import accuracy_score

from sklearn.gaussian_process import GaussianProcessClassifier
from sklearn.gaussian_process.kernels import RBF
X = iris.data[:, 0:2] # we only take the first two features for visualization
y = iris.target
C = 10
kernel = 1.0 * RBF([1.0, 1.0]) # for GPC
# Create different classifiers.

classifiers = {
'L1 logistic': LogisticRegression(C=C, penalty='l1',
solver='saga',
multi_class='multinomial',
max_iter=10000),
'L2 logistic (Multinomial)': LogisticRegression(C=C, penalty='l2',
solver='saga',
max_iter=10000),
'L2 logistic (OvR)': LogisticRegression(C=C, penalty='l2',
solver='saga',
multi_class='ovr',
max_iter=10000),
'Linear SVC': SVC(kernel='linear', C=C, probability=True,
random_state=0),
'GPC': GaussianProcessClassifier(kernel)
}
n_classifiers = len(classifiers)

plt.figure(figsize=(3 * 2, n_classifiers * 2))

plt.subplots_adjust(bottom=.2, top=.95)
xx = np.linspace(3, 9, 100)
yy = np.linspace(1, 5, 100).T
xx, yy = np.meshgrid(xx, yy)
Xfull = np.c_[xx.ravel(), yy.ravel()]
for index, (name, classifier) in enumerate(classifiers.items()):

classifier.fit(X, y)
y_pred = classifier.predict(X)
accuracy = accuracy_score(y, y_pred)
print("Accuracy (train) for %s: %0.1f%% " % (name, accuracy * 100))
# View probabilities:
probas = classifier.predict_proba(Xfull)
n_classes = np.unique(y_pred).size
for k in range(n_classes):
plt.subplot(n_classifiers, n_classes, index * n_classes + k + 1)
plt.title("Class %d" % k)
if k == 0:
plt.ylabel(name)
imshow_handle = plt.imshow(probas[:, k].reshape((100, 100)),
extent=(3, 9, 1, 5), origin='lower')
plt.xticks(())
plt.yticks(())
idx = (y_pred == k)
if idx.any():
plt.scatter(X[idx, 0], X[idx, 1], marker='o', c='w', edgecolor='k')
ax = plt.axes([0.15, 0.04, 0.7, 0.05])

plt.title("Probability")
plt.colorbar(imshow_handle, cax=ax, orientation='horizontal')
plt.show()
5.5.4 Classifier comparison
A comparison of a several classifiers in scikit-learn on synthetic datasets. The point of this example is to illustrate
the nature of decision boundaries of different classifiers. This should be taken with a grain of salt, as the intuition
conveyed by these examples does not necessarily carry over to real datasets.
Particularly in high-dimensional spaces, data can more easily be separated linearly and the simplicity of classifiers
such as naive Bayes and linear SVMs might lead to better generalization than is achieved by other classifiers.
The plots show training points in solid colors and testing points semi-transparent. The lower right shows the classifi-
cation accuracy on the test set.

print(__doc__)
# Code source: Gaël Varoquaux

# Andreas Müller
# Modified for documentation by Jaques Grobler
import numpy as np
from matplotlib.colors import ListedColormap
from sklearn.datasets import make_moons, make_circles, make_classification
from sklearn.neural_network import MLPClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier
from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis
h = .02 # step size in the mesh
names = ["Nearest Neighbors", "Linear SVM", "RBF SVM", "Gaussian Process",

"Decision Tree", "Random Forest", "Neural Net", "AdaBoost",
"Naive Bayes", "QDA"]
classifiers = [
KNeighborsClassifier(3),
SVC(kernel="linear", C=0.025),
SVC(gamma=2, C=1),
GaussianProcessClassifier(1.0 * RBF(1.0)),
DecisionTreeClassifier(max_depth=5),
RandomForestClassifier(max_depth=5, n_estimators=10, max_features=1),
MLPClassifier(alpha=1),
AdaBoostClassifier(),
GaussianNB(),
QuadraticDiscriminantAnalysis()]
X, y = make_classification(n_features=2, n_redundant=0, n_informative=2,

random_state=1, n_clusters_per_class=1)
X += 2 * rng.uniform(size=X.shape)
linearly_separable = (X, y)
datasets = [make_moons(noise=0.3, random_state=0),

make_circles(noise=0.2, factor=0.5, random_state=1),
linearly_separable
]
figure = plt.figure(figsize=(27, 9))

i = 1
# iterate over datasets
for ds_cnt, ds in enumerate(datasets):
# preprocess dataset, split into training and test part
X, y = ds
X = StandardScaler().fit_transform(X)
X_train, X_test, y_train, y_test = \
train_test_split(X, y, test_size=.4, random_state=42)
x_min, x_max = X[:, 0].min() - .5, X[:, 0].max() + .5

y_min, y_max = X[:, 1].min() - .5, X[:, 1].max() + .5
xx, yy = np.meshgrid(np.arange(x_min, x_max, h),
np.arange(y_min, y_max, h))
# just plot the dataset first

cm = plt.cm.RdBu
cm_bright = ListedColormap(['#FF0000', '#0000FF'])
ax = plt.subplot(len(datasets), len(classifiers) + 1, i)
if ds_cnt == 0:
ax.set_title("Input data")
# Plot the training points
ax.scatter(X_train[:, 0], X_train[:, 1], c=y_train, cmap=cm_bright,
edgecolors='k')
# Plot the testing points
ax.scatter(X_test[:, 0], X_test[:, 1], c=y_test, cmap=cm_bright, alpha=0.6,
edgecolors='k')
ax.set_xlim(xx.min(), xx.max())
ax.set_ylim(yy.min(), yy.max())
ax.set_xticks(())
ax.set_yticks(())
i += 1
# iterate over classifiers

for name, clf in zip(names, classifiers):
score = clf.score(X_test, y_test)
if hasattr(clf, "decision_function"):
Z = clf.decision_function(np.c_[xx.ravel(), yy.ravel()])
else:
Z = clf.predict_proba(np.c_[xx.ravel(), yy.ravel()])[:, 1]


ax.contourf(xx, yy, Z, cmap=cm, alpha=.8)

edgecolors='k')
# Plot the testing points
ax.scatter(X_test[:, 0], X_test[:, 1], c=y_test, cmap=cm_bright,
edgecolors='k', alpha=0.6)
ax.set_xticks(())
ax.set_yticks(())
if ds_cnt == 0:
ax.set_title(name)
ax.text(xx.max() - .3, yy.min() + .3, ('%.2f' % score).lstrip('0'),
size=15, horizontalalignment='right')
i += 1
plt.tight_layout()
plt.show()
5.5.5 Linear and Quadratic Discriminant Analysis with covariance ellipsoid
This example plots the covariance ellipsoids of each class and decision boundary learned by LDA and QDA. The
ellipsoids display the double standard deviation for each class. With LDA, the standard deviation is the same for all
the classes, while each class has its own standard deviation with QDA.

print(__doc__)
from scipy import linalg

import numpy as np
import matplotlib as mpl
from matplotlib import colors

from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis
# #############################################################################
# Colormap
cmap = colors.LinearSegmentedColormap(
'red_blue_classes',
{'red': [(0, 1, 1), (1, 0.7, 0.7)],
'green': [(0, 0.7, 0.7), (1, 0.7, 0.7)],
'blue': [(0, 0.7, 0.7), (1, 1, 1)]})
plt.cm.register_cmap(cmap=cmap)
# #############################################################################
# Generate datasets

def dataset_fixed_cov():
'''Generate 2 Gaussians samples with the same covariance matrix'''
n, dim = 300, 2
np.random.seed(0)
C = np.array([[0., -0.23], [0.83, .23]])
X = np.r_[np.dot(np.random.randn(n, dim), C),
np.dot(np.random.randn(n, dim), C) + np.array([1, 1])]
y = np.hstack((np.zeros(n), np.ones(n)))
return X, y
def dataset_cov():
'''Generate 2 Gaussians samples with different covariance matrices'''
n, dim = 300, 2
np.random.seed(0)
C = np.array([[0., -1.], [2.5, .7]]) * 2.
X = np.r_[np.dot(np.random.randn(n, dim), C),
np.dot(np.random.randn(n, dim), C.T) + np.array([1, 4])]
y = np.hstack((np.zeros(n), np.ones(n)))
return X, y
# #############################################################################
# Plot functions
def plot_data(lda, X, y, y_pred, fig_index):
splot = plt.subplot(2, 2, fig_index)
if fig_index == 1:
plt.title('Linear Discriminant Analysis')
plt.ylabel('Data with\n fixed covariance')
elif fig_index == 2:
plt.title('Quadratic Discriminant Analysis')
elif fig_index == 3:
plt.ylabel('Data with\n varying covariances')
tp = (y == y_pred) # True Positive

tp0, tp1 = tp[y == 0], tp[y == 1]
X0, X1 = X[y == 0], X[y == 1]
X0_tp, X0_fp = X0[tp0], X0[~tp0]
X1_tp, X1_fp = X1[tp1], X1[~tp1]
# class 0: dots
plt.scatter(X0_tp[:, 0], X0_tp[:, 1], marker='.', color='red')
plt.scatter(X0_fp[:, 0], X0_fp[:, 1], marker='x',
s=20, color='#990000') # dark red
# class 1: dots
plt.scatter(X1_tp[:, 0], X1_tp[:, 1], marker='.', color='blue')
plt.scatter(X1_fp[:, 0], X1_fp[:, 1], marker='x',
s=20, color='#000099') # dark blue
# class 0 and 1 : areas

nx, ny = 200, 100
x_min, x_max = plt.xlim()
y_min, y_max = plt.ylim()
xx, yy = np.meshgrid(np.linspace(x_min, x_max, nx),
np.linspace(y_min, y_max, ny))
Z = lda.predict_proba(np.c_[xx.ravel(), yy.ravel()])
Z = Z[:, 1].reshape(xx.shape)

plt.pcolormesh(xx, yy, Z, cmap='red_blue_classes',

norm=colors.Normalize(0., 1.), zorder=0)
plt.contour(xx, yy, Z, [0.5], linewidths=2., colors='white')
# means
plt.plot(lda.means_[0][0], lda.means_[0][1],
'*', color='yellow', markersize=15, markeredgecolor='grey')
plt.plot(lda.means_[1][0], lda.means_[1][1],
'*', color='yellow', markersize=15, markeredgecolor='grey')
return splot
def plot_ellipse(splot, mean, cov, color):

v, w = linalg.eigh(cov)
u = w[0] / linalg.norm(w[0])
angle = np.arctan(u[1] / u[0])
angle = 180 * angle / np.pi # convert to degrees
# filled Gaussian at 2 standard deviation
ell = mpl.patches.Ellipse(mean, 2 * v[0] ** 0.5, 2 * v[1] ** 0.5,
180 + angle, facecolor=color,
edgecolor='black', linewidth=2)
ell.set_clip_box(splot.bbox)
ell.set_alpha(0.2)
splot.add_artist(ell)
splot.set_xticks(())
splot.set_yticks(())
def plot_lda_cov(lda, splot):

plot_ellipse(splot, lda.means_[0], lda.covariance_, 'red')
plot_ellipse(splot, lda.means_[1], lda.covariance_, 'blue')
def plot_qda_cov(qda, splot):

plot_ellipse(splot, qda.means_[0], qda.covariance_[0], 'red')
plot_ellipse(splot, qda.means_[1], qda.covariance_[1], 'blue')
plt.figure(figsize=(10, 8), facecolor='white')

for i, (X, y) in enumerate([dataset_fixed_cov(), dataset_cov()]):
# Linear Discriminant Analysis
lda = LinearDiscriminantAnalysis(solver="svd", store_covariance=True)
y_pred = lda.fit(X, y).predict(X)
splot = plot_data(lda, X, y, y_pred, fig_index=2 * i + 1)
plot_lda_cov(lda, splot)
plt.axis('tight')
# Quadratic Discriminant Analysis

qda = QuadraticDiscriminantAnalysis(store_covariance=True)
y_pred = qda.fit(X, y).predict(X)
splot = plot_data(qda, X, y, y_pred, fig_index=2 * i + 2)
plot_qda_cov(qda, splot)
plt.axis('tight')
plt.suptitle('Linear Discriminant Analysis vs Quadratic Discriminant Analysis',
y=1.02, fontsize=15)
plt.tight_layout()
plt.show()

5.6 Clustering
Examples concerning the sklearn.cluster module.
5.6.1 Feature agglomeration
These images how similar features are merged together using feature agglomeration.
print(__doc__)

import numpy as np
from sklearn import datasets, cluster

images = digits.images
X = np.reshape(images, (len(images), -1))
connectivity = grid_to_graph(*images[0].shape)
agglo = cluster.FeatureAgglomeration(connectivity=connectivity,

n_clusters=32)
agglo.fit(X)
X_reduced = agglo.transform(X)
X_restored = agglo.inverse_transform(X_reduced)
images_restored = np.reshape(X_restored, images.shape)
plt.figure(1, figsize=(4, 3.5))
plt.clf()
plt.subplots_adjust(left=.01, right=.99, bottom=.01, top=.91)
for i in range(4):
plt.imshow(images[i], cmap=plt.cm.gray, vmax=16, interpolation='nearest')
plt.xticks(())
plt.yticks(())
if i == 1:
plt.title('Original data')
plt.subplot(3, 4, 4 + i + 1)
plt.imshow(images_restored[i], cmap=plt.cm.gray, vmax=16,
interpolation='nearest')
if i == 1:
plt.title('Agglomerated data')
plt.xticks(())
plt.yticks(())
plt.subplot(3, 4, 10)
plt.imshow(np.reshape(agglo.labels_, images[0].shape),
interpolation='nearest', cmap=plt.cm.nipy_spectral)
plt.xticks(())
plt.yticks(())
plt.title('Labels')
plt.show()
5.6.2 A demo of the mean-shift clustering algorithm
Reference:
Dorin Comaniciu and Peter Meer, “Mean Shift: A robust approach toward feature space analysis”. IEEE Transactions
on Pattern Analysis and Machine Intelligence. 2002. pp. 603-619.
5.6. Clustering 785

Out:
number of estimated clusters : 3
print(__doc__)
import numpy as np
from sklearn.cluster import MeanShift, estimate_bandwidth
from sklearn.datasets.samples_generator import make_blobs
# #############################################################################
centers = [[1, 1], [-1, -1], [1, -1]]
X, _ = make_blobs(n_samples=10000, centers=centers, cluster_std=0.6)
# #############################################################################
# Compute clustering with MeanShift
# The following bandwidth can be automatically detected using

bandwidth = estimate_bandwidth(X, quantile=0.2, n_samples=500)

ms = MeanShift(bandwidth=bandwidth, bin_seeding=True)
ms.fit(X)
labels = ms.labels_
cluster_centers = ms.cluster_centers_
labels_unique = np.unique(labels)
n_clusters_ = len(labels_unique)
print("number of estimated clusters : %d" % n_clusters_)
# #############################################################################
# Plot result
from itertools import cycle
plt.figure(1)
plt.clf()
colors = cycle('bgrcmykbgrcmykbgrcmykbgrcmyk')
for k, col in zip(range(n_clusters_), colors):
my_members = labels == k
cluster_center = cluster_centers[k]
plt.plot(X[my_members, 0], X[my_members, 1], col + '.')
plt.plot(cluster_center[0], cluster_center[1], 'o', markerfacecolor=col,
markeredgecolor='k', markersize=14)
plt.title('Estimated number of clusters: %d' % n_clusters_)
plt.show()
5.6.3 Demonstration of k-means assumptions
This example is meant to illustrate situations where k-means will produce unintuitive and possibly unexpected clusters.
In the first three plots, the input data does not conform to some implicit assumption that k-means makes and undesirable
clusters are produced as a result. In the last plot, k-means returns intuitive clusters despite unevenly sized blobs.
5.6. Clustering 787

print(__doc__)
# Author: Phil Roth <mr.phil.roth@gmail.com>

import numpy as np
from sklearn.cluster import KMeans

n_samples = 1500
random_state = 170

X, y = make_blobs(n_samples=n_samples, random_state=random_state)
# Incorrect number of clusters

y_pred = KMeans(n_clusters=2, random_state=random_state).fit_predict(X)
plt.subplot(221)
plt.scatter(X[:, 0], X[:, 1], c=y_pred)
plt.title("Incorrect Number of Blobs")
# Anisotropicly distributed data

transformation = [[0.60834549, -0.63667341], [-0.40887718, 0.85253229]]
X_aniso = np.dot(X, transformation)
y_pred = KMeans(n_clusters=3, random_state=random_state).fit_predict(X_aniso)
plt.subplot(222)
plt.scatter(X_aniso[:, 0], X_aniso[:, 1], c=y_pred)
plt.title("Anisotropicly Distributed Blobs")
# Different variance
X_varied, y_varied = make_blobs(n_samples=n_samples,
cluster_std=[1.0, 2.5, 0.5],
random_state=random_state)
y_pred = KMeans(n_clusters=3, random_state=random_state).fit_predict(X_varied)
plt.subplot(223)
plt.scatter(X_varied[:, 0], X_varied[:, 1], c=y_pred)
plt.title("Unequal Variance")
# Unevenly sized blobs

X_filtered = np.vstack((X[y == 0][:500], X[y == 1][:100], X[y == 2][:10]))
y_pred = KMeans(n_clusters=3,
random_state=random_state).fit_predict(X_filtered)
plt.subplot(224)
plt.scatter(X_filtered[:, 0], X_filtered[:, 1], c=y_pred)
plt.title("Unevenly Sized Blobs")
plt.show()
5.6.4 Online learning of a dictionary of parts of faces
This example uses a large dataset of faces to learn a set of 20 x 20 images patches that constitute faces.
From the programming standpoint, it is interesting because it shows how to use the online API of the scikit-learn to
process a very large dataset by chunks. The way we proceed is that we load an image at a time and extract randomly
50 patches from this image. Once we have accumulated 500 of these patches (using 10 images), we run the partial_fit
method of the online KMeans object, MiniBatchKMeans.
The verbose setting on the MiniBatchKMeans enables us to see that some clusters are reassigned during the successive
calls to partial-fit. This is because the number of patches that they represent has become too low, and it is better to
choose a random new cluster.
5.6. Clustering 789

Out:
Learning the dictionary...

Partial fit of 100 out of 2400
[MiniBatchKMeans] Reassigning 16 cluster centers.
done in 4.20s.

print(__doc__)
import time

import numpy as np

from sklearn.feature_extraction.image import extract_patches_2d
faces = datasets.fetch_olivetti_faces()
# #############################################################################
# Learn the dictionary of images
print('Learning the dictionary... ')

kmeans = MiniBatchKMeans(n_clusters=81, random_state=rng, verbose=True)
patch_size = (20, 20)
buffer = []
t0 = time.time()
# The online learning part: cycle over the whole dataset 6 times
index = 0
for _ in range(6):
for img in faces.images:
data = extract_patches_2d(img, patch_size, max_patches=50,
random_state=rng)
data = np.reshape(data, (len(data), -1))
buffer.append(data)
index += 1
if index % 10 == 0:
data = np.concatenate(buffer, axis=0)
data -= np.mean(data, axis=0)
data /= np.std(data, axis=0)
kmeans.partial_fit(data)
buffer = []
if index % 100 == 0:
print('Partial fit of %4i out of %i'
% (index, 6 * len(faces.images)))
dt = time.time() - t0
print('done in %.2fs.' % dt)
# #############################################################################
# Plot the results
plt.figure(figsize=(4.2, 4))
for i, patch in enumerate(kmeans.cluster_centers_):
plt.imshow(patch.reshape(patch_size), cmap=plt.cm.gray,
plt.xticks(())
5.6. Clustering 791

plt.yticks(())
plt.suptitle('Patches of faces\nTrain time %.1fs on %d patches' %

(dt, 8 * len(faces.images)), fontsize=16)
plt.subplots_adjust(0.08, 0.02, 0.92, 0.85, 0.08, 0.23)
plt.show()
5.6.5 Vector Quantization Example
Face, a 1024 x 768 size image of a raccoon face, is used here to illustrate how k-means is used for vector quantization.

print(__doc__)

import numpy as np
import scipy as sp
from sklearn import cluster
try: # SciPy >= 0.16 have face in misc

from scipy.misc import face
face = face(gray=True)
except ImportError:
face = sp.face(gray=True)
n_clusters = 5
np.random.seed(0)
X = face.reshape((-1, 1)) # We need an (n_sample, n_feature) array

k_means = cluster.KMeans(n_clusters=n_clusters, n_init=4)
k_means.fit(X)
values = k_means.cluster_centers_.squeeze()
labels = k_means.labels_
5.6. Clustering 793

# create an array from labels and values

face_compressed = np.choose(labels, values)
face_compressed.shape = face.shape
vmin = face.min()
vmax = face.max()
# original face
plt.imshow(face, cmap=plt.cm.gray, vmin=vmin, vmax=256)
# compressed face
plt.imshow(face_compressed, cmap=plt.cm.gray, vmin=vmin, vmax=vmax)
# equal bins face

regular_values = np.linspace(0, 256, n_clusters + 1)
regular_labels = np.searchsorted(regular_values, face) - 1
regular_values = .5 * (regular_values[1:] + regular_values[:-1]) # mean
regular_face = np.choose(regular_labels.ravel(), regular_values, mode="clip")
regular_face.shape = face.shape
plt.imshow(regular_face, cmap=plt.cm.gray, vmin=vmin, vmax=vmax)
# histogram
plt.clf()
plt.axes([.01, .01, .98, .98])
plt.hist(X, bins=256, color='.5', edgecolor='.5')
plt.yticks(())
plt.xticks(regular_values)
values = np.sort(values)
for center_1, center_2 in zip(values[:-1], values[1:]):
plt.axvline(.5 * (center_1 + center_2), color='b')
for center_1, center_2 in zip(regular_values[:-1], regular_values[1:]):

plt.axvline(.5 * (center_1 + center_2), color='b', linestyle='--')
plt.show()
5.6.6 Segmenting the picture of greek coins in regions
This example uses Spectral clustering on a graph created from voxel-to-voxel difference on an image to break this
image into multiple partly-homogeneous regions.
This procedure (spectral clustering on an image) is an efficient approximate solution for finding normalized graph cuts.
There are two options to assign labels:
• with ‘kmeans’ spectral clustering will cluster samples in the embedding space using a kmeans algorithm

• whereas ‘discrete’ will iteratively search for the closest partition space to the embedding space.
print(__doc__)
# Author: Gael Varoquaux <gael.varoquaux@normalesup.org>, Brian Cheung

import time
import numpy as np
from scipy.ndimage.filters import gaussian_filter
from sklearn.feature_extraction import image

from sklearn.cluster import spectral_clustering
# load the coins as a numpy array


# Convert the image into a graph with the value of the gradient on the
# edges.
graph = image.img_to_graph(rescaled_coins)
# Take a decreasing function of the gradient: an exponential

# The smaller beta is, the more independent the segmentation is of the
# actual image. For beta=1, the segmentation is close to a voronoi
beta = 10
eps = 1e-6
graph.data = np.exp(-beta * graph.data / graph.data.std()) + eps
# Apply spectral clustering (this step goes much faster if you have pyamg
# installed)
N_REGIONS = 25
Visualize the resulting regions

for assign_labels in ('kmeans', 'discretize'):
t0 = time.time()
labels = spectral_clustering(graph, n_clusters=N_REGIONS,
assign_labels=assign_labels, random_state=42)
t1 = time.time()
labels = labels.reshape(rescaled_coins.shape)
plt.imshow(rescaled_coins, cmap=plt.cm.gray)
for l in range(N_REGIONS):
plt.contour(labels == l,
colors=[plt.cm.nipy_spectral(l / float(N_REGIONS))])
plt.xticks(())
plt.yticks(())
5.6. Clustering 795

title = 'Spectral clustering: %s, %.2fs' % (assign_labels, (t1 - t0))

print(title)
plt.title(title)
plt.show()

•
Out:
Spectral clustering: kmeans, 4.66s

Spectral clustering: discretize, 3.96s
5.6.7 A demo of structured Ward hierarchical clustering on an image of coins
Compute the segmentation of a 2D image with Ward hierarchical clustering. The clustering is spatially constrained in
order for each segmented region to be in one piece.
5.6. Clustering 797

Out:
Compute structured hierarchical clustering...
Elapsed time: 0.3394918441772461
Number of pixels: 4697
Number of clusters: 27
# Author : Vincent Michel, 2010

# Alexandre Gramfort, 2011
print(__doc__)
import time as time
import numpy as np
from scipy.ndimage.filters import gaussian_filter



# #############################################################################
# Generate data

X = np.reshape(rescaled_coins, (-1, 1))
# #############################################################################
# Define the structure A of the data. Pixels connected to their neighbors.
connectivity = grid_to_graph(*rescaled_coins.shape)
# #############################################################################
# Compute clustering
print("Compute structured hierarchical clustering...")
st = time.time()
n_clusters = 27 # number of regions
ward = AgglomerativeClustering(n_clusters=n_clusters, linkage='ward',
connectivity=connectivity)
ward.fit(X)
label = np.reshape(ward.labels_, rescaled_coins.shape)
print("Elapsed time: ", time.time() - st)
print("Number of pixels: ", label.size)
print("Number of clusters: ", np.unique(label).size)
# #############################################################################
# Plot the results on an image
plt.imshow(rescaled_coins, cmap=plt.cm.gray)
for l in range(n_clusters):
plt.contour(label == l,
colors=[plt.cm.nipy_spectral(l / float(n_clusters)), ])
plt.xticks(())
plt.yticks(())
plt.show()
5.6.8 Agglomerative clustering with and without structure
This example shows the effect of imposing a connectivity graph to capture local structure in the data. The graph is
simply the graph of 20 nearest neighbors.
5.6. Clustering 799

Two consequences of imposing a connectivity can be seen. First clustering with a connectivity matrix is much faster.
Second, when using a connectivity matrix, single, average and complete linkage are unstable and tend to create a few
clusters that grow very quickly. Indeed, average and complete linkage fight this percolation behavior by considering all
the distances between two clusters when merging them ( while single linkage exaggerates the behaviour by considering
only the shortest distance between clusters). The connectivity graph breaks this mechanism for average and complete
linkage, making them resemble the more brittle single linkage. This effect is more pronounced for very sparse graphs
(try decreasing the number of neighbors in kneighbors_graph) and with complete linkage. In particular, having a very
small number of neighbors in the graph, imposes a geometry that is close to that of single linkage, which is well known
to have this percolation instability.

•
# Authors: Gael Varoquaux, Nelle Varoquaux
import time
import numpy as np

from sklearn.neighbors import kneighbors_graph

n_samples = 1500
np.random.seed(0)
t = 1.5 * np.pi * (1 + 3 * np.random.rand(1, n_samples))
x = t * np.cos(t)
y = t * np.sin(t)
X = np.concatenate((x, y))
X += .7 * np.random.randn(2, n_samples)
X = X.T
# Create a graph capturing local connectivity. Larger number of neighbors

# will give more homogeneous clusters to the cost of computation
# time. A very large number of neighbors gives more evenly distributed
5.6. Clustering 801

# cluster sizes, but may not impose the local manifold structure of
# the data
knn_graph = kneighbors_graph(X, 30, include_self=False)
for connectivity in (None, knn_graph):

for n_clusters in (30, 3):
for index, linkage in enumerate(('average',
'complete',
'ward',
'single')):
model = AgglomerativeClustering(linkage=linkage,
connectivity=connectivity,
n_clusters=n_clusters)
t0 = time.time()
model.fit(X)
elapsed_time = time.time() - t0
plt.scatter(X[:, 0], X[:, 1], c=model.labels_,
cmap=plt.cm.nipy_spectral)
plt.title('linkage=%s\n(time %.2fs)' % (linkage, elapsed_time),
fontdict=dict(verticalalignment='top'))
plt.axis('equal')
plt.axis('off')
plt.subplots_adjust(bottom=0, top=.89, wspace=0,

left=0, right=1)
plt.suptitle('n_cluster=%i, connectivity=%r' %
(n_clusters, connectivity is not None), size=17)
plt.show()
5.6.9 Demo of affinity propagation clustering algorithm
Reference: Brendan J. Frey and Delbert Dueck, “Clustering by Passing Messages Between Data Points”, Science Feb.
2007

Out:
Estimated number of clusters: 3

Homogeneity: 0.872
Completeness: 0.872
V-measure: 0.872
Adjusted Rand Index: 0.912
Adjusted Mutual Information: 0.871
Silhouette Coefficient: 0.753
print(__doc__)
from sklearn.cluster import AffinityPropagation

from sklearn import metrics
# #############################################################################
centers = [[1, 1], [-1, -1], [1, -1]]
X, labels_true = make_blobs(n_samples=300, centers=centers, cluster_std=0.5,
random_state=0)
5.6. Clustering 803

# #############################################################################
# Compute Affinity Propagation
af = AffinityPropagation(preference=-50).fit(X)
cluster_centers_indices = af.cluster_centers_indices_
labels = af.labels_
n_clusters_ = len(cluster_centers_indices)
print('Estimated number of clusters: %d' % n_clusters_)

print("Homogeneity: %0.3f" % metrics.homogeneity_score(labels_true, labels))
print("Completeness: %0.3f" % metrics.completeness_score(labels_true, labels))
print("V-measure: %0.3f" % metrics.v_measure_score(labels_true, labels))
print("Adjusted Rand Index: %0.3f"
% metrics.adjusted_rand_score(labels_true, labels))
print("Adjusted Mutual Information: %0.3f"
% metrics.adjusted_mutual_info_score(labels_true, labels))
print("Silhouette Coefficient: %0.3f"
% metrics.silhouette_score(X, labels, metric='sqeuclidean'))
# #############################################################################
# Plot result
plt.close('all')
plt.figure(1)
plt.clf()
colors = cycle('bgrcmykbgrcmykbgrcmykbgrcmyk')
for k, col in zip(range(n_clusters_), colors):
class_members = labels == k
cluster_center = X[cluster_centers_indices[k]]
plt.plot(X[class_members, 0], X[class_members, 1], col + '.')
plt.plot(cluster_center[0], cluster_center[1], 'o', markerfacecolor=col,
for x in X[class_members]:
plt.plot([cluster_center[0], x[0]], [cluster_center[1], x[1]], col)

plt.show()
5.6.10 K-means Clustering
The plots display firstly what a K-means algorithm would yield using three clusters. It is then shown what the effect
of a bad initialization is on the classification process: By setting n_init to only 1 (default is 10), the amount of times
that the algorithm will be run with different centroid seeds is reduced. The next plot displays what using eight clusters
would deliver and finally the ground truth.

5.6. Clustering 805

•
print(__doc__)

import numpy as np
# Though the following import is not directly being used, it is required
# for 3D projection to work
from mpl_toolkits.mplot3d import Axes3D

np.random.seed(5)

X = iris.data
y = iris.target
estimators = [('k_means_iris_8', KMeans(n_clusters=8)),

('k_means_iris_3', KMeans(n_clusters=3)),
('k_means_iris_bad_init', KMeans(n_clusters=3, n_init=1,
init='random'))]
fignum = 1
titles = ['8 clusters', '3 clusters', '3 clusters, bad initialization']
for name, est in estimators:
fig = plt.figure(fignum, figsize=(4, 3))
ax = Axes3D(fig, rect=[0, 0, .95, 1], elev=48, azim=134)
est.fit(X)
labels = est.labels_
ax.scatter(X[:, 3], X[:, 0], X[:, 2],

c=labels.astype(np.float), edgecolor='k')
ax.w_xaxis.set_ticklabels([])
ax.w_yaxis.set_ticklabels([])
ax.w_zaxis.set_ticklabels([])
ax.set_xlabel('Petal width')
ax.set_ylabel('Sepal length')
ax.set_zlabel('Petal length')
ax.set_title(titles[fignum - 1])
ax.dist = 12
fignum = fignum + 1
# Plot the ground truth

fig = plt.figure(fignum, figsize=(4, 3))
for name, label in [('Setosa', 0),

('Versicolour', 1),
('Virginica', 2)]:
ax.text3D(X[y == label, 3].mean(),
X[y == label, 0].mean(),
X[y == label, 2].mean() + 2, name,
horizontalalignment='center',
bbox=dict(alpha=.2, edgecolor='w', facecolor='w'))
# Reorder the labels to have colors matching the cluster results
y = np.choose(y, [1, 2, 0]).astype(np.float)
ax.scatter(X[:, 3], X[:, 0], X[:, 2], c=y, edgecolor='k')
ax.set_xlabel('Petal width')
ax.set_ylabel('Sepal length')
ax.set_zlabel('Petal length')
ax.set_title('Ground Truth')
ax.dist = 12
fig.show()
5.6. Clustering 807

5.6.11 Various Agglomerative Clustering on a 2D embedding of digits
An illustration of various linkage option for agglomerative clustering on a 2D embedding of the digits dataset.
The goal of this example is to show intuitively how the metrics behave, and not to find good clusters for the digits.
This is why the example works on a 2D embedding.
What this example shows us is the behavior “rich getting richer” of agglomerative clustering that tends to create uneven
cluster sizes. This behavior is pronounced for the average linkage strategy, that ends up with a couple of singleton
clusters, while in the case of single linkage we get a single central cluster with all other clusters being drawn from
noise points around the fringes.

5.6. Clustering 809

•
Out:
Computing embedding
Done.
ward : 0.35s
average : 0.29s
complete : 0.28s
single : 0.15s
# Authors: Gael Varoquaux

# License: BSD 3 clause (C) INRIA 2014
print(__doc__)
import numpy as np
from scipy import ndimage
from sklearn import manifold, datasets
X = digits.data
y = digits.target
n_samples, n_features = X.shape

np.random.seed(0)
def nudge_images(X, y):

# Having a larger dataset shows more clearly the behavior of the
# methods, but we multiply the size of the dataset only by 2, as the
# cost of the hierarchical clustering methods are strongly
# super-linear in n_samples
shift = lambda x: ndimage.shift(x.reshape((8, 8)),
.3 * np.random.normal(size=2),
mode='constant',
).ravel()
X = np.concatenate([X, np.apply_along_axis(shift, 1, X)])
Y = np.concatenate([y, y], axis=0)
return X, Y
X, y = nudge_images(X, y)
#----------------------------------------------------------------------
# Visualize the clustering
def plot_clustering(X_red, labels, title=None):
x_min, x_max = np.min(X_red, axis=0), np.max(X_red, axis=0)
X_red = (X_red - x_min) / (x_max - x_min)
for i in range(X_red.shape[0]):
plt.text(X_red[i, 0], X_red[i, 1], str(y[i]),
color=plt.cm.nipy_spectral(labels[i] / 10.),
fontdict={'weight': 'bold', 'size': 9})
plt.xticks([])
plt.yticks([])
if title is not None:
plt.title(title, size=17)
plt.axis('off')
plt.tight_layout(rect=[0, 0.03, 1, 0.95])
#----------------------------------------------------------------------
# 2D embedding of the digits dataset
print("Computing embedding")
X_red = manifold.SpectralEmbedding(n_components=2).fit_transform(X)
print("Done.")
for linkage in ('ward', 'average', 'complete', 'single'):

clustering = AgglomerativeClustering(linkage=linkage, n_clusters=10)
t0 = time()
clustering.fit(X_red)
print("%s :\t%.2fs" % (linkage, time() - t0))
plot_clustering(X_red, clustering.labels_, "%s linkage" % linkage)
plt.show()
5.6. Clustering 811

5.6.12 Spectral clustering for image segmentation
In this example, an image with connected circles is generated and spectral clustering is used to separate the circles.
In these settings, the Spectral clustering approach solves the problem know as ‘normalized graph cuts’: the image is
seen as a graph of connected voxels, and the spectral clustering algorithm amounts to choosing graph cuts defining
regions while minimizing the ratio of the gradient along the cut, and the volume of the region.
As the algorithm tries to balance the volume (ie balance the region sizes), if we take circles with different sizes, the
segmentation fails.
In addition, as there is no useful information in the intensity of the image, or its gradient, we choose to perform the
spectral clustering on a graph that is only weakly informed by the gradient. This is close to performing a Voronoi
partition of the graph.
In addition, we use the mask of the objects to restrict the graph to the outline of the objects. In this example, we are
interested in separating the objects one from the other, and not from the background.

5.6. Clustering 813


print(__doc__)
# Authors: Emmanuelle Gouillart <emmanuelle.gouillart@normalesup.org>

# Gael Varoquaux <gael.varoquaux@normalesup.org>
import numpy as np
from sklearn.feature_extraction import image

from sklearn.cluster import spectral_clustering
l = 100
x, y = np.indices((l, l))
center1 = (28, 24)

center2 = (40, 50)
center3 = (67, 58)
center4 = (24, 70)
radius1, radius2, radius3, radius4 = 16, 14, 15, 14
circle1 = (x - center1[0]) ** 2 + (y - center1[1]) ** 2 < radius1 ** 2

5.6. Clustering 815

# #############################################################################
# 4 circles
img = circle1 + circle2 + circle3 + circle4
# We use a mask that limits to the foreground: the problem that we are
# interested in here is not separating the objects from the background,
# but separating them one from the other.
mask = img.astype(bool)
img = img.astype(float)
img += 1 + 0.2 * np.random.randn(*img.shape)
# Convert the image into a graph with the value of the gradient on the
# edges.
graph = image.img_to_graph(img, mask=mask)
# Take a decreasing function of the gradient: we take it weakly

# dependent from the gradient the segmentation is close to a voronoi
graph.data = np.exp(-graph.data / graph.data.std())
# Force the solver to be arpack, since amg is numerically

# unstable on this example
labels = spectral_clustering(graph, n_clusters=4, eigen_solver='arpack')
label_im = np.full(mask.shape, -1.)
label_im[mask] = labels
plt.matshow(img)
plt.matshow(label_im)
# #############################################################################
# 2 circles
img = circle1 + circle2
mask = img.astype(bool)
img = img.astype(float)
img += 1 + 0.2 * np.random.randn(*img.shape)
graph = image.img_to_graph(img, mask=mask)

graph.data = np.exp(-graph.data / graph.data.std())
labels = spectral_clustering(graph, n_clusters=2, eigen_solver='arpack')

label_im = np.full(mask.shape, -1.)
label_im[mask] = labels
plt.matshow(img)
plt.matshow(label_im)
plt.show()

5.6.13 Demo of DBSCAN clustering algorithm
Finds core samples of high density and expands clusters from them.
Out:
Estimated number of clusters: 3

Estimated number of noise points: 18
Homogeneity: 0.953
Completeness: 0.883
V-measure: 0.917
Adjusted Rand Index: 0.952
Adjusted Mutual Information: 0.883
print(__doc__)
import numpy as np
from sklearn.cluster import DBSCAN

5.6. Clustering 817


# #############################################################################
centers = [[1, 1], [-1, -1], [1, -1]]
X, labels_true = make_blobs(n_samples=750, centers=centers, cluster_std=0.4,
random_state=0)
# #############################################################################
# Compute DBSCAN
db = DBSCAN(eps=0.3, min_samples=10).fit(X)
core_samples_mask = np.zeros_like(db.labels_, dtype=bool)
core_samples_mask[db.core_sample_indices_] = True
labels = db.labels_
# Number of clusters in labels, ignoring noise if present.

n_clusters_ = len(set(labels)) - (1 if -1 in labels else 0)
n_noise_ = list(labels).count(-1)
print('Estimated number of clusters: %d' % n_clusters_)

print('Estimated number of noise points: %d' % n_noise_)
print("Homogeneity: %0.3f" % metrics.homogeneity_score(labels_true, labels))
print("Completeness: %0.3f" % metrics.completeness_score(labels_true, labels))
print("V-measure: %0.3f" % metrics.v_measure_score(labels_true, labels))
print("Adjusted Rand Index: %0.3f"
% metrics.adjusted_rand_score(labels_true, labels))
print("Adjusted Mutual Information: %0.3f"
% metrics.adjusted_mutual_info_score(labels_true, labels))
% metrics.silhouette_score(X, labels))
# #############################################################################
# Plot result
# Black removed and is used for noise instead.

unique_labels = set(labels)
colors = [plt.cm.Spectral(each)
for each in np.linspace(0, 1, len(unique_labels))]
for k, col in zip(unique_labels, colors):
if k == -1:
# Black used for noise.
col = [0, 0, 0, 1]
class_member_mask = (labels == k)
xy = X[class_member_mask & core_samples_mask]

plt.plot(xy[:, 0], xy[:, 1], 'o', markerfacecolor=tuple(col),
xy = X[class_member_mask & ~core_samples_mask]

plt.plot(xy[:, 0], xy[:, 1], 'o', markerfacecolor=tuple(col),


plt.show()
5.6.14 Color Quantization using K-Means
Performs a pixel-wise Vector Quantization (VQ) of an image of the summer palace (China), reducing the number of
colors required to show the image from 96,615 unique colors to 64, while preserving the overall appearance quality.
In this example, pixels are represented in a 3D-space and K-means is used to find 64 color clusters. In the image
processing literature, the codebook obtained from K-means (the cluster centers) is called the color palette. Using a
single byte, up to 256 colors can be addressed, whereas an RGB encoding requires 3 bytes per pixel. The GIF file
format, for example, uses such a palette.
For comparison, a quantized image using a random codebook (colors picked up randomly) is also shown.
5.6. Clustering 819


•
Out:
Fitting model on a small sub-sample of the data

done in 0.242s.
Predicting color indices on the full image (k-means)
done in 0.264s.
Predicting color indices on the full image (random)
done in 0.400s.
# Authors: Robert Layton <robertlayton@gmail.com>

# Olivier Grisel <olivier.grisel@ensta.org>
# Mathieu Blondel <mathieu@mblondel.org>
#
print(__doc__)
import numpy as np
from sklearn.metrics import pairwise_distances_argmin
from sklearn.datasets import load_sample_image
5.6. Clustering 821

n_colors = 64
# Load the Summer Palace photo

china = load_sample_image("china.jpg")
# Convert to floats instead of the default 8 bits integer coding. Dividing by

# 255 is important so that plt.imshow behaves works well on float data (need to
# be in the range [0-1])
china = np.array(china, dtype=np.float64) / 255
# Load Image and transform to a 2D numpy array.

w, h, d = original_shape = tuple(china.shape)
assert d == 3
image_array = np.reshape(china, (w * h, d))
print("Fitting model on a small sub-sample of the data")

t0 = time()
image_array_sample = shuffle(image_array, random_state=0)[:1000]
kmeans = KMeans(n_clusters=n_colors, random_state=0).fit(image_array_sample)
# Get labels for all points

print("Predicting color indices on the full image (k-means)")
t0 = time()
labels = kmeans.predict(image_array)
codebook_random = shuffle(image_array, random_state=0)[:n_colors]

print("Predicting color indices on the full image (random)")
t0 = time()
labels_random = pairwise_distances_argmin(codebook_random,
image_array,
axis=0)
def recreate_image(codebook, labels, w, h):

"""Recreate the (compressed) image from the code book & labels"""
d = codebook.shape[1]
image = np.zeros((w, h, d))
label_idx = 0
for i in range(w):
for j in range(h):
image[i][j] = codebook[labels[label_idx]]
label_idx += 1
return image
# Display all results, alongside original image

plt.figure(1)
plt.clf()
plt.axis('off')
plt.title('Original image (96,615 colors)')
plt.imshow(china)
plt.figure(2)

plt.clf()
plt.axis('off')
plt.title('Quantized image (64 colors, K-Means)')
plt.imshow(recreate_image(kmeans.cluster_centers_, labels, w, h))
plt.figure(3)
plt.clf()
plt.axis('off')
plt.title('Quantized image (64 colors, Random)')
plt.imshow(recreate_image(codebook_random, labels_random, w, h))
plt.show()
5.6.15 Hierarchical clustering: structured vs unstructured ward
Example builds a swiss roll dataset and runs hierarchical clustering on their position.
For more information, see Hierarchical clustering.
In a first step, the hierarchical clustering is performed without connectivity constraints on the structure and is solely
based on distance, whereas in a second step the clustering is restricted to the k-Nearest Neighbors graph: it’s a
hierarchical clustering with structure prior.
Some of the clusters learned without connectivity constraints do not respect the structure of the swiss roll and extend
across different folds of the manifolds. On the opposite, when opposing connectivity constraints, the clusters form a
nice parcellation of the swiss roll.
5.6. Clustering 823


•
Out:
Compute unstructured hierarchical clustering...

Elapsed time: 0.04s
Number of points: 1500
Compute structured hierarchical clustering...
Elapsed time: 0.07s
Number of points: 1500
# Authors : Vincent Michel, 2010

# Alexandre Gramfort, 2010
# Gael Varoquaux, 2010
print(__doc__)
import time as time

import numpy as np
import mpl_toolkits.mplot3d.axes3d as p3
from sklearn.datasets.samples_generator import make_swiss_roll
5.6. Clustering 825

# #############################################################################
# Generate data (swiss roll dataset)
n_samples = 1500
noise = 0.05
X, _ = make_swiss_roll(n_samples, noise)
# Make it thinner
X[:, 1] *= .5
# #############################################################################
print("Compute unstructured hierarchical clustering...")
st = time.time()
ward = AgglomerativeClustering(n_clusters=6, linkage='ward').fit(X)
elapsed_time = time.time() - st
label = ward.labels_
print("Elapsed time: %.2fs" % elapsed_time)
print("Number of points: %i" % label.size)
# #############################################################################
# Plot result
fig = plt.figure()
ax = p3.Axes3D(fig)
ax.view_init(7, -80)
for l in np.unique(label):
ax.scatter(X[label == l, 0], X[label == l, 1], X[label == l, 2],
color=plt.cm.jet(np.float(l) / np.max(label + 1)),
s=20, edgecolor='k')
plt.title('Without connectivity constraints (time %.2fs)' % elapsed_time)
# #############################################################################
# Define the structure A of the data. Here a 10 nearest neighbors
connectivity = kneighbors_graph(X, n_neighbors=10, include_self=False)
# #############################################################################
print("Compute structured hierarchical clustering...")
st = time.time()
ward = AgglomerativeClustering(n_clusters=6, connectivity=connectivity,
linkage='ward').fit(X)
elapsed_time = time.time() - st
label = ward.labels_
print("Elapsed time: %.2fs" % elapsed_time)
print("Number of points: %i" % label.size)
# #############################################################################
# Plot result
fig = plt.figure()
ax = p3.Axes3D(fig)
for l in np.unique(label):
ax.scatter(X[label == l, 0], X[label == l, 1], X[label == l, 2],
color=plt.cm.jet(float(l) / np.max(label + 1)),
plt.title('With connectivity constraints (time %.2fs)' % elapsed_time)

plt.show()
5.6.16 Agglomerative clustering with different metrics
Demonstrates the effect of different metrics on the hierarchical clustering.

The example is engineered to show the effect of the choice of different metrics. It is applied to waveforms, which
can be seen as high-dimensional vector. Indeed, the difference between metrics is usually more pronounced in high
dimension (in particular for euclidean and cityblock).
We generate data from three groups of waveforms. Two of the waveforms (waveform 1 and waveform 2) are propor-
tional one to the other. The cosine distance is invariant to a scaling of the data, as a result, it cannot distinguish these
two waveforms. Thus even with no noise, clustering using this distance will not separate out waveform 1 and 2.
We add observation noise to these waveforms. We generate very sparse noise: only 6% of the time points contain
noise. As a result, the l1 norm of this noise (ie “cityblock” distance) is much smaller than it’s l2 norm (“euclidean”
distance). This can be seen on the inter-class distance matrices: the values on the diagonal, that characterize the spread
of the class, are much bigger for the Euclidean distance than for the cityblock distance.
When we apply clustering to the data, we find that the clustering reflects what was in the distance matrices. Indeed,
for the Euclidean distance, the classes are ill-separated because of the noise, and thus the clustering does not separate
the waveforms. For the cityblock distance, the separation is good and the waveform classes are recovered. Finally, the
cosine distance does not separate at all waveform 1 and 2, thus the clustering puts them in the same cluster.
5.6. Clustering 827


5.6. Clustering 829


5.6. Clustering 831


5.6. Clustering 833

# Author: Gael Varoquaux

# License: BSD 3-Clause or CC-0

import numpy as np

from sklearn.metrics import pairwise_distances
np.random.seed(0)
# Generate waveform data

n_features = 2000
t = np.pi * np.linspace(0, 1, n_features)
def sqr(x):
return np.sign(np.cos(x))
X = list()
y = list()
for i, (phi, a) in enumerate([(.5, .15), (.5, .6), (.3, .2)]):
for _ in range(30):
phase_noise = .01 * np.random.normal()
amplitude_noise = .04 * np.random.normal()
additional_noise = 1 - 2 * np.random.rand(n_features)
# Make the noise sparse

additional_noise[np.abs(additional_noise) < .997] = 0
X.append(12 * ((a + amplitude_noise)

* (sqr(6 * (t + phi + phase_noise)))
+ additional_noise))
y.append(i)
X = np.array(X)
y = np.array(y)
n_clusters = 3
labels = ('Waveform 1', 'Waveform 2', 'Waveform 3')
# Plot the ground-truth labelling

plt.figure()
plt.axes([0, 0, 1, 1])
for l, c, n in zip(range(n_clusters), 'rgb',
labels):
lines = plt.plot(X[y == l].T, c=c, alpha=.5)
lines[0].set_label(n)
plt.legend(loc='best')
plt.axis('tight')
plt.axis('off')
plt.suptitle("Ground truth", size=20)
# Plot the distances

for index, metric in enumerate(["cosine", "euclidean", "cityblock"]):
avg_dist = np.zeros((n_clusters, n_clusters))
for i in range(n_clusters):
for j in range(n_clusters):
avg_dist[i, j] = pairwise_distances(X[y == i], X[y == j],
metric=metric).mean()
avg_dist /= avg_dist.max()
for j in range(n_clusters):
plt.text(i, j, '%5.3f' % avg_dist[i, j],
verticalalignment='center',
horizontalalignment='center')
plt.imshow(avg_dist, interpolation='nearest', cmap=plt.cm.gnuplot2,

vmin=0)
plt.xticks(range(n_clusters), labels, rotation=45)
plt.yticks(range(n_clusters), labels)
plt.colorbar()
plt.suptitle("Interclass %s distances" % metric, size=18)
plt.tight_layout()
# Plot clustering results

for index, metric in enumerate(["cosine", "euclidean", "cityblock"]):
model = AgglomerativeClustering(n_clusters=n_clusters,
linkage="average", affinity=metric)
model.fit(X)
5.6. Clustering 835

plt.figure()
plt.axes([0, 0, 1, 1])
for l, c in zip(np.arange(model.n_clusters), 'rgbk'):
plt.plot(X[model.labels_ == l].T, c=c, alpha=.5)
plt.axis('tight')
plt.axis('off')
plt.suptitle("AgglomerativeClustering(affinity=%s)" % metric, size=20)
plt.show()
5.6.17 Compare BIRCH and MiniBatchKMeans
This example compares the timing of Birch (with and without the global clustering step) and MiniBatchKMeans on a
synthetic dataset having 100,000 samples and 2 features generated using make_blobs.
If n_clusters is set to None, the data is reduced from 100,000 samples to a set of 158 clusters. This can be viewed
as a preprocessing step before the final (global) clustering step that further reduces these 158 clusters to 100 clusters.
Out:
Birch without global clustering as the final step took 2.35 seconds
n_clusters : 158
Birch with global clustering as the final step took 2.32 seconds
n_clusters : 100
Time taken to run MiniBatchKMeans 3.26 seconds
# Authors: Manoj Kumar <manojkumarsivaraj334@gmail.com

# Alexandre Gramfort <alexandre.gramfort@telecom-paristech.fr>

print(__doc__)

import numpy as np
import matplotlib.colors as colors
from sklearn.cluster import Birch, MiniBatchKMeans

# Generate centers for the blobs so that it forms a 10 X 10 grid.

xx = np.linspace(-22, 22, 10)
yy = np.linspace(-22, 22, 10)
xx, yy = np.meshgrid(xx, yy)
n_centres = np.hstack((np.ravel(xx)[:, np.newaxis],
np.ravel(yy)[:, np.newaxis]))
# Generate blobs to do a comparison between MiniBatchKMeans and Birch.

X, y = make_blobs(n_samples=100000, centers=n_centres, random_state=0)
# Use all colors that matplotlib provides by default.

colors_ = cycle(colors.cnames.keys())
fig = plt.figure(figsize=(12, 4))

fig.subplots_adjust(left=0.04, right=0.98, bottom=0.1, top=0.9)
# Compute clustering with Birch with and without the final clustering step
# and plot.
birch_models = [Birch(threshold=1.7, n_clusters=None),
Birch(threshold=1.7, n_clusters=100)]
final_step = ['without global clustering', 'with global clustering']
for ind, (birch_model, info) in enumerate(zip(birch_models, final_step)):

t = time()
birch_model.fit(X)
time_ = time() - t
print("Birch %s as the final step took %0.2f seconds" % (
info, (time() - t)))
# Plot result
labels = birch_model.labels_
centroids = birch_model.subcluster_centers_
n_clusters = np.unique(labels).size
print("n_clusters : %d" % n_clusters)
ax = fig.add_subplot(1, 3, ind + 1)
for this_centroid, k, col in zip(centroids, range(n_clusters), colors_):
mask = labels == k
ax.scatter(X[mask, 0], X[mask, 1],
c='w', edgecolor=col, marker='.', alpha=0.5)
if birch_model.n_clusters is None:
ax.scatter(this_centroid[0], this_centroid[1], marker='+',
c='k', s=25)
ax.set_ylim([-25, 25])
ax.set_xlim([-25, 25])
ax.set_autoscaley_on(False)
5.6. Clustering 837

ax.set_title('Birch %s' % info)
# Compute clustering with MiniBatchKMeans.

mbk = MiniBatchKMeans(init='k-means++', n_clusters=100, batch_size=100,
n_init=10, max_no_improvement=10, verbose=0,
random_state=0)
t0 = time()
mbk.fit(X)
t_mini_batch = time() - t0
print("Time taken to run MiniBatchKMeans %0.2f seconds" % t_mini_batch)
mbk_means_labels_unique = np.unique(mbk.labels_)
ax = fig.add_subplot(1, 3, 3)
for this_centroid, k, col in zip(mbk.cluster_centers_,
range(n_clusters), colors_):
mask = mbk.labels_ == k
ax.scatter(X[mask, 0], X[mask, 1], marker='.',
c='w', edgecolor=col, alpha=0.5)
ax.scatter(this_centroid[0], this_centroid[1], marker='+',
c='k', s=25)
ax.set_xlim([-25, 25])
ax.set_ylim([-25, 25])
ax.set_title("MiniBatchKMeans")
ax.set_autoscaley_on(False)
plt.show()
5.6.18 Empirical evaluation of the impact of k-means initialization
Evaluate the ability of k-means initializations strategies to make the algorithm convergence robust as measured by
the relative standard deviation of the inertia of the clustering (i.e. the sum of squared distances to the nearest cluster
center).
The first plot shows the best inertia reached for each combination of the model (KMeans or MiniBatchKMeans)
and the init method (init="random" or init="kmeans++") for increasing values of the n_init parameter
that controls the number of initializations.
The second plot demonstrate one single run of the MiniBatchKMeans estimator using a init="random" and
n_init=1. This run leads to a bad convergence (local optimum) with estimated centers stuck between ground truth
clusters.
The dataset used for evaluation is a 2D grid of isotropic Gaussian clusters widely spaced.

5.6. Clustering 839

•
Out:
Evaluation of KMeans with k-means++ init

Evaluation of KMeans with random init
Evaluation of MiniBatchKMeans with k-means++ init
Evaluation of MiniBatchKMeans with random init
print(__doc__)

import numpy as np
import matplotlib.cm as cm

random_state = np.random.RandomState(0)

# Number of run (with randomly generated dataset) for each strategy so as

# to be able to compute an estimate of the standard deviation
n_runs = 5
# k-means models can do several random inits so as to be able to trade

# CPU time for convergence robustness
n_init_range = np.array([1, 5, 10, 15, 20])
# Datasets generation parameters

n_samples_per_center = 100
grid_size = 3
scale = 0.1
n_clusters = grid_size ** 2
def make_data(random_state, n_samples_per_center, grid_size, scale):

random_state = check_random_state(random_state)
centers = np.array([[i, j]
for i in range(grid_size)
for j in range(grid_size)])
n_clusters_true, n_features = centers.shape
noise = random_state.normal(
scale=scale, size=(n_samples_per_center, centers.shape[1]))
X = np.concatenate([c + noise for c in centers])

y = np.concatenate([[i] * n_samples_per_center
for i in range(n_clusters_true)])
return shuffle(X, y, random_state=random_state)
# Part 1: Quantitative evaluation of various init methods
plt.figure()
plots = []
legends = []
cases = [
(KMeans, 'k-means++', {}),
(KMeans, 'random', {}),
(MiniBatchKMeans, 'k-means++', {'max_no_improvement': 3}),
(MiniBatchKMeans, 'random', {'max_no_improvement': 3, 'init_size': 500}),
]
for factory, init, params in cases:

print("Evaluation of %s with %s init" % (factory.__name__, init))
inertia = np.empty((len(n_init_range), n_runs))
for run_id in range(n_runs):

X, y = make_data(run_id, n_samples_per_center, grid_size, scale)
for i, n_init in enumerate(n_init_range):
km = factory(n_clusters=n_clusters, init=init, random_state=run_id,
n_init=n_init, **params).fit(X)
inertia[i, run_id] = km.inertia_
p = plt.errorbar(n_init_range, inertia.mean(axis=1), inertia.std(axis=1))
plots.append(p[0])
legends.append("%s with %s init" % (factory.__name__, init))
5.6. Clustering 841

plt.xlabel('n_init')
plt.ylabel('inertia')
plt.legend(plots, legends)
plt.title("Mean inertia for various k-means init across %d runs" % n_runs)
# Part 2: Qualitative visual inspection of the convergence
X, y = make_data(random_state, n_samples_per_center, grid_size, scale)

km = MiniBatchKMeans(n_clusters=n_clusters, init='random', n_init=1,
random_state=random_state).fit(X)
plt.figure()
for k in range(n_clusters):
my_members = km.labels_ == k
color = cm.nipy_spectral(float(k) / n_clusters, 1)
plt.plot(X[my_members, 0], X[my_members, 1], 'o', marker='.', c=color)
cluster_center = km.cluster_centers_[k]
plt.plot(cluster_center[0], cluster_center[1], 'o',
markerfacecolor=color, markeredgecolor='k', markersize=6)
plt.title("Example cluster allocation with a single random init\n"
"with MiniBatchKMeans")
plt.show()
5.6.19 Adjustment for chance in clustering performance evaluation
The following plots demonstrate the impact of the number of clusters and number of samples on various clustering
performance evaluation metrics.
Non-adjusted measures such as the V-Measure show a dependency between the number of clusters and the number of
samples: the mean V-Measure of random labeling increases significantly as the number of clusters is closer to the total
number of samples used to compute the measure.
Adjusted for chance measure such as ARI display some random variations centered around a mean score of 0.0 for
any number of samples and clusters.
Only adjusted measures can hence safely be used as a consensus index to evaluate the average stability of clustering
algorithms for a given value of k on various overlapping sub-samples of the dataset.

5.6. Clustering 843

•
Out:
Computing adjusted_rand_score for 10 values of n_clusters and n_samples=100

done in 0.023s
Computing v_measure_score for 10 values of n_clusters and n_samples=100
done in 0.035s
Computing adjusted_mutual_info_score for 10 values of n_clusters and n_samples=100
done in 0.314s
Computing mutual_info_score for 10 values of n_clusters and n_samples=100
done in 0.029s
Computing adjusted_rand_score for 10 values of n_clusters and n_samples=1000
done in 0.038s
Computing v_measure_score for 10 values of n_clusters and n_samples=1000
done in 0.049s
Computing adjusted_mutual_info_score for 10 values of n_clusters and n_samples=1000
done in 0.197s
Computing mutual_info_score for 10 values of n_clusters and n_samples=1000
done in 0.038s
print(__doc__)

import numpy as np
def uniform_labelings_scores(score_func, n_samples, n_clusters_range,

fixed_n_classes=None, n_runs=5, seed=42):
"""Compute score for 2 random uniform cluster labelings.
Both random labelings have the same number of clusters for each value
possible value in ``n_clusters_range``.
When fixed_n_classes is not None the first labeling is considered a ground

truth class assignment with fixed number of classes.
"""
random_labels = np.random.RandomState(seed).randint
scores = np.zeros((len(n_clusters_range), n_runs))
if fixed_n_classes is not None:

labels_a = random_labels(low=0, high=fixed_n_classes, size=n_samples)
for i, k in enumerate(n_clusters_range):
for j in range(n_runs):
if fixed_n_classes is None:
labels_a = random_labels(low=0, high=k, size=n_samples)
labels_b = random_labels(low=0, high=k, size=n_samples)
scores[i, j] = score_func(labels_a, labels_b)
return scores
score_funcs = [
metrics.adjusted_rand_score,
metrics.v_measure_score,
metrics.adjusted_mutual_info_score,
metrics.mutual_info_score,
]
# 2 independent random clusterings with equal cluster number
n_samples = 100
n_clusters_range = np.linspace(2, n_samples, 10).astype(np.int)
plt.figure(1)
plots = []
names = []
for score_func in score_funcs:
print("Computing %s for %d values of n_clusters and n_samples=%d"
% (score_func.__name__, len(n_clusters_range), n_samples))
t0 = time()
scores = uniform_labelings_scores(score_func, n_samples, n_clusters_range)
plots.append(plt.errorbar(
n_clusters_range, np.median(scores, axis=1), scores.std(axis=1))[0])
names.append(score_func.__name__)
5.6. Clustering 845

plt.title("Clustering measures for 2 random uniform labelings\n"

"with equal number of clusters")
plt.xlabel('Number of clusters (Number of samples is fixed to %d)' % n_samples)
plt.ylabel('Score value')
plt.legend(plots, names)
plt.ylim(ymin=-0.05, ymax=1.05)
# Random labeling with varying n_clusters against ground class labels

# with fixed number of clusters
n_samples = 1000
n_clusters_range = np.linspace(2, 100, 10).astype(np.int)
n_classes = 10
plt.figure(2)
plots = []
names = []
for score_func in score_funcs:
print("Computing %s for %d values of n_clusters and n_samples=%d"
% (score_func.__name__, len(n_clusters_range), n_samples))
t0 = time()
scores = uniform_labelings_scores(score_func, n_samples, n_clusters_range,
fixed_n_classes=n_classes)
plots.append(plt.errorbar(
n_clusters_range, scores.mean(axis=1), scores.std(axis=1))[0])
names.append(score_func.__name__)
plt.title("Clustering measures for random uniform labeling\n"

"against reference assignment with %d classes" % n_classes)
plt.xlabel('Number of clusters (Number of samples is fixed to %d)' % n_samples)
plt.ylabel('Score value')
plt.ylim(ymin=-0.05, ymax=1.05)
plt.legend(plots, names)
plt.show()
5.6.20 Feature agglomeration vs. univariate selection
This example compares 2 dimensionality reduction strategies:

• univariate feature selection with Anova
• feature agglomeration with Ward hierarchical clustering
Both methods are compared in a regression problem using a BayesianRidge as supervised estimator.

Out:
________________________________________________________________________________
[Memory] Calling sklearn.cluster.hierarchical.ward_tree...
ward_tree(array([[-0.451933, ..., -0.675318],
...,
[ 0.275706, ..., -1.085711]]),
<1600x1600 sparse matrix of type '<class 'numpy.int64'>'
with 7840 stored elements in COOrdinate format>, n_clusters=None)
________________________________________________________ward_tree - 0.1s, 0.0min
________________________________________________________________________________
ward_tree(array([[ 0.905206, ..., 0.161245],
...,
[-0.849835, ..., -1.091621]]),
________________________________________________________ward_tree - 0.6s, 0.0min
________________________________________________________________________________
ward_tree(array([[ 0.905206, ..., -0.675318],
...,
[-0.849835, ..., -1.085711]]),
________________________________________________________ward_tree - 0.1s, 0.0min
________________________________________________________________________________
[Memory] Calling sklearn.feature_selection.univariate_selection.f_regression...
f_regression(array([[-0.451933, ..., 0.275706],
...,
[-0.675318, ..., -1.085711]]),
array([ 25.267703, ..., -25.026711]))
_____________________________________________________f_regression - 0.0s, 0.0min
________________________________________________________________________________
f_regression(array([[ 0.905206, ..., -0.849835],
...,
[ 0.161245, ..., -1.091621]]),
array([ -27.447268, ..., -112.638768]))
________________________________________________________________________________
5.6. Clustering 847

f_regression(array([[ 0.905206, ..., -0.849835],

...,
[-0.675318, ..., -1.085711]]),
array([-27.447268, ..., -25.026711]))

print(__doc__)
import shutil
import tempfile
import numpy as np
from scipy import linalg, ndimage

from sklearn import feature_selection
from sklearn.cluster import FeatureAgglomeration
from sklearn.linear_model import BayesianRidge
# #############################################################################
# Generate data
n_samples = 200
size = 40 # image size
roi_size = 15
snr = 5.
np.random.seed(0)
mask = np.ones([size, size], dtype=np.bool)
coef = np.zeros((size, size))

coef[0:roi_size, 0:roi_size] = -1.
coef[-roi_size:, -roi_size:] = 1.
X = np.random.randn(n_samples, size ** 2)
for x in X: # smooth data
x[:] = ndimage.gaussian_filter(x.reshape(size, size), sigma=1.0).ravel()
X -= X.mean(axis=0)
X /= X.std(axis=0)
y = np.dot(X, coef.ravel())
noise = np.random.randn(y.shape[0])
noise_coef = (linalg.norm(y, 2) / np.exp(snr / 20.)) / linalg.norm(noise, 2)
y += noise_coef * noise # add noise
# #############################################################################
# Compute the coefs of a Bayesian Ridge with GridSearch

cv = KFold(2) # cross-validation generator for model selection

ridge = BayesianRidge()
cachedir = tempfile.mkdtemp()
mem = Memory(cachedir=cachedir, verbose=1)
# Ward agglomeration followed by BayesianRidge

connectivity = grid_to_graph(n_x=size, n_y=size)
ward = FeatureAgglomeration(n_clusters=10, connectivity=connectivity,
memory=mem)
clf = Pipeline([('ward', ward), ('ridge', ridge)])
# Select the optimal number of parcels with grid search
clf = GridSearchCV(clf, {'ward__n_clusters': [10, 20, 30]}, n_jobs=1, cv=cv)
clf.fit(X, y) # set the best parameters
coef_ = clf.best_estimator_.steps[-1][1].coef_
coef_ = clf.best_estimator_.steps[0][1].inverse_transform(coef_)
coef_agglomeration_ = coef_.reshape(size, size)
# Anova univariate feature selection followed by BayesianRidge

f_regression = mem.cache(feature_selection.f_regression) # caching function
anova = feature_selection.SelectPercentile(f_regression)
clf = Pipeline([('anova', anova), ('ridge', ridge)])
# Select the optimal percentage of features with grid search
clf = GridSearchCV(clf, {'anova__percentile': [5, 10, 20]}, cv=cv)
clf.fit(X, y) # set the best parameters
coef_ = clf.best_estimator_.steps[-1][1].coef_
coef_ = clf.best_estimator_.steps[0][1].inverse_transform(coef_.reshape(1, -1))
coef_selection_ = coef_.reshape(size, size)
# #############################################################################
# Inverse the transformation to plot the results on an image
plt.close('all')
plt.figure(figsize=(7.3, 2.7))
plt.imshow(coef, interpolation="nearest", cmap=plt.cm.RdBu_r)
plt.title("True weights")
plt.imshow(coef_selection_, interpolation="nearest", cmap=plt.cm.RdBu_r)
plt.title("Feature Selection")
plt.imshow(coef_agglomeration_, interpolation="nearest", cmap=plt.cm.RdBu_r)
plt.title("Feature Agglomeration")
plt.show()
# Attempt to remove the temporary cachedir, but don't worry if it fails

shutil.rmtree(cachedir, ignore_errors=True)
5.6.21 A demo of K-Means clustering on the handwritten digits data
In this example we compare the various initialization strategies for K-means in terms of runtime and quality of the
results.
5.6. Clustering 849

As the ground truth is known here, we also apply different cluster quality metrics to judge the goodness of fit of the
cluster labels to the ground truth.
Cluster quality metrics evaluated (see Clustering performance evaluation for definitions and discussions of the met-
rics):
Shorthand full name

homo homogeneity score
compl completeness score
v-meas V measure
ARI adjusted Rand index
AMI adjusted mutual information
silhouette silhouette coefficient
Out:
n_digits: 10, n_samples 1797, n_features 64

__________________________________________________________________________________
init time inertia homo compl v-meas ARI AMI silhouette
k-means++ 0.48s 69432 0.602 0.650 0.625 0.465 0.598 0.146
random 0.28s 69694 0.669 0.710 0.689 0.553 0.666 0.147
PCA-based 0.03s 70804 0.671 0.698 0.684 0.561 0.668 0.118
__________________________________________________________________________________

print(__doc__)

import numpy as np

from sklearn.preprocessing import scale
np.random.seed(42)
digits = load_digits()
data = scale(digits.data)
n_samples, n_features = data.shape

n_digits = len(np.unique(digits.target))
labels = digits.target
sample_size = 300
print("n_digits: %d, \t n_samples %d, \t n_features %d"

% (n_digits, n_samples, n_features))
print(82 * '_')
print('init\t\ttime\tinertia\thomo\tcompl\tv-meas\tARI\tAMI\tsilhouette')
def bench_k_means(estimator, name, data):

t0 = time()
estimator.fit(data)
print('%-9s\t%.2fs\t%i\t%.3f\t%.3f\t%.3f\t%.3f\t%.3f\t%.3f'
% (name, (time() - t0), estimator.inertia_,
metrics.homogeneity_score(labels, estimator.labels_),
metrics.completeness_score(labels, estimator.labels_),
metrics.v_measure_score(labels, estimator.labels_),
metrics.adjusted_rand_score(labels, estimator.labels_),
metrics.adjusted_mutual_info_score(labels, estimator.labels_),
metrics.silhouette_score(data, estimator.labels_,
metric='euclidean',
sample_size=sample_size)))
bench_k_means(KMeans(init='k-means++', n_clusters=n_digits, n_init=10),

name="k-means++", data=data)
bench_k_means(KMeans(init='random', n_clusters=n_digits, n_init=10),

name="random", data=data)
# in this case the seeding of the centers is deterministic, hence we run the
# kmeans algorithm only once with n_init=1
pca = PCA(n_components=n_digits).fit(data)
bench_k_means(KMeans(init=pca.components_, n_clusters=n_digits, n_init=1),
5.6. Clustering 851

name="PCA-based",
data=data)
print(82 * '_')
# #############################################################################
# Visualize the results on PCA-reduced data
reduced_data = PCA(n_components=2).fit_transform(data)
kmeans = KMeans(init='k-means++', n_clusters=n_digits, n_init=10)
kmeans.fit(reduced_data)
# Step size of the mesh. Decrease to increase the quality of the VQ.
h = .02 # point in the mesh [x_min, x_max]x[y_min, y_max].
x_min, x_max = reduced_data[:, 0].min() - 1, reduced_data[:, 0].max() + 1
y_min, y_max = reduced_data[:, 1].min() - 1, reduced_data[:, 1].max() + 1
xx, yy = np.meshgrid(np.arange(x_min, x_max, h), np.arange(y_min, y_max, h))
# Obtain labels for each point in mesh. Use last trained model.
Z = kmeans.predict(np.c_[xx.ravel(), yy.ravel()])

plt.figure(1)
plt.clf()
plt.imshow(Z, interpolation='nearest',
extent=(xx.min(), xx.max(), yy.min(), yy.max()),
cmap=plt.cm.Paired,
aspect='auto', origin='lower')
plt.plot(reduced_data[:, 0], reduced_data[:, 1], 'k.', markersize=2)

# Plot the centroids as a white X
centroids = kmeans.cluster_centers_
plt.scatter(centroids[:, 0], centroids[:, 1],
marker='x', s=169, linewidths=3,
color='w', zorder=10)
plt.title('K-means clustering on the digits dataset (PCA-reduced data)\n'
'Centroids are marked with white cross')
plt.xlim(x_min, x_max)
plt.ylim(y_min, y_max)
plt.xticks(())
plt.yticks(())
plt.show()
5.6.22 Comparison of the K-Means and MiniBatchKMeans clustering algorithms
We want to compare the performance of the MiniBatchKMeans and KMeans: the MiniBatchKMeans is faster, but
gives slightly different results (see Mini Batch K-Means).
We will cluster a set of data, first with KMeans and then with MiniBatchKMeans, and plot the results. We will also

plot the points that are labelled differently between the two algorithms.
print(__doc__)
import time
import numpy as np
from sklearn.cluster import MiniBatchKMeans, KMeans

from sklearn.metrics.pairwise import pairwise_distances_argmin
# #############################################################################
np.random.seed(0)
batch_size = 45
centers = [[1, 1], [-1, -1], [1, -1]]
n_clusters = len(centers)
X, labels_true = make_blobs(n_samples=3000, centers=centers, cluster_std=0.7)
# #############################################################################
# Compute clustering with Means
k_means = KMeans(init='k-means++', n_clusters=3, n_init=10)

t0 = time.time()
k_means.fit(X)
t_batch = time.time() - t0
# #############################################################################
# Compute clustering with MiniBatchKMeans
mbk = MiniBatchKMeans(init='k-means++', n_clusters=3, batch_size=batch_size,

n_init=10, max_no_improvement=10, verbose=0)
t0 = time.time()
mbk.fit(X)
t_mini_batch = time.time() - t0
# #############################################################################
# Plot result
5.6. Clustering 853


fig.subplots_adjust(left=0.02, right=0.98, bottom=0.05, top=0.9)
colors = ['#4EACC5', '#FF9C34', '#4E9A06']
# We want to have the same colors for the same cluster from the
# MiniBatchKMeans and the KMeans algorithm. Let's pair the cluster centers per
# closest one.
k_means_cluster_centers = np.sort(k_means.cluster_centers_, axis=0)
mbk_means_cluster_centers = np.sort(mbk.cluster_centers_, axis=0)
k_means_labels = pairwise_distances_argmin(X, k_means_cluster_centers)
mbk_means_labels = pairwise_distances_argmin(X, mbk_means_cluster_centers)
order = pairwise_distances_argmin(k_means_cluster_centers,
mbk_means_cluster_centers)
# KMeans
for k, col in zip(range(n_clusters), colors):
my_members = k_means_labels == k
cluster_center = k_means_cluster_centers[k]
ax.plot(X[my_members, 0], X[my_members, 1], 'w',
markerfacecolor=col, marker='.')
ax.plot(cluster_center[0], cluster_center[1], 'o', markerfacecolor=col,
ax.set_title('KMeans')
ax.set_xticks(())
ax.set_yticks(())
plt.text(-3.5, 1.8, 'train time: %.2fs\ninertia: %f' % (
t_batch, k_means.inertia_))
# MiniBatchKMeans
for k, col in zip(range(n_clusters), colors):
my_members = mbk_means_labels == order[k]
cluster_center = mbk_means_cluster_centers[order[k]]
ax.plot(X[my_members, 0], X[my_members, 1], 'w',
markerfacecolor=col, marker='.')
ax.plot(cluster_center[0], cluster_center[1], 'o', markerfacecolor=col,
ax.set_title('MiniBatchKMeans')
ax.set_xticks(())
ax.set_yticks(())
plt.text(-3.5, 1.8, 'train time: %.2fs\ninertia: %f' %
(t_mini_batch, mbk.inertia_))
# Initialise the different array to all False

different = (mbk_means_labels == 4)
for k in range(n_clusters):
different += ((k_means_labels == k) != (mbk_means_labels == order[k]))
identic = np.logical_not(different)
ax.plot(X[identic, 0], X[identic, 1], 'w',
markerfacecolor='#bbbbbb', marker='.')
ax.plot(X[different, 0], X[different, 1], 'w',
markerfacecolor='m', marker='.')
ax.set_title('Difference')
ax.set_xticks(())

ax.set_yticks(())
plt.show()
5.6.23 Comparing different hierarchical linkage methods on toy datasets
This example shows characteristics of different linkage methods for hierarchical clustering on datasets that are “inter-
esting” but still in 2D.
The main observations to make are:
• single linkage is fast, and can perform well on non-globular data, but it performs poorly in the presence of noise.
• average and complete linkage perform well on cleanly separated globular clusters, but have mixed results other-
wise.
• Ward is the most effective method for noisy data.
data.
print(__doc__)
import time
import warnings
import numpy as np
from sklearn import cluster, datasets

from itertools import cycle, islice
np.random.seed(0)
Generate datasets. We choose the size big enough to see the scalability of the algorithms, but not too big to avoid too
long running times
n_samples = 1500
noisy_circles = datasets.make_circles(n_samples=n_samples, factor=.5,
noise=.05)
noisy_moons = datasets.make_moons(n_samples=n_samples, noise=.05)
blobs = datasets.make_blobs(n_samples=n_samples, random_state=8)
no_structure = np.random.rand(n_samples, 2), None

random_state = 170
X, y = datasets.make_blobs(n_samples=n_samples, random_state=random_state)
transformation = [[0.6, -0.6], [-0.4, 0.8]]
aniso = (X_aniso, y)
5.6. Clustering 855

# blobs with varied variances

varied = datasets.make_blobs(n_samples=n_samples,
cluster_std=[1.0, 2.5, 0.5],
Run the clustering and plot
# Set up cluster parameters

plt.figure(figsize=(9 * 1.3 + 2, 14.5))
hspace=.01)
plot_num = 1
default_base = {'n_neighbors': 10,

'n_clusters': 3}
datasets = [
(noisy_circles, {'n_clusters': 2}),
(noisy_moons, {'n_clusters': 2}),
(varied, {'n_neighbors': 2}),
(aniso, {'n_neighbors': 2}),
(blobs, {}),
(no_structure, {})]
for i_dataset, (dataset, algo_params) in enumerate(datasets):

# update parameters with dataset-specific values
params = default_base.copy()
params.update(algo_params)
X, y = dataset
# normalize dataset for easier parameter selection

# ============
# Create cluster objects
# ============
ward = cluster.AgglomerativeClustering(
n_clusters=params['n_clusters'], linkage='ward')
complete = cluster.AgglomerativeClustering(
n_clusters=params['n_clusters'], linkage='complete')
average = cluster.AgglomerativeClustering(
n_clusters=params['n_clusters'], linkage='average')
single = cluster.AgglomerativeClustering(
n_clusters=params['n_clusters'], linkage='single')
clustering_algorithms = (
('Single Linkage', single),
('Average Linkage', average),
('Complete Linkage', complete),
('Ward Linkage', ward),
)
for name, algorithm in clustering_algorithms:

t0 = time.time()
# catch warnings related to kneighbors_graph

with warnings.catch_warnings():
warnings.filterwarnings(
"ignore",
message="the number of connected components of the " +
"connectivity matrix is [0-9]{1,2}" +
" > 1. Completing it to avoid stopping the tree early.",
category=UserWarning)
algorithm.fit(X)
t1 = time.time()
if hasattr(algorithm, 'labels_'):
y_pred = algorithm.labels_.astype(np.int)
else:
y_pred = algorithm.predict(X)
plt.subplot(len(datasets), len(clustering_algorithms), plot_num)

if i_dataset == 0:
colors = np.array(list(islice(cycle(['#377eb8', '#ff7f00', '#4daf4a',

'#f781bf', '#a65628', '#984ea3',
'#999999', '#e41a1c', '#dede00']),
int(max(y_pred) + 1))))
plt.scatter(X[:, 0], X[:, 1], s=10, color=colors[y_pred])
plt.xlim(-2.5, 2.5)
plt.ylim(-2.5, 2.5)
plt.xticks(())
plt.yticks(())
plot_num += 1
plt.show()
5.6. Clustering 857

5.6.24 Selecting the number of clusters with silhouette analysis on KMeans clus-
tering
Silhouette analysis can be used to study the separation distance between the resulting clusters. The silhouette plot
displays a measure of how close each point in one cluster is to points in the neighboring clusters and thus provides a

way to assess parameters like number of clusters visually. This measure has a range of [-1, 1].
Silhouette coefficients (as these values are referred to as) near +1 indicate that the sample is far away from the neigh-
boring clusters. A value of 0 indicates that the sample is on or very close to the decision boundary between two
neighboring clusters and negative values indicate that those samples might have been assigned to the wrong cluster.
In this example the silhouette analysis is used to choose an optimal value for n_clusters. The silhouette plot shows
that the n_clusters value of 3, 5 and 6 are a bad pick for the given data due to the presence of clusters with below
average silhouette scores and also due to wide fluctuations in the size of the silhouette plots. Silhouette analysis is
more ambivalent in deciding between 2 and 4.
Also from the thickness of the silhouette plot the cluster size can be visualized. The silhouette plot for cluster 0 when
n_clusters is equal to 2, is bigger in size owing to the grouping of the 3 sub clusters into one big cluster. However
when the n_clusters is equal to 4, all the plots are more or less of similar thickness and hence are of similar sizes
as can be also verified from the labelled scatter plot on the right.
5.6. Clustering 859

•
Out:
For n_clusters = 2 The average silhouette_score is : 0.7049787496083262



from sklearn.metrics import silhouette_samples, silhouette_score

import matplotlib.cm as cm
import numpy as np
print(__doc__)
# Generating the sample data from make_blobs

# This particular setting has one distinct cluster and 3 clusters placed close
# together.
X, y = make_blobs(n_samples=500,
n_features=2,
centers=4,
cluster_std=1,
center_box=(-10.0, 10.0),
shuffle=True,
random_state=1) # For reproducibility
range_n_clusters = [2, 3, 4, 5, 6]
for n_clusters in range_n_clusters:

# Create a subplot with 1 row and 2 columns
fig, (ax1, ax2) = plt.subplots(1, 2)
fig.set_size_inches(18, 7)
# The 1st subplot is the silhouette plot

# The silhouette coefficient can range from -1, 1 but in this example all
# lie within [-0.1, 1]
ax1.set_xlim([-0.1, 1])
# The (n_clusters+1)*10 is for inserting blank space between silhouette
# plots of individual clusters, to demarcate them clearly.
ax1.set_ylim([0, len(X) + (n_clusters + 1) * 10])
# Initialize the clusterer with n_clusters value and a random generator

# seed of 10 for reproducibility.
clusterer = KMeans(n_clusters=n_clusters, random_state=10)
cluster_labels = clusterer.fit_predict(X)
# The silhouette_score gives the average value for all the samples.
# This gives a perspective into the density and separation of the formed
# clusters
silhouette_avg = silhouette_score(X, cluster_labels)
print("For n_clusters =", n_clusters,
"The average silhouette_score is :", silhouette_avg)
# Compute the silhouette scores for each sample

sample_silhouette_values = silhouette_samples(X, cluster_labels)
y_lower = 10
5.6. Clustering 861

# Aggregate the silhouette scores for samples belonging to

# cluster i, and sort them
ith_cluster_silhouette_values = \
sample_silhouette_values[cluster_labels == i]
ith_cluster_silhouette_values.sort()
size_cluster_i = ith_cluster_silhouette_values.shape[0]
y_upper = y_lower + size_cluster_i
color = cm.nipy_spectral(float(i) / n_clusters)

ax1.fill_betweenx(np.arange(y_lower, y_upper),
0, ith_cluster_silhouette_values,
facecolor=color, edgecolor=color, alpha=0.7)
# Label the silhouette plots with their cluster numbers at the middle
ax1.text(-0.05, y_lower + 0.5 * size_cluster_i, str(i))
# Compute the new y_lower for next plot

y_lower = y_upper + 10 # 10 for the 0 samples
ax1.set_title("The silhouette plot for the various clusters.")

ax1.set_xlabel("The silhouette coefficient values")
ax1.set_ylabel("Cluster label")
# The vertical line for average silhouette score of all the values
ax1.axvline(x=silhouette_avg, color="red", linestyle="--")
ax1.set_yticks([]) # Clear the yaxis labels / ticks

ax1.set_xticks([-0.1, 0, 0.2, 0.4, 0.6, 0.8, 1])
# 2nd Plot showing the actual clusters formed

colors = cm.nipy_spectral(cluster_labels.astype(float) / n_clusters)
ax2.scatter(X[:, 0], X[:, 1], marker='.', s=30, lw=0, alpha=0.7,
c=colors, edgecolor='k')
# Labeling the clusters

centers = clusterer.cluster_centers_
# Draw white circles at cluster centers
ax2.scatter(centers[:, 0], centers[:, 1], marker='o',
c="white", alpha=1, s=200, edgecolor='k')
for i, c in enumerate(centers):
ax2.scatter(c[0], c[1], marker='$%d$' % i, alpha=1,
ax2.set_title("The visualization of the clustered data.")

ax2.set_xlabel("Feature space for the 1st feature")
ax2.set_ylabel("Feature space for the 2nd feature")
plt.suptitle(("Silhouette analysis for KMeans clustering on sample data "

"with n_clusters = %d" % n_clusters),
fontsize=14, fontweight='bold')
plt.show()

5.6.25 Comparing different clustering algorithms on toy datasets
This example shows characteristics of different clustering algorithms on datasets that are “interesting” but still in 2D.
With the exception of the last dataset, the parameters of each of these dataset-algorithm pairs has been tuned to produce
good clustering results. Some algorithms are more sensitive to parameter values than others.
The last dataset is an example of a ‘null’ situation for clustering: the data is homogeneous, and there is no good
clustering. For this example, the null dataset uses the same parameters as the dataset in the row above it, which
represents a mismatch in the parameter values and the data structure.
data.
print(__doc__)
import time
import warnings
import numpy as np
from sklearn import cluster, datasets, mixture

from itertools import cycle, islice
np.random.seed(0)
5.6. Clustering 863

# ============
# Generate datasets. We choose the size big enough to see the scalability
# of the algorithms, but not too big to avoid too long running times
# ============
n_samples = 1500
noisy_circles = datasets.make_circles(n_samples=n_samples, factor=.5,
noise=.05)
noisy_moons = datasets.make_moons(n_samples=n_samples, noise=.05)
blobs = datasets.make_blobs(n_samples=n_samples, random_state=8)
no_structure = np.random.rand(n_samples, 2), None

random_state = 170
X, y = datasets.make_blobs(n_samples=n_samples, random_state=random_state)
transformation = [[0.6, -0.6], [-0.4, 0.8]]
aniso = (X_aniso, y)
# blobs with varied variances

varied = datasets.make_blobs(n_samples=n_samples,
cluster_std=[1.0, 2.5, 0.5],
# ============
# Set up cluster parameters
# ============
plt.figure(figsize=(9 * 2 + 3, 12.5))
hspace=.01)
plot_num = 1
default_base = {'quantile': .3,

'eps': .3,
'damping': .9,
'preference': -200,
'n_neighbors': 10,
'n_clusters': 3}
datasets = [
(noisy_circles, {'damping': .77, 'preference': -240,
'quantile': .2, 'n_clusters': 2}),
(noisy_moons, {'damping': .75, 'preference': -220, 'n_clusters': 2}),
(varied, {'eps': .18, 'n_neighbors': 2}),
(aniso, {'eps': .15, 'n_neighbors': 2}),
(blobs, {}),
(no_structure, {})]
for i_dataset, (dataset, algo_params) in enumerate(datasets):

# update parameters with dataset-specific values
params = default_base.copy()
params.update(algo_params)
X, y = dataset
# normalize dataset for easier parameter selection


# estimate bandwidth for mean shift

bandwidth = cluster.estimate_bandwidth(X, quantile=params['quantile'])
# connectivity matrix for structured Ward

connectivity = kneighbors_graph(
X, n_neighbors=params['n_neighbors'], include_self=False)
# make connectivity symmetric
connectivity = 0.5 * (connectivity + connectivity.T)
# ============
# Create cluster objects
# ============
ms = cluster.MeanShift(bandwidth=bandwidth, bin_seeding=True)
two_means = cluster.MiniBatchKMeans(n_clusters=params['n_clusters'])
ward = cluster.AgglomerativeClustering(
n_clusters=params['n_clusters'], linkage='ward',
connectivity=connectivity)
spectral = cluster.SpectralClustering(
n_clusters=params['n_clusters'], eigen_solver='arpack',
affinity="nearest_neighbors")
dbscan = cluster.DBSCAN(eps=params['eps'])
affinity_propagation = cluster.AffinityPropagation(
damping=params['damping'], preference=params['preference'])
average_linkage = cluster.AgglomerativeClustering(
linkage="average", affinity="cityblock",
n_clusters=params['n_clusters'], connectivity=connectivity)
birch = cluster.Birch(n_clusters=params['n_clusters'])
gmm = mixture.GaussianMixture(
n_components=params['n_clusters'], covariance_type='full')
clustering_algorithms = (
('MiniBatchKMeans', two_means),
('AffinityPropagation', affinity_propagation),
('MeanShift', ms),
('SpectralClustering', spectral),
('Ward', ward),
('AgglomerativeClustering', average_linkage),
('DBSCAN', dbscan),
('Birch', birch),
('GaussianMixture', gmm)
)
for name, algorithm in clustering_algorithms:

t0 = time.time()
# catch warnings related to kneighbors_graph

with warnings.catch_warnings():
"ignore",
message="the number of connected components of the " +
"connectivity matrix is [0-9]{1,2}" +
" > 1. Completing it to avoid stopping the tree early.",
"ignore",
message="Graph is not fully connected, spectral embedding" +
" may not work as expected.",
5.6. Clustering 865

algorithm.fit(X)
t1 = time.time()
if hasattr(algorithm, 'labels_'):
y_pred = algorithm.labels_.astype(np.int)
else:
y_pred = algorithm.predict(X)
plt.subplot(len(datasets), len(clustering_algorithms), plot_num)

if i_dataset == 0:
colors = np.array(list(islice(cycle(['#377eb8', '#ff7f00', '#4daf4a',

'#f781bf', '#a65628', '#984ea3',
'#999999', '#e41a1c', '#dede00']),
int(max(y_pred) + 1))))
# add black color for outliers (if any)
colors = np.append(colors, ["#000000"])
plt.scatter(X[:, 0], X[:, 1], s=10, color=colors[y_pred])
plt.xlim(-2.5, 2.5)
plt.ylim(-2.5, 2.5)
plt.xticks(())
plt.yticks(())
plot_num += 1
plt.show()
5.7 Pipelines and composite estimators
Examples of how to compose transformers and pipelines from other estimators. See the User Guide.
5.7.1 Concatenating multiple feature extraction methods
In many real-world examples, there are many ways to extract features from a dataset. Often it is beneficial to combine
several methods to obtain good performance. This example shows how to use FeatureUnion to combine features
obtained by PCA and univariate selection.
Combining features using this transformer has the benefit that it allows cross validation and grid searches over the
whole process.
The combination used in this example is not particularly helpful on this dataset and is only used to illustrate the usage
of FeatureUnion.
Out:

Combined space has 3 features

Fitting 5 folds for each of 18 candidates, totalling 90 fits
[CV] features__pca__n_components=1, features__univ_select__k=1, svm__C=0.1
[CV] features__pca__n_components=1, features__univ_select__k=1, svm__C=0.1, score=0.
˓→9333333333333333, total= 0.0s
˓→9333333333333333, total= 0.0s
˓→8666666666666667, total= 0.0s
˓→9333333333333333, total= 0.0s
˓→0, total= 0.0s
[CV] features__pca__n_components=1, features__univ_select__k=1, svm__C=1
[CV] features__pca__n_components=1, features__univ_select__k=1, svm__C=1, score=0.9,
˓→total= 0.0s
˓→total= 0.0s
[CV] features__pca__n_components=1, features__univ_select__k=1, svm__C=1, score=0.
˓→8666666666666667, total= 0.0s
˓→9333333333333333, total= 0.0s
˓→total= 0.0s
˓→9333333333333333, total= 0.0s
˓→ total= 0.0s
˓→ total= 0.0s
˓→9333333333333333, total= 0.0s
˓→ total= 0.0s
˓→9333333333333333, total= 0.0s
˓→9666666666666667, total= 0.0s
˓→9333333333333333, total= 0.0s
˓→9333333333333333, total= 0.0s
5.7. Pipelines and composite estimators 867


˓→0, total= 0.0s
˓→9333333333333333, total= 0.0s
˓→9666666666666667, total= 0.0s
˓→9333333333333333, total= 0.0s
˓→9333333333333333, total= 0.0s
˓→total= 0.0s
˓→9666666666666667, total= 0.0s
˓→9666666666666667, total= 0.1s
˓→9333333333333333, total= 0.0s
˓→9333333333333333, total= 0.0s
˓→ total= 0.0s
˓→9333333333333333, total= 0.0s
˓→0, total= 0.0s
˓→8666666666666667, total= 0.0s
˓→9333333333333333, total= 0.0s
˓→0, total= 0.0s
˓→9666666666666667, total= 0.0s
˓→total= 0.0s
˓→9333333333333333, total= 0.1s


˓→9333333333333333, total= 0.0s
˓→total= 0.0s
˓→9666666666666667, total= 0.0s
˓→9666666666666667, total= 0.0s
˓→ total= 0.0s
˓→9333333333333333, total= 0.0s
˓→ total= 0.0s
˓→9666666666666667, total= 0.0s
˓→0, total= 0.0s
˓→9333333333333333, total= 0.0s
˓→9333333333333333, total= 0.0s
˓→0, total= 0.1s
˓→9666666666666667, total= 0.0s
˓→total= 0.0s
˓→9333333333333333, total= 0.0s
˓→9666666666666667, total= 0.0s
˓→total= 0.0s
˓→9666666666666667, total= 0.0s
˓→ total= 0.0s
˓→ total= 0.0s


˓→9333333333333333, total= 0.0s
˓→ total= 0.0s
˓→9666666666666667, total= 0.0s
˓→0, total= 0.1s
˓→9333333333333333, total= 0.0s
˓→9666666666666667, total= 0.0s
˓→0, total= 0.0s
˓→9666666666666667, total= 0.0s
˓→total= 0.0s
˓→9333333333333333, total= 0.0s
˓→9666666666666667, total= 0.0s
˓→total= 0.0s
˓→ total= 0.0s
˓→ total= 0.0s
˓→9333333333333333, total= 0.0s
˓→9666666666666667, total= 0.0s
˓→ total= 0.0s
˓→9666666666666667, total= 0.0s
˓→0, total= 0.0s


˓→9333333333333333, total= 0.0s
˓→9666666666666667, total= 0.0s
˓→0, total= 0.0s
˓→9666666666666667, total= 0.0s
˓→total= 0.0s
˓→9666666666666667, total= 0.0s
˓→9666666666666667, total= 0.0s
˓→total= 0.0s
˓→ total= 0.0s
˓→ total= 0.0s
˓→ total= 0.0s
˓→9666666666666667, total= 0.0s
˓→ total= 0.0s
steps=[('features', FeatureUnion(n_jobs=None,
transformer_list=[('pca', PCA(copy=True, iterated_power='auto', n_components=3,
˓→ random_state=None,
svd_solver='auto', tol=0.0, whiten=False)), ('univ_select', SelectKBest(k=1, score_

˓→func=<function f_classif at 0x7ff1f89b1510>))],
transfor...r', max_iter=-1, probability=False, random_state=None,

shrinking=True, tol=0.001, verbose=False))])
# Author: Andreas Mueller <amueller@ais.uni-bonn.de>

#

from sklearn.pipeline import Pipeline, FeatureUnion


from sklearn.datasets import load_iris
from sklearn.feature_selection import SelectKBest
iris = load_iris()
X, y = iris.data, iris.target
# This dataset is way too high-dimensional. Better do PCA:

pca = PCA(n_components=2)
# Maybe some original features where good, too?

selection = SelectKBest(k=1)
# Build estimator from PCA and Univariate selection:
combined_features = FeatureUnion([("pca", pca), ("univ_select", selection)])
# Use combined features to transform dataset:

X_features = combined_features.fit(X, y).transform(X)
print("Combined space has", X_features.shape[1], "features")
svm = SVC(kernel="linear")
# Do grid search over k, n_components and C:
pipeline = Pipeline([("features", combined_features), ("svm", svm)])
param_grid = dict(features__pca__n_components=[1, 2, 3],

features__univ_select__k=[1, 2],
svm__C=[0.1, 1, 10])
grid_search = GridSearchCV(pipeline, param_grid=param_grid, cv=5, verbose=10)

grid_search.fit(X, y)
print(grid_search.best_estimator_)
5.7.2 Pipelining: chaining a PCA and a logistic regression
The PCA does an unsupervised dimensionality reduction, while the logistic regression does the prediction.
We use a GridSearchCV to set the dimensionality of the PCA

Out:
Best parameter (CV score=0.920):

{'logistic__alpha': 0.01, 'pca__n_components': 64}
print(__doc__)


import numpy as np
import pandas as pd

# Define a pipeline to search for the best combination of PCA truncation

# and classifier regularization.
logistic = SGDClassifier(loss='log', penalty='l2', early_stopping=True,
max_iter=10000, tol=1e-5, random_state=0)
pca = PCA()
pipe = Pipeline(steps=[('pca', pca), ('logistic', logistic)])
X_digits = digits.data
# Parameters of pipelines can be set using ‘__’ separated parameter names:

param_grid = {
'pca__n_components': [5, 20, 30, 40, 50, 64],
'logistic__alpha': np.logspace(-4, 4, 5),
}
search = GridSearchCV(pipe, param_grid, iid=False, cv=5,
return_train_score=False)
search.fit(X_digits, y_digits)
print("Best parameter (CV score=%0.3f):" % search.best_score_)
print(search.best_params_)
# Plot the PCA spectrum

pca.fit(X_digits)
fig, (ax0, ax1) = plt.subplots(nrows=2, sharex=True, figsize=(6, 6))

ax0.plot(pca.explained_variance_ratio_, linewidth=2)
ax0.set_ylabel('PCA explained variance')
ax0.axvline(search.best_estimator_.named_steps['pca'].n_components,
linestyle=':', label='n_components chosen')
ax0.legend(prop=dict(size=12))
# For each number of components, find the best classifier results

results = pd.DataFrame(search.cv_results_)
components_col = 'param_pca__n_components'
best_clfs = results.groupby(components_col).apply(
lambda g: g.nlargest(1, 'mean_test_score'))
best_clfs.plot(x=components_col, y='mean_test_score', yerr='std_test_score',

legend=False, ax=ax1)
ax1.set_ylabel('Classification accuracy (val)')
ax1.set_xlabel('n_components')
plt.tight_layout()
plt.show()

5.7.3 Column Transformer with Mixed Types
This example illustrates how to apply different preprocessing and feature extraction pipelines to different subsets of
features, using sklearn.compose.ColumnTransformer. This is particularly handy for the case of datasets
that contain heterogeneous data types, since we may want to scale the numeric features and one-hot encode the cate-
gorical ones.
In this example, the numeric data is standard-scaled after mean-imputation, while the categorical data is one-hot
encoded after imputing missing values with a new category ('missing').
Finally, the preprocessing pipeline is integrated in a full prediction pipeline using sklearn.pipeline.
Pipeline, together with a simple classification model.
# Author: Pedro Morales <part.morales@gmail.com>
#
import pandas as pd
import numpy as np
from sklearn.compose import ColumnTransformer

from sklearn.impute import SimpleImputer
from sklearn.preprocessing import StandardScaler, OneHotEncoder
from sklearn.model_selection import train_test_split, GridSearchCV
np.random.seed(0)
# Read data from Titanic dataset.

titanic_url = ('https://raw.githubusercontent.com/amueller/'
'scipy-2017-sklearn/091d371/notebooks/datasets/titanic3.csv')
data = pd.read_csv(titanic_url)
# We will train our classifier with the following features:

# Numeric Features:
# - age: float.
# - fare: float.
# Categorical Features:
# - embarked: categories encoded as strings {'C', 'S', 'Q'}.
# - sex: categories encoded as strings {'female', 'male'}.
# - pclass: ordinal integers {1, 2, 3}.
# We create the preprocessing pipelines for both numeric and categorical data.
numeric_features = ['age', 'fare']
numeric_transformer = Pipeline(steps=[
('imputer', SimpleImputer(strategy='median')),
('scaler', StandardScaler())])
categorical_features = ['embarked', 'sex', 'pclass']

categorical_transformer = Pipeline(steps=[
('imputer', SimpleImputer(strategy='constant', fill_value='missing')),
('onehot', OneHotEncoder(handle_unknown='ignore'))])
preprocessor = ColumnTransformer(
transformers=[
('num', numeric_transformer, numeric_features),
('cat', categorical_transformer, categorical_features)])
# Append classifier to preprocessing pipeline.

# Now we have a full prediction pipeline.
clf = Pipeline(steps=[('preprocessor', preprocessor),
('classifier', LogisticRegression(solver='lbfgs'))])
X = data.drop('survived', axis=1)
y = data['survived']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
print("model score: %.3f" % clf.score(X_test, y_test))
Out:
model score: 0.790
Using the prediction pipeline in a grid search
Grid search can also be performed on the different preprocessing steps defined in the
ColumnTransformer object, together with the classifier’s hyperparameters as part of the Pipeline.
We will search for both the imputer strategy of the numeric preprocessing and the regularization parameter
of the logistic regression using sklearn.model_selection.GridSearchCV .
param_grid = {
'preprocessor__num__imputer__strategy': ['mean', 'median'],
'classifier__C': [0.1, 1.0, 10, 100],
}
grid_search = GridSearchCV(clf, param_grid, cv=10, iid=False)

grid_search.fit(X_train, y_train)
print(("best logistic regression from grid search: %.3f"

% grid_search.score(X_test, y_test)))
Out:
best logistic regression from grid search: 0.798

5.7.4 Selecting dimensionality reduction with Pipeline and GridSearchCV
This example constructs a pipeline that does dimensionality reduction followed by prediction with a support vector
classifier. It demonstrates the use of GridSearchCV and Pipeline to optimize over different classes of estimators
in a single CV run – unsupervised PCA and NMF dimensionality reductions are compared to univariate feature selection
during the grid search.
Additionally, Pipeline can be instantiated with the memory argument to memoize the transformers within the
pipeline, avoiding to fit again the same transformers over and over.
Note that the use of memory to enable caching becomes interesting when the fitting of a transformer is costly.
Illustration of Pipeline and GridSearchCV
This section illustrates the use of a Pipeline with GridSearchCV

# Authors: Robert McGibbon, Joel Nothman, Guillaume Lemaitre
from __future__ import print_function, division
import numpy as np
from sklearn.decomposition import PCA, NMF
from sklearn.feature_selection import SelectKBest, chi2
print(__doc__)
pipe = Pipeline([
# the reduce_dim stage is populated by the param_grid
('reduce_dim', None),
('classify', LinearSVC())
])
N_FEATURES_OPTIONS = [2, 4, 8]
C_OPTIONS = [1, 10, 100, 1000]
param_grid = [
{
'reduce_dim': [PCA(iterated_power=7), NMF()],
'reduce_dim__n_components': N_FEATURES_OPTIONS,
'classify__C': C_OPTIONS
},
{
'reduce_dim': [SelectKBest(chi2)],
'reduce_dim__k': N_FEATURES_OPTIONS,
'classify__C': C_OPTIONS
},
]
reducer_labels = ['PCA', 'NMF', 'KBest(chi2)']
grid = GridSearchCV(pipe, cv=5, n_jobs=1, param_grid=param_grid)

grid.fit(digits.data, digits.target)
mean_scores = np.array(grid.cv_results_['mean_test_score'])

# scores are in the order of param_grid iteration, which is alphabetical

mean_scores = mean_scores.reshape(len(C_OPTIONS), -1, len(N_FEATURES_OPTIONS))
# select score for best C
mean_scores = mean_scores.max(axis=0)
bar_offsets = (np.arange(len(N_FEATURES_OPTIONS)) *
(len(reducer_labels) + 1) + .5)
plt.figure()
COLORS = 'bgrcmyk'
for i, (label, reducer_scores) in enumerate(zip(reducer_labels, mean_scores)):
plt.bar(bar_offsets + i, reducer_scores, label=label, color=COLORS[i])
plt.title("Comparing feature reduction techniques")

plt.xlabel('Reduced number of features')
plt.xticks(bar_offsets + len(reducer_labels) / 2, N_FEATURES_OPTIONS)
plt.ylabel('Digit classification accuracy')
plt.ylim((0, 1))
plt.legend(loc='upper left')
plt.show()

Caching transformers within a Pipeline
It is sometimes worthwhile storing the state of a specific transformer since it could be used again. Using a
pipeline in GridSearchCV triggers such situations. Therefore, we use the argument memory to enable
caching.
Warning: Note that this example is, however, only an illustration since for this specific case fitting
PCA is not necessarily slower than loading the cache. Hence, use the memory constructor parameter
when the fitting of a transformer is costly.
from tempfile import mkdtemp

from shutil import rmtree
# Create a temporary folder to store the transformers of the pipeline

cachedir = mkdtemp()
memory = Memory(cachedir=cachedir, verbose=10)
cached_pipe = Pipeline([('reduce_dim', PCA()),
('classify', LinearSVC())],
memory=memory)
# This time, a cached pipeline will be used within the grid search
grid = GridSearchCV(cached_pipe, cv=5, n_jobs=1, param_grid=param_grid)
grid.fit(digits.data, digits.target)
# Delete the temporary cache before exiting

rmtree(cachedir)
Out:
________________________________________________________________________________
[Memory] Calling sklearn.pipeline._fit_transform_one...
_fit_transform_one(PCA(copy=True, iterated_power=7, n_components=2, random_state=None,
svd_solver='auto', tol=0.0, whiten=False),
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________fit_transform_one - 0.0s, 0.0min
________________________________________________________________________________
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________


array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 9]), None)
________________________________________________________________________________
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 9]), None)
________________________________________________________________________________


array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 9]), None)
________________________________________________________________________________
_fit_transform_one(NMF(alpha=0.0, beta_loss='frobenius', init=None, l1_ratio=0.0, max_
˓→iter=200,
n_components=2, random_state=None, shuffle=False, solver='cd',

tol=0.0001, verbose=0),
array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________


˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 9]), None)
________________________________________________________________________________
˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)


________________________________________________________________________________
˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 9]), None)
________________________________________________________________________________
˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→iter=200,

array([[0., ..., 0.],

...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 9]), None)
[Memory]0.0s, 0.0min : Loading _fit_transform_one from /tmp/tmpd22i53nw/joblib/
˓→sklearn/pipeline/_fit_transform_one/ce059dd4031139a6b2806c6eeb882177
___________________________________fit_transform_one cache loaded - 0.0s, 0.0min

˓→sklearn/pipeline/_fit_transform_one/fd2f9ad0f7503004c6c4e5daa84b3d81

˓→sklearn/pipeline/_fit_transform_one/f7699cba03794b00c4e7ad12918c3870

˓→sklearn/pipeline/_fit_transform_one/637c1c59f2a2bebcf577f17752da8e7b

˓→sklearn/pipeline/_fit_transform_one/662bd46dd1274c93e6633e1814aa6d18

˓→sklearn/pipeline/_fit_transform_one/340ab152c19b30c5094b0cc3d48088ce

˓→sklearn/pipeline/_fit_transform_one/5f5b28b918cd25901092d13d441e4d26

˓→sklearn/pipeline/_fit_transform_one/cfb5ed9e762b82923af42c9cdd8a029f

˓→sklearn/pipeline/_fit_transform_one/f43b55f88ba83096d0f246bc7cdeeacc

˓→sklearn/pipeline/_fit_transform_one/4a4f9f5b5aef5a99552c40788a041c3a

˓→sklearn/pipeline/_fit_transform_one/4d9657eb86031c9c3f4d22e2f3c9cb43

˓→sklearn/pipeline/_fit_transform_one/1251d528711edc2cb2ff337ef4d2b930

˓→sklearn/pipeline/_fit_transform_one/5899433ec17984cc1dc0c0b58c092f48

˓→sklearn/pipeline/_fit_transform_one/ac4979a637124e8de3021326e803952f

˓→sklearn/pipeline/_fit_transform_one/62198984c58175f6c8de85a08adac475


˓→sklearn/pipeline/_fit_transform_one/6fff13cab5b44100846e3d5ebcf5b777

˓→sklearn/pipeline/_fit_transform_one/8d0ec968ae6c20d5b13136352e9b3221

˓→sklearn/pipeline/_fit_transform_one/9130f54ef4a686fa57506a07f278a220

˓→sklearn/pipeline/_fit_transform_one/0553e2fdefbc628d4fc448fe0485529b

˓→sklearn/pipeline/_fit_transform_one/01ae613cb5d86163f07ea3ad33679ea1

˓→sklearn/pipeline/_fit_transform_one/7a8c73996adeba89c910d9043e4c8065

˓→sklearn/pipeline/_fit_transform_one/39cb0260b02feb8c2c86cd648f109e38

˓→sklearn/pipeline/_fit_transform_one/a2dfe8b09115d1094021450fa7dea05a

˓→sklearn/pipeline/_fit_transform_one/e3695525c67b22b8572c03b14c0dc50b

˓→sklearn/pipeline/_fit_transform_one/03bc3c8801621ecb55d332f660286f0f

˓→sklearn/pipeline/_fit_transform_one/714bb222f2d2b3701cde1ff62ed70e6b

˓→sklearn/pipeline/_fit_transform_one/9a0bbb81ae7f721edae06009b7927f36

˓→sklearn/pipeline/_fit_transform_one/07ed5c7ffed979168bc6b10dde11fedf

˓→sklearn/pipeline/_fit_transform_one/67e7ced7c33c8539bd4754a3a43ccb25

˓→sklearn/pipeline/_fit_transform_one/53f9f651406e0182dfcfe2d0ca68db80
































































________________________________________________________________________________
_fit_transform_one(SelectKBest(k=2, score_func=<function chi2 at 0x7ff1f89b1620>),
˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________


˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 9]), None)
________________________________________________________________________________
˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→array([[0., ..., 0.],

...,
[0., ..., 0.]]), array([0, ..., 9]), None)
________________________________________________________________________________
˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
________________________________________________________________________________
˓→array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 9]), None)
˓→sklearn/pipeline/_fit_transform_one/957fad013ca010289fccaf07eba68f62

˓→sklearn/pipeline/_fit_transform_one/481779f7d475f07d3a5e4f0bc06eafab

˓→sklearn/pipeline/_fit_transform_one/caf47ab6c6c883d05da5da8684cefb47

˓→sklearn/pipeline/_fit_transform_one/e31b0f98339c869d8c03a3273af6511e

˓→sklearn/pipeline/_fit_transform_one/6f27e62b861a27d22aec7689f36f86af

˓→sklearn/pipeline/_fit_transform_one/bca47aac37d2a7d757128c6a4d334ae8

˓→sklearn/pipeline/_fit_transform_one/bb5dbd9ce947c86b88a0430ac920077c


˓→sklearn/pipeline/_fit_transform_one/6f7ed41e81892936065cf6e231300d92

˓→sklearn/pipeline/_fit_transform_one/4e4cf1344179544b063154ff81b1cc2c

˓→sklearn/pipeline/_fit_transform_one/0f59c30e1159b6229e5f0a97fd042cc2

˓→sklearn/pipeline/_fit_transform_one/c35366c71fb804f90ed882ea82f61800

˓→sklearn/pipeline/_fit_transform_one/2eb4e2b554c826175d5861fbe4121cf1

˓→sklearn/pipeline/_fit_transform_one/68544d0e24f376a2a0172b8c07eb235f

˓→sklearn/pipeline/_fit_transform_one/09042604ce6aea29e15cc3faafb0284d

˓→sklearn/pipeline/_fit_transform_one/5732ea62c8724f7e20a4cc267ed111d6
































________________________________________________________________________________


˓→iter=200,

array([[0., ..., 0.],
...,
[0., ..., 0.]]), array([0, ..., 8]), None)
The PCA fitting is only computed at the evaluation of the first configuration of the C parameter of the LinearSVC
classifier. The other configurations of C will trigger the loading of the cached PCA estimator data, leading to save pro-
cessing time. Therefore, the use of caching the pipeline using memory is highly beneficial when fitting a transformer
is costly.
5.7.5 Column Transformer with Heterogeneous Data Sources
Datasets can often contain components of that require different feature extraction and processing pipelines. This
scenario might occur when:
1. Your dataset consists of heterogeneous data types (e.g. raster images and text captions)
2. Your dataset is stored in a Pandas DataFrame and different columns require different processing pipelines.
This example demonstrates how to use sklearn.compose.ColumnTransformer on a dataset containing dif-
ferent types of features. We use the 20-newsgroups dataset and compute standard bag-of-words features for the subject
line and body in separate pipelines as well as ad hoc features on the body. We combine them (with weights) using a
ColumnTransformer and finally train a classifier on the combined set of features.
The choice of features is not particularly helpful, but serves to illustrate the technique.
Out:
0 0.96 0.62 0.76 494

1 0.25 0.84 0.39 76
micro avg 0.65 0.65 0.65 570

macro avg 0.61 0.73 0.57 570
weighted avg 0.87 0.65 0.71 570
# Author: Matt Terry <matt.terry@gmail.com>

#

import numpy as np
from sklearn.base import BaseEstimator, TransformerMixin

from sklearn.datasets.twenty_newsgroups import strip_newsgroup_footer
from sklearn.datasets.twenty_newsgroups import strip_newsgroup_quoting
from sklearn.decomposition import TruncatedSVD
from sklearn.feature_extraction import DictVectorizer
from sklearn.compose import ColumnTransformer
class TextStats(BaseEstimator, TransformerMixin):

"""Extract features from each document for DictVectorizer"""
def fit(self, x, y=None):

return self
def transform(self, posts):

return [{'length': len(text),
'num_sentences': text.count('.')}
for text in posts]
class SubjectBodyExtractor(BaseEstimator, TransformerMixin):

"""Extract the subject & body from a usenet post in a single pass.
Takes a sequence of strings and produces a dict of sequences. Keys are

`subject` and `body`.
"""
def fit(self, x, y=None):
return self
def transform(self, posts):

# construct object dtype array with two columns
# first column = 'subject' and second column = 'body'
features = np.empty(shape=(len(posts), 2), dtype=object)
for i, text in enumerate(posts):
headers, _, bod = text.partition('\n\n')
bod = strip_newsgroup_footer(bod)
bod = strip_newsgroup_quoting(bod)
features[i, 1] = bod
prefix = 'Subject:'
sub = ''
for line in headers.split('\n'):
if line.startswith(prefix):
sub = line[len(prefix):]
break
features[i, 0] = sub
return features
pipeline = Pipeline([

# Extract the subject & body

('subjectbody', SubjectBodyExtractor()),
# Use ColumnTransformer to combine the features from subject and body

('union', ColumnTransformer(
[
# Pulling features from the post's subject line (first column)
('subject', TfidfVectorizer(min_df=50), 0),
# Pipeline for standard bag-of-words model for body (second column)

('body_bow', Pipeline([
('tfidf', TfidfVectorizer()),
('best', TruncatedSVD(n_components=50)),
]), 1),
# Pipeline for pulling ad hoc features from post's body

('body_stats', Pipeline([
('stats', TextStats()), # returns a list of dicts
('vect', DictVectorizer()), # list of dicts -> feature matrix
]), 1),
],
# weight components in ColumnTransformer

transformer_weights={
'subject': 0.8,
'body_bow': 0.5,
'body_stats': 1.0,
}
)),
# Use a SVC classifier on the combined features

('svc', LinearSVC()),
])
# limit the list of categories to make running this example faster.

categories = ['alt.atheism', 'talk.religion.misc']
train = fetch_20newsgroups(random_state=1,
subset='train',
categories=categories,
)
test = fetch_20newsgroups(random_state=1,
subset='test',
categories=categories,
)
pipeline.fit(train.data, train.target)
y = pipeline.predict(test.data)
print(classification_report(y, test.target))

5.7.6 Effect of transforming the targets in regression model
In this example, we give an overview of the sklearn.compose.TransformedTargetRegressor. Two

examples illustrate the benefit of transforming the targets before learning a linear regression model. The first example
uses synthetic data while the second example is based on the Boston housing data set.
# Author: Guillaume Lemaitre <guillaume.lemaitre@inria.fr>

from __future__ import print_function, division
import numpy as np
import matplotlib
print(__doc__)
Synthetic example
from sklearn.datasets import make_regression

from sklearn.linear_model import RidgeCV
from sklearn.compose import TransformedTargetRegressor
from sklearn.metrics import median_absolute_error, r2_score

else:
A synthetic random regression problem is generated. The targets y are modified by: (i) translating all targets such
that all entries are non-negative and (ii) applying an exponential function to obtain non-linear targets which cannot be
fitted using a simple linear model.
Therefore, a logarithmic (np.log1p) and an exponential function (np.expm1) will be used to transform the targets before
training a linear regression model and using it for prediction.
X, y = make_regression(n_samples=10000, noise=100, random_state=0)

y = np.exp((y + abs(y.min())) / 200)
y_trans = np.log1p(y)
The following illustrate the probability density functions of the target before and after applying the logarithmic func-
tions.
f, (ax0, ax1) = plt.subplots(1, 2)
ax0.hist(y, bins=100, **density_param)

ax0.set_xlim([0, 2000])
ax0.set_ylabel('Probability')
ax0.set_xlabel('Target')
ax0.set_title('Target distribution')
ax1.hist(y_trans, bins=100, **density_param)

ax1.set_title('Transformed target distribution')
f.suptitle("Synthetic data", y=0.035)

f.tight_layout(rect=[0.05, 0.05, 0.95, 0.95])
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)
At first, a linear model will be applied on the original targets. Due to the non-linearity, the model trained will not
be precise during the prediction. Subsequently, a logarithmic function is used to linearize the targets, allowing better
prediction even with a similar linear model as reported by the median absolute error (MAE).
f, (ax0, ax1) = plt.subplots(1, 2, sharey=True)
regr = RidgeCV()
regr.fit(X_train, y_train)
y_pred = regr.predict(X_test)
ax0.scatter(y_test, y_pred)
ax0.plot([0, 2000], [0, 2000], '--k')
ax0.set_ylabel('Target predicted')
ax0.set_xlabel('True Target')
ax0.set_title('Ridge regression \n without target transformation')
ax0.text(100, 1750, r'$R^2$=%.2f, MAE=%.2f' % (
r2_score(y_test, y_pred), median_absolute_error(y_test, y_pred)))

ax0.set_xlim([0, 2000])
ax0.set_ylim([0, 2000])
regr_trans = TransformedTargetRegressor(regressor=RidgeCV(),
func=np.log1p,
inverse_func=np.expm1)
regr_trans.fit(X_train, y_train)
y_pred = regr_trans.predict(X_test)
ax1.plot([0, 2000], [0, 2000], '--k')
ax1.set_title('Ridge regression \n with target transformation')
ax1.text(100, 1750, r'$R^2$=%.2f, MAE=%.2f' % (
ax1.set_xlim([0, 2000])
ax1.set_ylim([0, 2000])
f.suptitle("Synthetic data", y=0.035)


Real-world data set
In a similar manner, the boston housing data set is used to show the impact of transforming the targets before learning a
model. In this example, the targets to be predicted corresponds to the weighted distances to the five Boston employment
centers.

from sklearn.preprocessing import QuantileTransformer, quantile_transform
dataset = load_boston()
target = np.array(dataset.feature_names) == "DIS"
X = dataset.data[:, np.logical_not(target)]
y = dataset.data[:, target].squeeze()
y_trans = quantile_transform(dataset.data[:, target],
output_distribution='normal').squeeze()
A sklearn.preprocessing.QuantileTransformer is used such that the targets follows a normal distri-

bution before applying a sklearn.linear_model.RidgeCV model.
f, (ax0, ax1) = plt.subplots(1, 2)
ax0.hist(y, bins=100, **density_param)

ax0.set_title('Target distribution')
ax1.hist(y_trans, bins=100, **density_param)

ax1.set_title('Transformed target distribution')
f.suptitle("Boston housing data: distance to employment centers", y=0.035)


The effect of the transformer is weaker than on the synthetic data. However, the transform induces a decrease of the
MAE.
f, (ax0, ax1) = plt.subplots(1, 2, sharey=True)
regr = RidgeCV()
regr.fit(X_train, y_train)
y_pred = regr.predict(X_test)
ax0.plot([0, 10], [0, 10], '--k')
ax0.set_title('Ridge regression \n without target transformation')
ax0.text(1, 9, r'$R^2$=%.2f, MAE=%.2f' % (
ax0.set_xlim([0, 10])
ax0.set_ylim([0, 10])
regr_trans = TransformedTargetRegressor(
regressor=RidgeCV(),
transformer=QuantileTransformer(output_distribution='normal'))
regr_trans.fit(X_train, y_train)
y_pred = regr_trans.predict(X_test)

ax1.plot([0, 10], [0, 10], '--k')

ax1.set_title('Ridge regression \n with target transformation')
ax1.text(1, 9, r'$R^2$=%.2f, MAE=%.2f' % (
ax1.set_xlim([0, 10])
ax1.set_ylim([0, 10])
f.suptitle("Boston housing data: distance to employment centers", y=0.035)

plt.show()
5.8 Covariance estimation
Examples concerning the sklearn.covariance module.
5.8. Covariance estimation 901

5.8.1 Ledoit-Wolf vs OAS estimation
The usual covariance maximum likelihood estimate can be regularized using shrinkage. Ledoit and Wolf proposed a
close formula to compute the asymptotically optimal shrinkage parameter (minimizing a MSE criterion), yielding the
Ledoit-Wolf covariance estimate.
Chen et al. proposed an improvement of the Ledoit-Wolf shrinkage parameter, the OAS coefficient, whose convergence
is significantly better under the assumption that the data are Gaussian.
This example, inspired from Chen’s publication [1], shows a comparison of the estimated MSE of the LW and OAS
methods, using Gaussian distributed data.
[1] “Shrinkage Algorithms for MMSE Covariance Estimation” Chen et al., IEEE Trans. on Sign. Proc., Volume 58,
Issue 10, October 2010.
print(__doc__)
import numpy as np
from scipy.linalg import toeplitz, cholesky
from sklearn.covariance import LedoitWolf, OAS
np.random.seed(0)
n_features = 100
# simulation covariance matrix (AR(1) process)
r = 0.1
real_cov = toeplitz(r ** np.arange(n_features))
coloring_matrix = cholesky(real_cov)
n_samples_range = np.arange(6, 31, 1)

repeat = 100
lw_mse = np.zeros((n_samples_range.size, repeat))
oa_mse = np.zeros((n_samples_range.size, repeat))
lw_shrinkage = np.zeros((n_samples_range.size, repeat))
oa_shrinkage = np.zeros((n_samples_range.size, repeat))
for i, n_samples in enumerate(n_samples_range):
for j in range(repeat):
X = np.dot(
np.random.normal(size=(n_samples, n_features)), coloring_matrix.T)
lw = LedoitWolf(store_precision=False, assume_centered=True)
lw.fit(X)
lw_mse[i, j] = lw.error_norm(real_cov, scaling=False)
lw_shrinkage[i, j] = lw.shrinkage_
oa = OAS(store_precision=False, assume_centered=True)
oa.fit(X)
oa_mse[i, j] = oa.error_norm(real_cov, scaling=False)
oa_shrinkage[i, j] = oa.shrinkage_
# plot MSE
plt.errorbar(n_samples_range, lw_mse.mean(1), yerr=lw_mse.std(1),
label='Ledoit-Wolf', color='navy', lw=2)
plt.errorbar(n_samples_range, oa_mse.mean(1), yerr=oa_mse.std(1),
label='OAS', color='darkorange', lw=2)
plt.ylabel("Squared error")

plt.legend(loc="upper right")
plt.title("Comparison of covariance estimators")
plt.xlim(5, 31)
# plot shrinkage coefficient

plt.errorbar(n_samples_range, lw_shrinkage.mean(1), yerr=lw_shrinkage.std(1),
label='Ledoit-Wolf', color='navy', lw=2)
plt.errorbar(n_samples_range, oa_shrinkage.mean(1), yerr=oa_shrinkage.std(1),
label='OAS', color='darkorange', lw=2)
plt.xlabel("n_samples")
plt.ylabel("Shrinkage")
plt.legend(loc="lower right")
plt.ylim(plt.ylim()[0], 1. + (plt.ylim()[1] - plt.ylim()[0]) / 10.)
plt.xlim(5, 31)
plt.show()

5.8.2 Sparse inverse covariance estimation
Using the GraphicalLasso estimator to learn a covariance and sparse precision from a small number of samples.
To estimate a probabilistic model (e.g. a Gaussian model), estimating the precision matrix, that is the inverse covari-
ance matrix, is as important as estimating the covariance matrix. Indeed a Gaussian model is parametrized by the
precision matrix.
To be in favorable recovery conditions, we sample the data from a model with a sparse inverse covariance matrix. In
addition, we ensure that the data is not too much correlated (limiting the largest coefficient of the precision matrix) and
that there a no small coefficients in the precision matrix that cannot be recovered. In addition, with a small number of
observations, it is easier to recover a correlation matrix rather than a covariance, thus we scale the time series.
Here, the number of samples is slightly larger than the number of dimensions, thus the empirical covariance is still
invertible. However, as the observations are strongly correlated, the empirical covariance matrix is ill-conditioned and
as a result its inverse –the empirical precision matrix– is very far from the ground truth.
If we use l2 shrinkage, as with the Ledoit-Wolf estimator, as the number of samples is small, we need to shrink a lot.
As a result, the Ledoit-Wolf precision is fairly close to the ground truth precision, that is not far from being diagonal,
but the off-diagonal structure is lost.
The l1-penalized estimator can recover part of this off-diagonal structure. It learns a sparse precision. It is not
able to recover the exact sparsity pattern: it detects too many non-zero coefficients. However, the highest non-zero
coefficients of the l1 estimated correspond to the non-zero coefficients in the ground truth. Finally, the coefficients of
the l1 precision estimate are biased toward zero: because of the penalty, they are all smaller than the corresponding
ground truth value, as can be seen on the figure.
Note that, the color range of the precision matrices is tweaked to improve readability of the figure. The full range of
values of the empirical precision is not displayed.
The alpha parameter of the GraphicalLasso setting the sparsity of the model is set by internal cross-validation in the
GraphicalLassoCV. As can be seen on figure 2, the grid to compute the cross-validation score is iteratively refined in
the neighborhood of the maximum.

•
print(__doc__)
# author: Gael Varoquaux <gael.varoquaux@inria.fr>
# Copyright: INRIA
import numpy as np
from sklearn.datasets import make_sparse_spd_matrix
from sklearn.covariance import GraphicalLassoCV, ledoit_wolf
# #############################################################################
# Generate the data
n_samples = 60
n_features = 20
prng = np.random.RandomState(1)
prec = make_sparse_spd_matrix(n_features, alpha=.98,
smallest_coef=.4,
largest_coef=.7,
random_state=prng)
cov = linalg.inv(prec)
d = np.sqrt(np.diag(cov))
cov /= d
cov /= d[:, np.newaxis]
prec *= d
prec *= d[:, np.newaxis]
X = prng.multivariate_normal(np.zeros(n_features), cov, size=n_samples)
X -= X.mean(axis=0)
X /= X.std(axis=0)
# #############################################################################
# Estimate the covariance
emp_cov = np.dot(X.T, X) / n_samples
model = GraphicalLassoCV(cv=5)
model.fit(X)
cov_ = model.covariance_

prec_ = model.precision_
lw_cov_, _ = ledoit_wolf(X)
lw_prec_ = linalg.inv(lw_cov_)
# #############################################################################
# Plot the results
plt.subplots_adjust(left=0.02, right=0.98)
# plot the covariances

covs = [('Empirical', emp_cov), ('Ledoit-Wolf', lw_cov_),
('GraphicalLassoCV', cov_), ('True', cov)]
vmax = cov_.max()
for i, (name, this_cov) in enumerate(covs):
plt.imshow(this_cov, interpolation='nearest', vmin=-vmax, vmax=vmax,
cmap=plt.cm.RdBu_r)
plt.xticks(())
plt.yticks(())
plt.title('%s covariance' % name)
# plot the precisions

precs = [('Empirical', linalg.inv(emp_cov)), ('Ledoit-Wolf', lw_prec_),
('GraphicalLasso', prec_), ('True', prec)]
vmax = .9 * prec_.max()
for i, (name, this_prec) in enumerate(precs):
ax = plt.subplot(2, 4, i + 5)
plt.imshow(np.ma.masked_equal(this_prec, 0),
interpolation='nearest', vmin=-vmax, vmax=vmax,
cmap=plt.cm.RdBu_r)
plt.xticks(())
plt.yticks(())
plt.title('%s precision' % name)
if hasattr(ax, 'set_facecolor'):
ax.set_facecolor('.7')
else:
ax.set_axis_bgcolor('.7')
# plot the model selection metric

plt.axes([.2, .15, .75, .7])
plt.plot(model.cv_alphas_, np.mean(model.grid_scores_, axis=1), 'o-')
plt.axvline(model.alpha_, color='.5')
plt.title('Model selection')
plt.ylabel('Cross-validation score')
plt.xlabel('alpha')
plt.show()

5.8.3 Shrinkage covariance estimation: LedoitWolf vs OAS and max-likelihood
When working with covariance estimation, the usual approach is to use a maximum likelihood estimator, such as
the sklearn.covariance.EmpiricalCovariance. It is unbiased, i.e. it converges to the true (population)
covariance when given many observations. However, it can also be beneficial to regularize it, in order to reduce
its variance; this, in turn, introduces some bias. This example illustrates the simple regularization used in Shrunk
Covariance estimators. In particular, it focuses on how to set the amount of regularization, i.e. how to choose the
bias-variance trade-off.
Here we compare 3 approaches:
• Setting the parameter by cross-validating the likelihood on three folds according to a grid of potential shrinkage
parameters.
• A close formula proposed by Ledoit and Wolf to compute the asymptotically optimal regularization parameter
(minimizing a MSE criterion), yielding the sklearn.covariance.LedoitWolf covariance estimate.
• An improvement of the Ledoit-Wolf shrinkage, the sklearn.covariance.OAS, proposed by Chen et al.
Its convergence is significantly better under the assumption that the data are Gaussian, in particular for small
samples.
To quantify estimation error, we plot the likelihood of unseen data for different values of the shrinkage parameter. We
also show the choices by cross-validation, or with the LedoitWolf and OAS estimates.
Note that the maximum likelihood estimate corresponds to no shrinkage, and thus performs poorly. The Ledoit-Wolf
estimate performs really well, as it is close to the optimal and is computational not costly. In this example, the OAS
estimate is a bit further away. Interestingly, both approaches outperform cross-validation, which is significantly most
computationally costly.

print(__doc__)
import numpy as np
from sklearn.covariance import LedoitWolf, OAS, ShrunkCovariance, \

log_likelihood, empirical_covariance
# #############################################################################
n_features, n_samples = 40, 20
np.random.seed(42)
base_X_train = np.random.normal(size=(n_samples, n_features))
base_X_test = np.random.normal(size=(n_samples, n_features))
# Color samples
coloring_matrix = np.random.normal(size=(n_features, n_features))
X_train = np.dot(base_X_train, coloring_matrix)
X_test = np.dot(base_X_test, coloring_matrix)
# #############################################################################
# Compute the likelihood on test data

# spanning a range of possible shrinkage coefficient values

shrinkages = np.logspace(-2, 0, 30)
negative_logliks = [-ShrunkCovariance(shrinkage=s).fit(X_train).score(X_test)
for s in shrinkages]
# under the ground-truth model, which we would not have access to in real
# settings
real_cov = np.dot(coloring_matrix.T, coloring_matrix)
emp_cov = empirical_covariance(X_train)
loglik_real = -log_likelihood(emp_cov, linalg.inv(real_cov))
# #############################################################################
# Compare different approaches to setting the parameter
# GridSearch for an optimal shrinkage coefficient

tuned_parameters = [{'shrinkage': shrinkages}]
cv = GridSearchCV(ShrunkCovariance(), tuned_parameters, cv=5)
cv.fit(X_train)
# Ledoit-Wolf optimal shrinkage coefficient estimate

lw = LedoitWolf()
loglik_lw = lw.fit(X_train).score(X_test)
# OAS coefficient estimate

oa = OAS()
loglik_oa = oa.fit(X_train).score(X_test)
# #############################################################################
# Plot results
fig = plt.figure()
plt.title("Regularized covariance: likelihood and shrinkage coefficient")
plt.xlabel('Regularization parameter: shrinkage coefficient')
plt.ylabel('Error: negative log-likelihood on test data')
# range shrinkage curve
plt.loglog(shrinkages, negative_logliks, label="Negative log-likelihood")
plt.plot(plt.xlim(), 2 * [loglik_real], '--r',

label="Real covariance likelihood")
# adjust view
lik_max = np.amax(negative_logliks)
lik_min = np.amin(negative_logliks)
ymin = lik_min - 6. * np.log((plt.ylim()[1] - plt.ylim()[0]))
ymax = lik_max + 10. * np.log(lik_max - lik_min)
xmin = shrinkages[0]
xmax = shrinkages[-1]
# LW likelihood
plt.vlines(lw.shrinkage_, ymin, -loglik_lw, color='magenta',
linewidth=3, label='Ledoit-Wolf estimate')
# OAS likelihood
plt.vlines(oa.shrinkage_, ymin, -loglik_oa, color='purple',
linewidth=3, label='OAS estimate')
# best CV estimator likelihood
plt.vlines(cv.best_estimator_.shrinkage, ymin,
-cv.best_estimator_.score(X_test), color='cyan',
linewidth=3, label='Cross-validation best estimate')
plt.ylim(ymin, ymax)

plt.xlim(xmin, xmax)
plt.legend()
plt.show()
5.8.4 Robust covariance estimation and Mahalanobis distances relevance
An example to show covariance estimation with the Mahalanobis distances on Gaussian distributed data.
For Gaussian distributed data, the distance of an observation 𝑥𝑖 to the mode of the distribution can be computed using
its Mahalanobis distance: 𝑑(𝜇,Σ) (𝑥𝑖 )2 = (𝑥𝑖 − 𝜇)′ Σ−1 (𝑥𝑖 − 𝜇) where 𝜇 and Σ are the location and the covariance of
the underlying Gaussian distribution.
In practice, 𝜇 and Σ are replaced by some estimates. The usual covariance maximum likelihood estimate is very
sensitive to the presence of outliers in the data set and therefor, the corresponding Mahalanobis distances are. One
would better have to use a robust estimator of covariance to guarantee that the estimation is resistant to “erroneous”
observations in the data set and that the associated Mahalanobis distances accurately reflect the true organisation of
the observations.
The Minimum Covariance Determinant estimator is a robust, high-breakdown point (i.e. it can be used to estimate the
𝑛 −𝑛 −1
covariance matrix of highly contaminated datasets, up to samples 2 features outliers) estimator of covariance. The idea is
𝑛samples +𝑛features +1
to find 2 observations whose empirical covariance has the smallest determinant, yielding a “pure” subset
of observations from which to compute standards estimates of location and covariance.
The Minimum Covariance Determinant estimator (MCD) has been introduced by P.J.Rousseuw in [1].
This example illustrates how the Mahalanobis distances are affected by outlying data: observations drawn from a
contaminating distribution are not distinguishable from the observations coming from the real, Gaussian distribution
that one may want to work with. Using MCD-based Mahalanobis distances, the two populations become distinguish-
able. Associated applications are outliers detection, observations ranking, clustering, . . . For visualization purpose,
the cubic root of the Mahalanobis distances are represented in the boxplot, as Wilson and Hilferty suggest [2]
[1] P. J. Rousseeuw. Least median of squares regression. J. Am Stat Ass, 79:871, 1984.
[2] Wilson, E. B., & Hilferty, M. M. (1931). The distribution of chi-square. Proceedings of the National
Academy of Sciences of the United States of America, 17, 684-688.

print(__doc__)
import numpy as np
from sklearn.covariance import EmpiricalCovariance, MinCovDet
n_samples = 125
n_outliers = 25
n_features = 2
# generate data
gen_cov = np.eye(n_features)
gen_cov[0, 0] = 2.
X = np.dot(np.random.randn(n_samples, n_features), gen_cov)
# add some outliers
outliers_cov = np.eye(n_features)
outliers_cov[np.arange(1, n_features), np.arange(1, n_features)] = 7.
X[-n_outliers:] = np.dot(np.random.randn(n_outliers, n_features), outliers_cov)
# fit a Minimum Covariance Determinant (MCD) robust estimator to data

robust_cov = MinCovDet().fit(X)
# compare estimators learnt from the full data set with true parameters
emp_cov = EmpiricalCovariance().fit(X)

# #############################################################################
# Display results
fig = plt.figure()
plt.subplots_adjust(hspace=-.1, wspace=.4, top=.95, bottom=.05)
# Show data set

subfig1 = plt.subplot(3, 1, 1)
inlier_plot = subfig1.scatter(X[:, 0], X[:, 1],
color='black', label='inliers')
outlier_plot = subfig1.scatter(X[:, 0][-n_outliers:], X[:, 1][-n_outliers:],
color='red', label='outliers')
subfig1.set_xlim(subfig1.get_xlim()[0], 11.)
subfig1.set_title("Mahalanobis distances of a contaminated data set:")
# Show contours of the distance functions

xx, yy = np.meshgrid(np.linspace(plt.xlim()[0], plt.xlim()[1], 100),
np.linspace(plt.ylim()[0], plt.ylim()[1], 100))
zz = np.c_[xx.ravel(), yy.ravel()]
mahal_emp_cov = emp_cov.mahalanobis(zz)
mahal_emp_cov = mahal_emp_cov.reshape(xx.shape)
emp_cov_contour = subfig1.contour(xx, yy, np.sqrt(mahal_emp_cov),
cmap=plt.cm.PuBu_r,
linestyles='dashed')
mahal_robust_cov = robust_cov.mahalanobis(zz)
mahal_robust_cov = mahal_robust_cov.reshape(xx.shape)
robust_contour = subfig1.contour(xx, yy, np.sqrt(mahal_robust_cov),
cmap=plt.cm.YlOrBr_r, linestyles='dotted')
subfig1.legend([emp_cov_contour.collections[1], robust_contour.collections[1],
inlier_plot, outlier_plot],
['MLE dist', 'robust dist', 'inliers', 'outliers'],
loc="upper right", borderaxespad=0)
plt.xticks(())
plt.yticks(())
# Plot the scores for each point

emp_mahal = emp_cov.mahalanobis(X - np.mean(X, 0)) ** (0.33)
subfig2.boxplot([emp_mahal[:-n_outliers], emp_mahal[-n_outliers:]], widths=.25)
subfig2.plot(np.full(n_samples - n_outliers, 1.26),
emp_mahal[:-n_outliers], '+k', markeredgewidth=1)
subfig2.plot(np.full(n_outliers, 2.26),
emp_mahal[-n_outliers:], '+k', markeredgewidth=1)
subfig2.axes.set_xticklabels(('inliers', 'outliers'), size=15)
subfig2.set_ylabel(r"$\sqrt[3]{\rm{(Mahal. dist.)}}$", size=16)
subfig2.set_title("1. from non-robust estimates\n(Maximum Likelihood)")
plt.yticks(())
robust_mahal = robust_cov.mahalanobis(X - robust_cov.location_) ** (0.33)

subfig3.boxplot([robust_mahal[:-n_outliers], robust_mahal[-n_outliers:]],
widths=.25)
subfig3.plot(np.full(n_samples - n_outliers, 1.26),
robust_mahal[:-n_outliers], '+k', markeredgewidth=1)
subfig3.plot(np.full(n_outliers, 2.26),
robust_mahal[-n_outliers:], '+k', markeredgewidth=1)

subfig3.axes.set_xticklabels(('inliers', 'outliers'), size=15)

subfig3.set_ylabel(r"$\sqrt[3]{\rm{(Mahal. dist.)}}$", size=16)
subfig3.set_title("2. from robust estimates\n(Minimum Covariance Determinant)")
plt.yticks(())
plt.show()
5.8.5 Robust vs Empirical covariance estimate
The usual covariance maximum likelihood estimate is very sensitive to the presence of outliers in the data set. In
such a case, it would be better to use a robust estimator of covariance to guarantee that the estimation is resistant to
“erroneous” observations in the data set.
Minimum Covariance Determinant Estimator
The Minimum Covariance Determinant estimator is a robust, high-breakdown point (i.e. it can be used to estimate the
𝑛 −𝑛 −1
covariance matrix of highly contaminated datasets, up to samples 2 features outliers) estimator of covariance. The idea is
𝑛samples +𝑛features +1
to find 2 observations whose empirical covariance has the smallest determinant, yielding a “pure” subset
of observations from which to compute standards estimates of location and covariance. After a correction step aiming
at compensating the fact that the estimates were learned from only a portion of the initial data, we end up with robust
estimates of the data set location and covariance.
The Minimum Covariance Determinant estimator (MCD) has been introduced by P.J.Rousseuw in1 .
Evaluation
In this example, we compare the estimation errors that are made when using various types of location and covariance
estimates on contaminated Gaussian distributed data sets:
• The mean and the empirical covariance of the full dataset, which break down as soon as there are outliers in the
data set
• The robust MCD, that has a low error provided 𝑛samples > 5𝑛features
• The mean and the empirical covariance of the observations that are known to be good ones. This can be consid-
ered as a “perfect” MCD estimation, so one can trust our implementation by comparing to this case.
1 P. J. Rousseeuw. Least median of squares regression. Journal of American Statistical Ass., 79:871, 1984.

References
print(__doc__)
import numpy as np
from sklearn.covariance import EmpiricalCovariance, MinCovDet
# example settings
n_samples = 80
n_features = 5
repeat = 10
range_n_outliers = np.concatenate(
(np.linspace(0, n_samples / 8, 5),
np.linspace(n_samples / 8, n_samples / 2, 5)[1:-1])).astype(np.int)
# definition of arrays to store results

err_loc_mcd = np.zeros((range_n_outliers.size, repeat))
err_cov_mcd = np.zeros((range_n_outliers.size, repeat))
err_loc_emp_full = np.zeros((range_n_outliers.size, repeat))
err_cov_emp_full = np.zeros((range_n_outliers.size, repeat))
err_loc_emp_pure = np.zeros((range_n_outliers.size, repeat))
err_cov_emp_pure = np.zeros((range_n_outliers.size, repeat))

# computation
for i, n_outliers in enumerate(range_n_outliers):
for j in range(repeat):
rng = np.random.RandomState(i * j)
# generate data
X = rng.randn(n_samples, n_features)
# add some outliers
outliers_index = rng.permutation(n_samples)[:n_outliers]
outliers_offset = 10. * \
(np.random.randint(2, size=(n_outliers, n_features)) - 0.5)
X[outliers_index] += outliers_offset
inliers_mask = np.ones(n_samples).astype(bool)
inliers_mask[outliers_index] = False
# fit a Minimum Covariance Determinant (MCD) robust estimator to data

mcd = MinCovDet().fit(X)
# compare raw robust estimates with the true location and covariance
err_loc_mcd[i, j] = np.sum(mcd.location_ ** 2)
err_cov_mcd[i, j] = mcd.error_norm(np.eye(n_features))
# compare estimators learned from the full data set with true
# parameters
err_loc_emp_full[i, j] = np.sum(X.mean(0) ** 2)
err_cov_emp_full[i, j] = EmpiricalCovariance().fit(X).error_norm(
np.eye(n_features))
# compare with an empirical covariance learned from a pure data set

# (i.e. "perfect" mcd)
pure_X = X[inliers_mask]
pure_location = pure_X.mean(0)
pure_emp_cov = EmpiricalCovariance().fit(pure_X)
err_loc_emp_pure[i, j] = np.sum(pure_location ** 2)
err_cov_emp_pure[i, j] = pure_emp_cov.error_norm(np.eye(n_features))
# Display results
font_prop = matplotlib.font_manager.FontProperties(size=11)
lw = 2
plt.errorbar(range_n_outliers, err_loc_mcd.mean(1),
yerr=err_loc_mcd.std(1) / np.sqrt(repeat),
label="Robust location", lw=lw, color='m')
plt.errorbar(range_n_outliers, err_loc_emp_full.mean(1),
yerr=err_loc_emp_full.std(1) / np.sqrt(repeat),
label="Full data set mean", lw=lw, color='green')
plt.errorbar(range_n_outliers, err_loc_emp_pure.mean(1),
yerr=err_loc_emp_pure.std(1) / np.sqrt(repeat),
label="Pure data set mean", lw=lw, color='black')
plt.title("Influence of outliers on the location estimation")
plt.ylabel(r"Error ($||\mu - \hat{\mu}||_2^2$)")
plt.legend(loc="upper left", prop=font_prop)
x_size = range_n_outliers.size
plt.errorbar(range_n_outliers, err_cov_mcd.mean(1),
yerr=err_cov_mcd.std(1),

label="Robust covariance (mcd)", color='m')

plt.errorbar(range_n_outliers[:(x_size // 5 + 1)],
err_cov_emp_full.mean(1)[:(x_size // 5 + 1)],
yerr=err_cov_emp_full.std(1)[:(x_size // 5 + 1)],
label="Full data set empirical covariance", color='green')
plt.plot(range_n_outliers[(x_size // 5):(x_size // 2 - 1)],
err_cov_emp_full.mean(1)[(x_size // 5):(x_size // 2 - 1)],
color='green', ls='--')
plt.errorbar(range_n_outliers, err_cov_emp_pure.mean(1),
yerr=err_cov_emp_pure.std(1),
label="Pure data set empirical covariance", color='black')
plt.title("Influence of outliers on the covariance estimation")
plt.xlabel("Amount of contamination (%)")
plt.ylabel("RMSE")
plt.legend(loc="upper center", prop=font_prop)
plt.show()
5.9 Cross decomposition
Examples concerning the sklearn.cross_decomposition module.
5.9.1 Compare cross decomposition methods
Simple usage of various cross decomposition algorithms: - PLSCanonical - PLSRegression, with multivariate re-
sponse, a.k.a. PLS2 - PLSRegression, with univariate response, a.k.a. PLS1 - CCA
Given 2 multivariate covarying two-dimensional datasets, X, and Y, PLS extracts the ‘directions of covariance’, i.e.
the components of each datasets that explain the most shared variance between both datasets. This is apparent on the
scatterplot matrix display: components 1 in dataset X and dataset Y are maximally correlated (points lie around the
first diagonal). This is also true for components 2 in both dataset, however, the correlation across datasets for different
components is weak: the point cloud is very spherical.

Out:
Corr(X)
[[ 1. 0.51 0.07 -0.05]
[ 0.51 1. 0.11 -0.01]
[ 0.07 0.11 1. 0.49]
[-0.05 -0.01 0.49 1. ]]
Corr(Y)
[[1. 0.48 0.05 0.03]
[0.48 1. 0.04 0.12]
[0.05 0.04 1. 0.51]
[0.03 0.12 0.51 1. ]]
True B (such that: Y = XB + Err)
[[1 1 1]
[2 2 2]
[0 0 0]
[0 0 0]
[0 0 0]
[0 0 0]
[0 0 0]
[0 0 0]
[0 0 0]
[0 0 0]]
Estimated B
[[ 1. 1. 1. ]
[ 2. 2. 2. ]
[-0. -0. 0. ]
[ 0. 0. 0. ]
[ 0. 0. 0. ]
5.9. Cross decomposition 917

[ 0. 0. -0. ]
[-0. -0. -0.1]
[-0. -0. 0. ]
[ 0. 0. 0.1]
[ 0. 0. -0. ]]
Estimated betas
[[ 1. ]
[ 2.1]
[ 0. ]
[ 0. ]
[ 0. ]
[-0. ]
[-0. ]
[ 0. ]
[-0. ]
[-0. ]]
print(__doc__)
import numpy as np
from sklearn.cross_decomposition import PLSCanonical, PLSRegression, CCA
# #############################################################################
# Dataset based latent variables model
n = 500
# 2 latents vars:
l1 = np.random.normal(size=n)
l2 = np.random.normal(size=n)
latents = np.array([l1, l1, l2, l2]).T

X = latents + np.random.normal(size=4 * n).reshape((n, 4))
Y = latents + np.random.normal(size=4 * n).reshape((n, 4))
X_train = X[:n // 2]
Y_train = Y[:n // 2]
X_test = X[n // 2:]
Y_test = Y[n // 2:]
print("Corr(X)")
print(np.round(np.corrcoef(X.T), 2))
print("Corr(Y)")
print(np.round(np.corrcoef(Y.T), 2))
# #############################################################################
# Canonical (symmetric) PLS
# Transform data
# ~~~~~~~~~~~~~~
plsca = PLSCanonical(n_components=2)
plsca.fit(X_train, Y_train)
X_train_r, Y_train_r = plsca.transform(X_train, Y_train)

X_test_r, Y_test_r = plsca.transform(X_test, Y_test)
# Scatter plot of scores

# ~~~~~~~~~~~~~~~~~~~~~~
# 1) On diagonal plot X vs Y scores on each components
plt.subplot(221)
plt.scatter(X_train_r[:, 0], Y_train_r[:, 0], label="train",
marker="o", c="b", s=25)
plt.scatter(X_test_r[:, 0], Y_test_r[:, 0], label="test",
marker="o", c="r", s=25)
plt.xlabel("x scores")
plt.ylabel("y scores")
plt.title('Comp. 1: X vs Y (test corr = %.2f)' %
np.corrcoef(X_test_r[:, 0], Y_test_r[:, 0])[0, 1])
plt.xticks(())
plt.yticks(())
plt.subplot(224)
plt.scatter(X_train_r[:, 1], Y_train_r[:, 1], label="train",
marker="o", c="b", s=25)
plt.scatter(X_test_r[:, 1], Y_test_r[:, 1], label="test",
marker="o", c="r", s=25)
plt.xlabel("x scores")
plt.ylabel("y scores")
plt.title('Comp. 2: X vs Y (test corr = %.2f)' %
np.corrcoef(X_test_r[:, 1], Y_test_r[:, 1])[0, 1])
plt.xticks(())
plt.yticks(())
# 2) Off diagonal plot components 1 vs 2 for X and Y

plt.subplot(222)
plt.scatter(X_train_r[:, 0], X_train_r[:, 1], label="train",
marker="*", c="b", s=50)
plt.scatter(X_test_r[:, 0], X_test_r[:, 1], label="test",
marker="*", c="r", s=50)
plt.xlabel("X comp. 1")
plt.ylabel("X comp. 2")
plt.title('X comp. 1 vs X comp. 2 (test corr = %.2f)'
% np.corrcoef(X_test_r[:, 0], X_test_r[:, 1])[0, 1])
plt.xticks(())
plt.yticks(())
plt.subplot(223)
plt.scatter(Y_train_r[:, 0], Y_train_r[:, 1], label="train",
marker="*", c="b", s=50)
plt.scatter(Y_test_r[:, 0], Y_test_r[:, 1], label="test",
marker="*", c="r", s=50)
plt.xlabel("Y comp. 1")
plt.ylabel("Y comp. 2")
plt.title('Y comp. 1 vs Y comp. 2 , (test corr = %.2f)'
% np.corrcoef(Y_test_r[:, 0], Y_test_r[:, 1])[0, 1])
plt.xticks(())
plt.yticks(())
5.9. Cross decomposition 919

plt.show()
# #############################################################################
# PLS regression, with multivariate response, a.k.a. PLS2
n = 1000
q = 3
p = 10
X = np.random.normal(size=n * p).reshape((n, p))
B = np.array([[1, 2] + [0] * (p - 2)] * q).T
# each Yj = 1*X1 + 2*X2 + noize
Y = np.dot(X, B) + np.random.normal(size=n * q).reshape((n, q)) + 5
pls2 = PLSRegression(n_components=3)
pls2.fit(X, Y)
print("True B (such that: Y = XB + Err)")
print(B)
# compare pls2.coef_ with B
print("Estimated B")
print(np.round(pls2.coef_, 1))
pls2.predict(X)
# PLS regression, with univariate response, a.k.a. PLS1
n = 1000
p = 10
X = np.random.normal(size=n * p).reshape((n, p))
y = X[:, 0] + 2 * X[:, 1] + np.random.normal(size=n * 1) + 5
pls1 = PLSRegression(n_components=3)
pls1.fit(X, y)
# note that the number of components exceeds 1 (the dimension of y)
print("Estimated betas")
print(np.round(pls1.coef_, 1))
# #############################################################################
# CCA (PLS mode B with symmetric deflation)
cca = CCA(n_components=2)
cca.fit(X_train, Y_train)
X_train_r, Y_train_r = cca.transform(X_train, Y_train)
X_test_r, Y_test_r = cca.transform(X_test, Y_test)
5.10 Dataset examples
Examples concerning the sklearn.datasets module.

5.10.1 The Digit Dataset
This dataset is made up of 1797 8x8 images. Each image, like the one shown below, is of a hand-written digit. In order
to utilize an 8x8 figure like this, we’d have to first transform it into a feature vector with length 64.
See here for more information about this dataset.
print(__doc__)

#Load the digits dataset

#Display the first digit

plt.figure(1, figsize=(3, 3))
plt.imshow(digits.images[-1], cmap=plt.cm.gray_r, interpolation='nearest')
plt.show()
5.10.2 The Iris Dataset
This data sets consists of 3 different types of irises’ (Setosa, Versicolour, and Virginica) petal and sepal length, stored
in a 150x4 numpy.ndarray
The rows being the samples and the columns being: Sepal Length, Sepal Width, Petal Length and Petal Width.
5.10. Dataset examples 921

The below plot uses the first two features. See here for more information on this dataset.

print(__doc__)


# import some data to play with

X = iris.data[:, :2] # we only take the first two features.
y = iris.target


plt.clf()

plt.scatter(X[:, 0], X[:, 1], c=y, cmap=plt.cm.Set1,
edgecolor='k')
plt.xlabel('Sepal length')

plt.ylabel('Sepal width')
plt.xticks(())
plt.yticks(())
# To getter a better understanding of interaction of the dimensions

# plot the first three PCA dimensions
fig = plt.figure(1, figsize=(8, 6))
ax = Axes3D(fig, elev=-150, azim=110)
X_reduced = PCA(n_components=3).fit_transform(iris.data)
ax.scatter(X_reduced[:, 0], X_reduced[:, 1], X_reduced[:, 2], c=y,
cmap=plt.cm.Set1, edgecolor='k', s=40)
ax.set_title("First three PCA directions")
ax.set_xlabel("1st eigenvector")
ax.set_ylabel("2nd eigenvector")
ax.set_zlabel("3rd eigenvector")
plt.show()
5.10.3 Plot randomly generated classification dataset
Plot several randomly generated 2D classification datasets. This example illustrates the datasets.
make_classification datasets.make_blobs and datasets.make_gaussian_quantiles func-
tions.
For make_classification, three binary and two multi-class classification datasets are generated, with different
numbers of informative features and clusters per class.

print(__doc__)
from sklearn.datasets import make_classification

from sklearn.datasets import make_gaussian_quantiles
plt.subplots_adjust(bottom=.05, top=.9, left=.05, right=.95)
plt.subplot(321)
plt.title("One informative feature, one cluster per class", fontsize='small')
X1, Y1 = make_classification(n_features=2, n_redundant=0, n_informative=1,
n_clusters_per_class=1)

plt.scatter(X1[:, 0], X1[:, 1], marker='o', c=Y1,

plt.subplot(322)
plt.title("Two informative features, one cluster per class", fontsize='small')
plt.subplot(323)
plt.title("Two informative features, two clusters per class",
fontsize='small')
X2, Y2 = make_classification(n_features=2, n_redundant=0, n_informative=2)
plt.subplot(324)
plt.title("Multi-class, two informative features, one cluster",
fontsize='small')
n_clusters_per_class=1, n_classes=3)
plt.subplot(325)
plt.title("Three blobs", fontsize='small')
X1, Y1 = make_blobs(n_features=2, centers=3)
plt.subplot(326)
plt.title("Gaussian divided into three quantiles", fontsize='small')
X1, Y1 = make_gaussian_quantiles(n_features=2, n_classes=3)
plt.show()
5.10.4 Plot randomly generated multilabel dataset
This illustrates the datasets.make_multilabel_classification dataset generator. Each sample consists of counts of two
features (up to 50 in total), which are differently distributed in each of two classes.
Points are labeled as follows, where Y means the class is present:

1 2 3 Color
Y N N Red
N Y N Blue
N N Y Yellow
Y Y N Purple
Y N Y Orange
Y Y N Green
Y Y Y Brown
A star marks the expected sample for each class; its size reflects the probability of selecting that class label.
The left and right examples highlight the n_labels parameter: more of the samples in the right plot have 2 or 3
labels.
Note that this two-dimensional example is very degenerate: generally the number of features would be much greater
than the “document length”, while here we have much larger documents than vocabulary. Similarly, with n_classes
> n_features, it is much less likely that a feature distinguishes a particular class.
Out:
The data was generated from (random_state=1013):
Class P(C) P(w0|C) P(w1|C)
red 0.64 0.97 0.03
blue 0.06 0.60 0.40
yellow 0.30 0.09 0.91

import numpy as np

from sklearn.datasets import make_multilabel_classification as make_ml_clf
print(__doc__)
COLORS = np.array(['!',
'#FF3333', # red
'#0198E1', # blue
'#BF5FFF', # purple
'#FCD116', # yellow
'#FF7216', # orange
'#4DBD33', # green
'#87421F' # brown
])
# Use same random seed for multiple calls to make_multilabel_classification to

# ensure same distributions
RANDOM_SEED = np.random.randint(2 ** 10)
def plot_2d(ax, n_labels=1, n_classes=3, length=50):

X, Y, p_c, p_w_c = make_ml_clf(n_samples=150, n_features=2,
n_classes=n_classes, n_labels=n_labels,
length=length, allow_unlabeled=False,
return_distributions=True,
random_state=RANDOM_SEED)
ax.scatter(X[:, 0], X[:, 1], color=COLORS.take((Y * [1, 2, 4]

).sum(axis=1)),
marker='.')
ax.scatter(p_w_c[0] * length, p_w_c[1] * length,
marker='*', linewidth=.5, edgecolor='black',
s=20 + 1500 * p_c ** 2,
color=COLORS.take([1, 2, 4]))
ax.set_xlabel('Feature 0 count')
return p_c, p_w_c
_, (ax1, ax2) = plt.subplots(1, 2, sharex='row', sharey='row', figsize=(8, 4))

plt.subplots_adjust(bottom=.15)
p_c, p_w_c = plot_2d(ax1, n_labels=1)

ax1.set_title('n_labels=1, length=50')
ax1.set_ylabel('Feature 1 count')
plot_2d(ax2, n_labels=3)
ax2.set_title('n_labels=3, length=50')
ax2.set_xlim(left=0, auto=True)
ax2.set_ylim(bottom=0, auto=True)
plt.show()
print('The data was generated from (random_state=%d):' % RANDOM_SEED)

print('Class', 'P(C)', 'P(w0|C)', 'P(w1|C)', sep='\t')
for k, p, p_w in zip(['red', 'blue', 'yellow'], p_c, p_w_c.T):
print('%s\t%0.2f\t%0.2f\t%0.2f' % (k, p, p_w[0], p_w[1]))

5.11 Decomposition
Examples concerning the sklearn.decomposition module.
5.11.1 Beta-divergence loss functions
A plot that compares the various Beta-divergence loss functions supported by the Multiplicative-Update (‘mu’) solver
in sklearn.decomposition.NMF.
import numpy as np
from sklearn.decomposition.nmf import _beta_divergence
print(__doc__)
x = np.linspace(0.001, 4, 1000)
y = np.zeros(x.shape)
colors = 'mbgyr'
for j, beta in enumerate((0., 0.5, 1., 1.5, 2.)):
5.11. Decomposition 929

for i, xi in enumerate(x):
y[i] = _beta_divergence(1, xi, 1, beta)
name = "beta = %1.1f" % beta
plt.plot(x, y, label=name, color=colors[j])
plt.xlabel("x")
plt.title("beta-divergence(1, x)")
plt.legend(loc=0)
plt.axis([0, 4, 0, 3])
plt.show()
5.11.2 PCA example with Iris Data-set
Principal Component Analysis applied to the Iris dataset.

See here for more information on this dataset.
print(__doc__)

import numpy as np
from sklearn import decomposition


np.random.seed(5)
centers = [[1, 1], [-1, -1], [1, -1]]

X = iris.data
y = iris.target
fig = plt.figure(1, figsize=(4, 3))

plt.clf()
plt.cla()
pca = decomposition.PCA(n_components=3)
pca.fit(X)
X = pca.transform(X)
for name, label in [('Setosa', 0), ('Versicolour', 1), ('Virginica', 2)]:

ax.text3D(X[y == label, 0].mean(),
X[y == label, 1].mean() + 1.5,
X[y == label, 2].mean(), name,
horizontalalignment='center',
bbox=dict(alpha=.5, edgecolor='w', facecolor='w'))
# Reorder the labels to have colors matching the cluster results
y = np.choose(y, [1, 2, 0]).astype(np.float)
ax.scatter(X[:, 0], X[:, 1], X[:, 2], c=y, cmap=plt.cm.nipy_spectral,
edgecolor='k')
plt.show()
5.11.3 Incremental PCA
Incremental principal component analysis (IPCA) is typically used as a replacement for principal component analysis
(PCA) when the dataset to be decomposed is too large to fit in memory. IPCA builds a low-rank approximation for the
input data using an amount of memory which is independent of the number of input data samples. It is still dependent
on the input data features, but changing the batch size allows for control of memory usage.
This example serves as a visual check that IPCA is able to find a similar projection of the data to PCA (to a sign flip),
while only processing a few samples at a time. This can be considered a “toy example”, as IPCA is intended for large
datasets which do not fit in main memory, requiring incremental approaches.


print(__doc__)
# Authors: Kyle Kastner

import numpy as np

from sklearn.decomposition import PCA, IncrementalPCA
iris = load_iris()
X = iris.data
y = iris.target
n_components = 2
ipca = IncrementalPCA(n_components=n_components, batch_size=10)

X_ipca = ipca.fit_transform(X)
pca = PCA(n_components=n_components)
X_pca = pca.fit_transform(X)
colors = ['navy', 'turquoise', 'darkorange']
for X_transformed, title in [(X_ipca, "Incremental PCA"), (X_pca, "PCA")]:

for color, i, target_name in zip(colors, [0, 1, 2], iris.target_names):
plt.scatter(X_transformed[y == i, 0], X_transformed[y == i, 1],
color=color, lw=2, label=target_name)
if "Incremental" in title:
err = np.abs(np.abs(X_pca) - np.abs(X_ipca)).mean()
plt.title(title + " of iris dataset\nMean absolute unsigned error "
"%.6f" % err)
else:
plt.title(title + " of iris dataset")
plt.legend(loc="best", shadow=False, scatterpoints=1)
plt.axis([-4, 4, -1.5, 1.5])
plt.show()
5.11.4 Comparison of LDA and PCA 2D projection of Iris dataset
The Iris dataset represents 3 kind of Iris flowers (Setosa, Versicolour and Virginica) with 4 attributes: sepal length,
sepal width, petal length and petal width.
Principal Component Analysis (PCA) applied to this data identifies the combination of attributes (principal compo-
nents, or directions in the feature space) that account for the most variance in the data. Here we plot the different
samples on the 2 first principal components.
Linear Discriminant Analysis (LDA) tries to identify attributes that account for the most variance between classes. In
particular, LDA, in contrast to PCA, is a supervised method, using known class labels.


•
Out:
explained variance ratio (first two components): [0.92461872 0.05306648]
print(__doc__)

X = iris.data
y = iris.target
target_names = iris.target_names
X_r = pca.fit(X).transform(X)
lda = LinearDiscriminantAnalysis(n_components=2)

X_r2 = lda.fit(X, y).transform(X)
# Percentage of variance explained for each components

print('explained variance ratio (first two components): %s'
% str(pca.explained_variance_ratio_))
plt.figure()
lw = 2
for color, i, target_name in zip(colors, [0, 1, 2], target_names):

plt.scatter(X_r[y == i, 0], X_r[y == i, 1], color=color, alpha=.8, lw=lw,
label=target_name)
plt.legend(loc='best', shadow=False, scatterpoints=1)
plt.title('PCA of IRIS dataset')
plt.figure()
for color, i, target_name in zip(colors, [0, 1, 2], target_names):
plt.scatter(X_r2[y == i, 0], X_r2[y == i, 1], alpha=.8, color=color,
label=target_name)
plt.legend(loc='best', shadow=False, scatterpoints=1)
plt.title('LDA of IRIS dataset')
plt.show()
5.11.5 Blind source separation using FastICA
An example of estimating sources from noisy data.

Independent component analysis (ICA) is used to estimate sources given noisy measurements. Imagine 3 instruments
playing simultaneously and 3 microphones recording the mixed signals. ICA is used to recover the sources ie. what
is played by each instrument. Importantly, PCA fails at recovering our instruments since the related signals reflect
non-Gaussian processes.

print(__doc__)
import numpy as np
from scipy import signal
from sklearn.decomposition import FastICA, PCA
# #############################################################################
np.random.seed(0)
n_samples = 2000
time = np.linspace(0, 8, n_samples)
s1 = np.sin(2 * time) # Signal 1 : sinusoidal signal

s2 = np.sign(np.sin(3 * time)) # Signal 2 : square signal
s3 = signal.sawtooth(2 * np.pi * time) # Signal 3: saw tooth signal
S = np.c_[s1, s2, s3]

S += 0.2 * np.random.normal(size=S.shape) # Add noise
S /= S.std(axis=0) # Standardize data

# Mix data
A = np.array([[1, 1, 1], [0.5, 2, 1.0], [1.5, 1.0, 2.0]]) # Mixing matrix
X = np.dot(S, A.T) # Generate observations

# Compute ICA
ica = FastICA(n_components=3)
S_ = ica.fit_transform(X) # Reconstruct signals
A_ = ica.mixing_ # Get estimated mixing matrix
# We can `prove` that the ICA model applies by reverting the unmixing.
assert np.allclose(X, np.dot(S_, A_.T) + ica.mean_)
# For comparison, compute PCA

H = pca.fit_transform(X) # Reconstruct signals based on orthogonal components
# #############################################################################
# Plot results
plt.figure()
models = [X, S, S_, H]

names = ['Observations (mixed signal)',
'True Sources',
'ICA recovered signals',
'PCA recovered signals']
colors = ['red', 'steelblue', 'orange']
for ii, (model, name) in enumerate(zip(models, names), 1):

plt.subplot(4, 1, ii)
plt.title(name)
for sig, color in zip(model.T, colors):
plt.plot(sig, color=color)

plt.show()
5.11.6 Principal components analysis (PCA)
These figures aid in illustrating how a point cloud can be very flat in one direction–which is where PCA comes in to
choose a direction that is not flat.

•
print(__doc__)
# Authors: Gael Varoquaux

# Jaques Grobler
# Kevin Hughes

import numpy as np
from scipy import stats
# #############################################################################
# Create the data

e = np.exp(1)
np.random.seed(4)
def pdf(x):
return 0.5 * (stats.norm(scale=0.25 / e).pdf(x)
+ stats.norm(scale=4 / e).pdf(x))
y = np.random.normal(scale=0.5, size=(30000))
x = np.random.normal(scale=0.5, size=(30000))
z = np.random.normal(scale=0.1, size=len(x))
density = pdf(x) * pdf(y)

pdf_z = pdf(5 * z)
density *= pdf_z
a = x + y
b = 2 * y
c = a - b + z
norm = np.sqrt(a.var() + b.var())

a /= norm
b /= norm
# #############################################################################
# Plot the figures
def plot_figs(fig_num, elev, azim):
fig = plt.figure(fig_num, figsize=(4, 3))
plt.clf()
ax = Axes3D(fig, rect=[0, 0, .95, 1], elev=elev, azim=azim)
ax.scatter(a[::10], b[::10], c[::10], c=density[::10], marker='+', alpha=.4)

Y = np.c_[a, b, c]
# Using SciPy's SVD, this would be:

# _, pca_score, V = scipy.linalg.svd(Y, full_matrices=False)
pca.fit(Y)
pca_score = pca.explained_variance_ratio_
V = pca.components_
x_pca_axis, y_pca_axis, z_pca_axis = 3 * V.T

x_pca_plane = np.r_[x_pca_axis[:2], - x_pca_axis[1::-1]]
y_pca_plane = np.r_[y_pca_axis[:2], - y_pca_axis[1::-1]]
z_pca_plane = np.r_[z_pca_axis[:2], - z_pca_axis[1::-1]]
x_pca_plane.shape = (2, 2)
y_pca_plane.shape = (2, 2)
z_pca_plane.shape = (2, 2)
ax.plot_surface(x_pca_plane, y_pca_plane, z_pca_plane)
elev = -40

azim = -80
plot_figs(1, elev, azim)
elev = 30
azim = 20
plot_figs(2, elev, azim)
plt.show()
5.11.7 FastICA on 2D point clouds
This example illustrates visually in the feature space a comparison by results using two different component analysis
techniques.
Independent component analysis (ICA) vs Principal component analysis (PCA).
Representing ICA in the feature space gives the view of ‘geometric ICA’: ICA is an algorithm that finds directions in
the feature space corresponding to projections with high non-Gaussianity. These directions need not be orthogonal in
the original feature space, but they are orthogonal in the whitened feature space, in which all directions correspond to
the same variance.
PCA, on the other hand, finds orthogonal directions in the raw feature space that correspond to directions accounting
for maximum variance.
Here we simulate independent sources using a highly non-Gaussian process, 2 student T with a low number of degrees
of freedom (top left figure). We mix them to create observations (top right figure). In this raw observation space,
directions identified by PCA are represented by orange vectors. We represent the signal in the PCA space, after
whitening by the variance corresponding to the PCA vectors (lower left). Running ICA corresponds to finding a
rotation in this space to identify the directions of largest non-Gaussianity (lower right).

print(__doc__)
# Authors: Alexandre Gramfort, Gael Varoquaux

import numpy as np
from sklearn.decomposition import PCA, FastICA
# #############################################################################
S = rng.standard_t(1.5, size=(20000, 2))
S[:, 0] *= 2.
# Mix data
A = np.array([[1, 1], [0, 2]]) # Mixing matrix
X = np.dot(S, A.T) # Generate observations
pca = PCA()
S_pca_ = pca.fit(X).transform(X)
ica = FastICA(random_state=rng)
S_ica_ = ica.fit(X).transform(X) # Estimate the sources

S_ica_ /= S_ica_.std(axis=0)
# #############################################################################
# Plot results
def plot_samples(S, axis_list=None):

plt.scatter(S[:, 0], S[:, 1], s=2, marker='o', zorder=10,
color='steelblue', alpha=0.5)
if axis_list is not None:
colors = ['orange', 'red']
for color, axis in zip(colors, axis_list):
axis /= axis.std()
x_axis, y_axis = axis
# Trick to get legend to work
plt.plot(0.1 * x_axis, 0.1 * y_axis, linewidth=2, color=color)
plt.quiver(0, 0, x_axis, y_axis, zorder=11, width=0.01, scale=6,
color=color)
plt.hlines(0, -3, 3)
plt.vlines(0, -3, 3)
plt.xlim(-3, 3)
plt.ylim(-3, 3)
plt.xlabel('x')
plt.ylabel('y')
plt.figure()
plot_samples(S / S.std())
plt.title('True Independent Sources')
axis_list = [pca.components_.T, ica.mixing_]

plot_samples(X / np.std(X), axis_list=axis_list)
legend = plt.legend(['PCA', 'ICA'], loc='upper right')
legend.set_zorder(100)
plt.title('Observations')
plot_samples(S_pca_ / np.std(S_pca_, axis=0))
plt.title('PCA recovered signals')
plot_samples(S_ica_ / np.std(S_ica_))
plt.title('ICA recovered signals')

plt.show()

5.11.8 Kernel PCA
This example shows that Kernel PCA is able to find a projection of the data that makes data linearly separable.
print(__doc__)
# Authors: Mathieu Blondel

# Andreas Mueller
import numpy as np
from sklearn.decomposition import PCA, KernelPCA

from sklearn.datasets import make_circles
np.random.seed(0)
X, y = make_circles(n_samples=400, factor=.3, noise=.05)
kpca = KernelPCA(kernel="rbf", fit_inverse_transform=True, gamma=10)

X_kpca = kpca.fit_transform(X)
X_back = kpca.inverse_transform(X_kpca)
pca = PCA()
X_pca = pca.fit_transform(X)

# Plot results
plt.figure()
plt.subplot(2, 2, 1, aspect='equal')
plt.title("Original space")
reds = y == 0
blues = y == 1
plt.scatter(X[reds, 0], X[reds, 1], c="red",

plt.scatter(X[blues, 0], X[blues, 1], c="blue",
plt.xlabel("$x_1$")
plt.ylabel("$x_2$")
X1, X2 = np.meshgrid(np.linspace(-1.5, 1.5, 50), np.linspace(-1.5, 1.5, 50))

X_grid = np.array([np.ravel(X1), np.ravel(X2)]).T
# projection on the first principal component (in the phi space)
Z_grid = kpca.transform(X_grid)[:, 0].reshape(X1.shape)
plt.contour(X1, X2, Z_grid, colors='grey', linewidths=1, origin='lower')
plt.scatter(X_pca[reds, 0], X_pca[reds, 1], c="red",
plt.scatter(X_pca[blues, 0], X_pca[blues, 1], c="blue",
plt.title("Projection by PCA")
plt.xlabel("1st principal component")
plt.ylabel("2nd component")
plt.scatter(X_kpca[reds, 0], X_kpca[reds, 1], c="red",
plt.scatter(X_kpca[blues, 0], X_kpca[blues, 1], c="blue",
plt.title("Projection by KPCA")
plt.xlabel(r"1st principal component in space induced by $\phi$")
plt.ylabel("2nd component")
plt.scatter(X_back[reds, 0], X_back[reds, 1], c="red",
plt.scatter(X_back[blues, 0], X_back[blues, 1], c="blue",
plt.title("Original space after inverse transform")
plt.xlabel("$x_1$")
plt.ylabel("$x_2$")
plt.show()

5.11.9 Model selection with Probabilistic PCA and Factor Analysis (FA)
Probabilistic PCA and Factor Analysis are probabilistic models. The consequence is that the likelihood of new data
can be used for model selection and covariance estimation. Here we compare PCA and FA with cross-validation on
low rank data corrupted with homoscedastic noise (noise variance is the same for each feature) or heteroscedastic noise
(noise variance is the different for each feature). In a second step we compare the model likelihood to the likelihoods
obtained from shrinkage covariance estimators.
One can observe that with homoscedastic noise both FA and PCA succeed in recovering the size of the low rank
subspace. The likelihood with PCA is higher than FA in this case. However PCA fails and overestimates the rank
when heteroscedastic noise is present. Under appropriate circumstances the low rank models are more likely than
shrinkage models.
The automatic estimation from Automatic Choice of Dimensionality for PCA. NIPS 2000: 598-604 by Thomas P.
Minka is also compared.

•
Out:
best n_components by PCA CV = 10

best n_components by FactorAnalysis CV = 10
best n_components by PCA MLE = 10
best n_components by PCA CV = 35
best n_components by FactorAnalysis CV = 10
best n_components by PCA MLE = 38
# Authors: Alexandre Gramfort

# Denis A. Engemann
import numpy as np
from sklearn.decomposition import PCA, FactorAnalysis

from sklearn.covariance import ShrunkCovariance, LedoitWolf

print(__doc__)
# #############################################################################
# Create the data
n_samples, n_features, rank = 1000, 50, 10

sigma = 1.
U, _, _ = linalg.svd(rng.randn(n_features, n_features))
X = np.dot(rng.randn(n_samples, rank), U[:, :rank].T)
# Adding homoscedastic noise

X_homo = X + sigma * rng.randn(n_samples, n_features)
# Adding heteroscedastic noise

sigmas = sigma * rng.rand(n_features) + sigma / 2.
X_hetero = X + rng.randn(n_samples, n_features) * sigmas
# #############################################################################
# Fit the models
n_components = np.arange(0, n_features, 5) # options for n_components
def compute_scores(X):
pca = PCA(svd_solver='full')
fa = FactorAnalysis()
pca_scores, fa_scores = [], []

for n in n_components:
pca.n_components = n
fa.n_components = n
pca_scores.append(np.mean(cross_val_score(pca, X, cv=5)))
fa_scores.append(np.mean(cross_val_score(fa, X, cv=5)))
return pca_scores, fa_scores
def shrunk_cov_score(X):
shrinkages = np.logspace(-2, 0, 30)
cv = GridSearchCV(ShrunkCovariance(), {'shrinkage': shrinkages}, cv=5)
return np.mean(cross_val_score(cv.fit(X).best_estimator_, X, cv=5))
def lw_score(X):
return np.mean(cross_val_score(LedoitWolf(), X, cv=5))
for X, title in [(X_homo, 'Homoscedastic Noise'),

(X_hetero, 'Heteroscedastic Noise')]:
pca_scores, fa_scores = compute_scores(X)
n_components_pca = n_components[np.argmax(pca_scores)]
n_components_fa = n_components[np.argmax(fa_scores)]
pca = PCA(svd_solver='full', n_components='mle')

pca.fit(X)
n_components_pca_mle = pca.n_components_

print("best n_components by PCA CV = %d" % n_components_pca)

print("best n_components by FactorAnalysis CV = %d" % n_components_fa)
print("best n_components by PCA MLE = %d" % n_components_pca_mle)
plt.figure()
plt.plot(n_components, pca_scores, 'b', label='PCA scores')
plt.plot(n_components, fa_scores, 'r', label='FA scores')
plt.axvline(rank, color='g', label='TRUTH: %d' % rank, linestyle='-')
plt.axvline(n_components_pca, color='b',
label='PCA CV: %d' % n_components_pca, linestyle='--')
plt.axvline(n_components_fa, color='r',
label='FactorAnalysis CV: %d' % n_components_fa,
linestyle='--')
plt.axvline(n_components_pca_mle, color='k',
label='PCA MLE: %d' % n_components_pca_mle, linestyle='--')
# compare with other covariance estimators

plt.axhline(shrunk_cov_score(X), color='violet',
label='Shrunk Covariance MLE', linestyle='-.')
plt.axhline(lw_score(X), color='orange',
label='LedoitWolf MLE' % n_components_pca_mle, linestyle='-.')
plt.xlabel('nb of components')
plt.ylabel('CV scores')
plt.legend(loc='lower right')
plt.title(title)
plt.show()
5.11.10 Sparse coding with a precomputed dictionary
Transform a signal as a sparse combination of Ricker wavelets. This example visually compares different sparse coding
methods using the sklearn.decomposition.SparseCoder estimator. The Ricker (also known as Mexican
hat or the second derivative of a Gaussian) is not a particularly good kernel to represent piecewise constant signals
like this one. It can therefore be seen how much adding different widths of atoms matters and it therefore motivates
learning the dictionary to best fit your type of signals.
The richer dictionary on the right is not larger in size, heavier subsampling is performed in order to stay on the same
order of magnitude.

print(__doc__)
import numpy as np
from sklearn.decomposition import SparseCoder
def ricker_function(resolution, center, width):

"""Discrete sub-sampled Ricker (Mexican hat) wavelet"""
x = np.linspace(0, resolution - 1, resolution)
x = ((2 / ((np.sqrt(3 * width) * np.pi ** 1 / 4)))
* (1 - ((x - center) ** 2 / width ** 2))
* np.exp((-(x - center) ** 2) / (2 * width ** 2)))
return x
def ricker_matrix(width, resolution, n_components):

"""Dictionary of Ricker (Mexican hat) wavelets"""
centers = np.linspace(0, resolution - 1, n_components)
D = np.empty((n_components, resolution))
for i, center in enumerate(centers):
D[i] = ricker_function(resolution, center, width)
D /= np.sqrt(np.sum(D ** 2, axis=1))[:, np.newaxis]
return D
resolution = 1024
subsampling = 3 # subsampling factor
width = 100
n_components = resolution // subsampling
# Compute a wavelet dictionary

D_fixed = ricker_matrix(width=width, resolution=resolution,
n_components=n_components)
D_multi = np.r_[tuple(ricker_matrix(width=w, resolution=resolution,

n_components=n_components // 5)
for w in (10, 50, 100, 500, 1000))]
# Generate a signal
y = np.linspace(0, resolution - 1, resolution)
first_quarter = y < resolution / 4
y[first_quarter] = 3.
y[np.logical_not(first_quarter)] = -1.
# List the different sparse coding methods in the following format:

# (title, transform_algorithm, transform_alpha, transform_n_nozero_coefs)
estimators = [('OMP', 'omp', None, 15, 'navy'),
('Lasso', 'lasso_cd', 2, None, 'turquoise'), ]
lw = 2
# Avoid FutureWarning about default value change when numpy >= 1.14
lstsq_rcond = None if LooseVersion(np.__version__) >= '1.14' else -1
for subplot, (D, title) in enumerate(zip((D_fixed, D_multi),
('fixed width', 'multiple widths'))):
plt.subplot(1, 2, subplot + 1)
plt.title('Sparse coding against %s dictionary' % title)
plt.plot(y, lw=lw, linestyle='--', label='Original signal')
# Do a wavelet approximation
for title, algo, alpha, n_nonzero, color in estimators:
coder = SparseCoder(dictionary=D, transform_n_nonzero_coefs=n_nonzero,
transform_alpha=alpha, transform_algorithm=algo)
x = coder.transform(y.reshape(1, -1))
density = len(np.flatnonzero(x))
x = np.ravel(np.dot(x, D))
squared_error = np.sum((y - x) ** 2)
plt.plot(x, color=color, lw=lw,
label='%s: %s nonzero coefs,\n%.2f error'
% (title, density, squared_error))
# Soft thresholding debiasing

coder = SparseCoder(dictionary=D, transform_algorithm='threshold',
transform_alpha=20)
x = coder.transform(y.reshape(1, -1))
_, idx = np.where(x != 0)
x[0, idx], _, _, _ = np.linalg.lstsq(D[idx, :].T, y, rcond=lstsq_rcond)
x = np.ravel(np.dot(x, D))
squared_error = np.sum((y - x) ** 2)
plt.plot(x, color='darkorange', lw=lw,
label='Thresholding w/ debiasing:\n%d nonzero coefs, %.2f error'
% (len(idx), squared_error))
plt.axis('tight')
plt.legend(shadow=False, loc='best')
plt.subplots_adjust(.04, .07, .97, .90, .09, .2)
plt.show()

5.11.11 Image denoising using dictionary learning
An example comparing the effect of reconstructing noisy fragments of a raccoon face image using firstly online
Dictionary Learning and various transform methods.
The dictionary is fitted on the distorted left half of the image, and subsequently used to reconstruct the right half. Note
that even better performance could be achieved by fitting to an undistorted (i.e. noiseless) image, but here we start
from the assumption that it is not available.
A common practice for evaluating the results of image denoising is by looking at the difference between the recon-
struction and the original image. If the reconstruction is perfect this will look like Gaussian noise.
It can be seen from the plots that the results of Orthogonal Matching Pursuit (OMP) with two non-zero coefficients is
a bit less biased than when keeping only one (the edges look less prominent). It is in addition closer from the ground
truth in Frobenius norm.
The result of Least Angle Regression is much more strongly biased: the difference is reminiscent of the local intensity
value of the original image.
Thresholding is clearly not useful for denoising, but it is here to show that it can produce a suggestive output with
very high speed, and thus be useful for other tasks such as object classification, where performance is not necessarily
related to visualisation.



•
Out:
Distorting image...
Extracting reference patches...
done in 0.06s.
Learning the dictionary...
done in 8.09s.
Extracting noisy patches...
done in 0.03s.
Orthogonal Matching Pursuit
1 atom...
done in 5.24s.
Orthogonal Matching Pursuit
2 atoms...
done in 11.23s.
Least-angle regression
5 atoms...
done in 59.46s.
Thresholding
alpha=0.1...
done in 0.84s.
print(__doc__)

import numpy as np
import scipy as sp
from sklearn.decomposition import MiniBatchDictionaryLearning

from sklearn.feature_extraction.image import extract_patches_2d

from sklearn.feature_extraction.image import reconstruct_from_patches_2d
try: # SciPy >= 0.16 have face in misc

from scipy.misc import face
face = face(gray=True)
except ImportError:
face = sp.face(gray=True)
# Convert from uint8 representation with values between 0 and 255 to

# a floating point representation with values between 0 and 1.
face = face / 255.
# downsample for higher speed

face = face[::2, ::2] + face[1::2, ::2] + face[::2, 1::2] + face[1::2, 1::2]
face /= 4.0
height, width = face.shape
# Distort the right half of the image

print('Distorting image...')
distorted = face.copy()
distorted[:, width // 2:] += 0.075 * np.random.randn(height, width // 2)
# Extract all reference patches from the left half of the image
print('Extracting reference patches...')
t0 = time()
patch_size = (7, 7)
data = extract_patches_2d(distorted[:, :width // 2], patch_size)
data = data.reshape(data.shape[0], -1)
data -= np.mean(data, axis=0)
data /= np.std(data, axis=0)
print('done in %.2fs.' % (time() - t0))
# #############################################################################
# Learn the dictionary from reference patches
print('Learning the dictionary...')

t0 = time()
dico = MiniBatchDictionaryLearning(n_components=100, alpha=1, n_iter=500)
V = dico.fit(data).components_
dt = time() - t0
for i, comp in enumerate(V[:100]):
plt.imshow(comp.reshape(patch_size), cmap=plt.cm.gray_r,
plt.xticks(())
plt.yticks(())
plt.suptitle('Dictionary learned from face patches\n' +
'Train time %.1fs on %d patches' % (dt, len(data)),
fontsize=16)
# #############################################################################

# Display the distorted image
def show_with_diff(image, reference, title):

"""Helper function to display denoising"""
plt.title('Image')
plt.imshow(image, vmin=0, vmax=1, cmap=plt.cm.gray,
plt.xticks(())
plt.yticks(())
difference = image - reference
plt.title('Difference (norm: %.2f)' % np.sqrt(np.sum(difference ** 2)))

plt.imshow(difference, vmin=-0.5, vmax=0.5, cmap=plt.cm.PuOr,
plt.xticks(())
plt.yticks(())
plt.suptitle(title, size=16)
show_with_diff(distorted, face, 'Distorted image')
# #############################################################################
# Extract noisy patches and reconstruct them using the dictionary
print('Extracting noisy patches... ')

t0 = time()
data = extract_patches_2d(distorted[:, width // 2:], patch_size)
data = data.reshape(data.shape[0], -1)
intercept = np.mean(data, axis=0)
data -= intercept
print('done in %.2fs.' % (time() - t0))
transform_algorithms = [
('Orthogonal Matching Pursuit\n1 atom', 'omp',
{'transform_n_nonzero_coefs': 1}),
('Orthogonal Matching Pursuit\n2 atoms', 'omp',
('Least-angle regression\n5 atoms', 'lars',
('Thresholding\n alpha=0.1', 'threshold', {'transform_alpha': .1})]
reconstructions = {}
for title, transform_algorithm, kwargs in transform_algorithms:
print(title + '...')
reconstructions[title] = face.copy()
t0 = time()
dico.set_params(transform_algorithm=transform_algorithm, **kwargs)
code = dico.transform(data)
patches = np.dot(code, V)
patches += intercept
patches = patches.reshape(len(data), *patch_size)
if transform_algorithm == 'threshold':
patches -= patches.min()
patches /= patches.max()

reconstructions[title][:, width // 2:] = reconstruct_from_patches_2d(

patches, (height, width // 2))
dt = time() - t0
show_with_diff(reconstructions[title], face,
title + ' (time: %.1fs)' % dt)
plt.show()
5.11.12 Faces dataset decompositions
This example applies to olivetti_faces different unsupervised matrix decomposition (dimension reduction) methods
from the module sklearn.decomposition (see the documentation chapter Decomposing signals in components
(matrix factorization problems)) .












•
Out:
Dataset consists of 400 faces

Extracting the top 6 Eigenfaces - PCA using randomized SVD...
done in 0.058s
Extracting the top 6 Non-negative components - NMF...
done in 0.196s
Extracting the top 6 Independent components - FastICA...
done in 0.164s
Extracting the top 6 Sparse comp. - MiniBatchSparsePCA...
done in 0.849s
Extracting the top 6 MiniBatchDictionaryLearning...
done in 4.693s
Extracting the top 6 Cluster centers - MiniBatchKMeans...
done in 0.201s
Extracting the top 6 Factor Analysis components - FA...
done in 0.089s
Extracting the top 6 Dictionary learning...
done in 1.498s
Extracting the top 6 Dictionary learning - positive dictionary...
done in 1.575s
Extracting the top 6 Dictionary learning - positive code...
done in 0.709s
Extracting the top 6 Dictionary learning - positive dictionary & code...
done in 0.371s

print(__doc__)
# Authors: Vlad Niculae, Alexandre Gramfort

import logging
from numpy.random import RandomState


from sklearn import decomposition

logging.basicConfig(level=logging.INFO,
format='%(asctime)s %(levelname)s %(message)s')
n_row, n_col = 2, 3
n_components = n_row * n_col
image_shape = (64, 64)
rng = RandomState(0)
# #############################################################################
# Load faces data
dataset = fetch_olivetti_faces(shuffle=True, random_state=rng)
faces = dataset.data
n_samples, n_features = faces.shape
# global centering
faces_centered = faces - faces.mean(axis=0)
# local centering
faces_centered -= faces_centered.mean(axis=1).reshape(n_samples, -1)
print("Dataset consists of %d faces" % n_samples)
def plot_gallery(title, images, n_col=n_col, n_row=n_row, cmap=plt.cm.gray):

plt.figure(figsize=(2. * n_col, 2.26 * n_row))
plt.suptitle(title, size=16)
for i, comp in enumerate(images):
vmax = max(comp.max(), -comp.min())
plt.imshow(comp.reshape(image_shape), cmap=cmap,
interpolation='nearest',
vmin=-vmax, vmax=vmax)
plt.xticks(())
plt.yticks(())
plt.subplots_adjust(0.01, 0.05, 0.99, 0.93, 0.04, 0.)
# #############################################################################
# List of the different estimators, whether to center and transpose the
# problem, and whether the transformer uses the clustering API.
estimators = [
('Eigenfaces - PCA using randomized SVD',
decomposition.PCA(n_components=n_components, svd_solver='randomized',

whiten=True),
True),
('Non-negative components - NMF',

decomposition.NMF(n_components=n_components, init='nndsvda', tol=5e-3),
False),
('Independent components - FastICA',

decomposition.FastICA(n_components=n_components, whiten=True),
True),
('Sparse comp. - MiniBatchSparsePCA',

decomposition.MiniBatchSparsePCA(n_components=n_components, alpha=0.8,
n_iter=100, batch_size=3,
random_state=rng,
normalize_components=True),
True),
('MiniBatchDictionaryLearning',
decomposition.MiniBatchDictionaryLearning(n_components=15, alpha=0.1,
random_state=rng),
True),
('Cluster centers - MiniBatchKMeans',

MiniBatchKMeans(n_clusters=n_components, tol=1e-3, batch_size=20,
max_iter=50, random_state=rng),
True),
('Factor Analysis components - FA',

decomposition.FactorAnalysis(n_components=n_components, max_iter=2),
True),
]
# #############################################################################
# Plot a sample of the input data
plot_gallery("First centered Olivetti faces", faces_centered[:n_components])
# #############################################################################
# Do the estimation and plot it
for name, estimator, center in estimators:

print("Extracting the top %d %s..." % (n_components, name))
t0 = time()
data = faces
if center:
data = faces_centered
estimator.fit(data)
train_time = (time() - t0)
print("done in %0.3fs" % train_time)
if hasattr(estimator, 'cluster_centers_'):
components_ = estimator.cluster_centers_
else:
components_ = estimator.components_
# Plot an image representing the pixelwise variance provided by the

# estimator e.g its noise_variance_ attribute. The Eigenfaces estimator,

# via the PCA decomposition, also provides a scalar noise_variance_
# (the mean of pixelwise variance) that cannot be displayed as an image
# so we skip it.
if (hasattr(estimator, 'noise_variance_') and
estimator.noise_variance_.ndim > 0): # Skip the Eigenfaces case
plot_gallery("Pixelwise variance",
estimator.noise_variance_.reshape(1, -1), n_col=1,
n_row=1)
plot_gallery('%s - Train time %.1fs' % (name, train_time),
components_[:n_components])
plt.show()
# #############################################################################
# Various positivity constraints applied to dictionary learning.
estimators = [
('Dictionary learning',
random_state=rng),
True),
('Dictionary learning - positive dictionary',
random_state=rng,
positive_dict=True),
True),
('Dictionary learning - positive code',
random_state=rng,
positive_code=True),
True),
('Dictionary learning - positive dictionary & code',
random_state=rng,
positive_dict=True,
positive_code=True),
True),
]
# #############################################################################
# Plot a sample of the input data
plot_gallery("First centered Olivetti faces", faces_centered[:n_components],

cmap=plt.cm.RdBu)
# #############################################################################
# Do the estimation and plot it
for name, estimator, center in estimators:

print("Extracting the top %d %s..." % (n_components, name))
t0 = time()
data = faces
if center:

data = faces_centered
estimator.fit(data)
train_time = (time() - t0)
print("done in %0.3fs" % train_time)
components_ = estimator.components_
plot_gallery(name, components_[:n_components], cmap=plt.cm.RdBu)
plt.show()
5.12 Ensemble methods
Examples concerning the sklearn.ensemble module.
5.12.1 Decision Tree Regression with AdaBoost
A decision tree is boosted using the AdaBoost.R21 algorithm on a 1D sinusoidal dataset with a small amount of
Gaussian noise. 299 boosts (300 decision trees) is compared with a single decision tree regressor. As the number of
boosts is increased the regressor can fit more detail.
1
8. Drucker, “Improving Regressors using Boosting Techniques”, 1997.
5.12. Ensemble methods 975

print(__doc__)
# Author: Noel Dawe <noel.dawe@gmail.com>

#
# importing necessary libraries

import numpy as np
from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import AdaBoostRegressor
# Create the dataset

X = np.linspace(0, 6, 100)[:, np.newaxis]
y = np.sin(X).ravel() + np.sin(6 * X).ravel() + rng.normal(0, 0.1, X.shape[0])

regr_1 = DecisionTreeRegressor(max_depth=4)
regr_2 = AdaBoostRegressor(DecisionTreeRegressor(max_depth=4),
n_estimators=300, random_state=rng)
regr_1.fit(X, y)
regr_2.fit(X, y)

# Predict
y_1 = regr_1.predict(X)
y_2 = regr_2.predict(X)
# Plot the results

plt.figure()
plt.scatter(X, y, c="k", label="training samples")
plt.plot(X, y_1, c="g", label="n_estimators=1", linewidth=2)
plt.plot(X, y_2, c="r", label="n_estimators=300", linewidth=2)
plt.xlabel("data")
plt.ylabel("target")
plt.title("Boosted Decision Tree Regression")
plt.legend()
plt.show()
5.12.2 Pixel importances with a parallel forest of trees
This example shows the use of forests of trees to evaluate the importance of the pixels in an image classification task
(faces). The hotter the pixel, the more important.
The code below also illustrates how the construction and the computation of the predictions can be parallelized within
multiple jobs.

Out:
Fitting ExtraTreesClassifier on faces data with 1 cores...

done in 0.924s
print(__doc__)


from sklearn.ensemble import ExtraTreesClassifier
# Number of cores to use to perform parallel fitting of the forest model

n_jobs = 1
# Load the faces dataset

data = fetch_olivetti_faces()
X = data.images.reshape((len(data.images), -1))
y = data.target

mask = y < 5 # Limit to 5 classes

X = X[mask]
y = y[mask]
# Build a forest and compute the pixel importances

print("Fitting ExtraTreesClassifier on faces data with %d cores..." % n_jobs)
t0 = time()
forest = ExtraTreesClassifier(n_estimators=1000,
max_features=128,
n_jobs=n_jobs,
random_state=0)
forest.fit(X, y)
importances = forest.feature_importances_
importances = importances.reshape(data.images[0].shape)
# Plot pixel importances

plt.matshow(importances, cmap=plt.cm.hot)
plt.title("Pixel importances with forests of trees")
plt.show()
5.12.3 Feature importances with forests of trees
This examples shows the use of forests of trees to evaluate the importance of features on an artificial classification
task. The red bars are the feature importances of the forest, along with their inter-trees variability.
As expected, the plot suggests that 3 features are informative, while the remaining are not.

Out:
Feature ranking:
1. feature 1 (0.295902)
2. feature 2 (0.208351)
3. feature 0 (0.177632)
4. feature 3 (0.047121)
5. feature 6 (0.046303)
6. feature 8 (0.046013)
7. feature 7 (0.045575)
8. feature 4 (0.044614)
9. feature 9 (0.044577)
10. feature 5 (0.043912)
print(__doc__)
import numpy as np

from sklearn.ensemble import ExtraTreesClassifier

# Build a classification task using 3 informative features

X, y = make_classification(n_samples=1000,
n_features=10,
n_informative=3,
n_redundant=0,
n_repeated=0,
n_classes=2,
random_state=0,
shuffle=False)
# Build a forest and compute the feature importances

forest = ExtraTreesClassifier(n_estimators=250,
random_state=0)
forest.fit(X, y)
importances = forest.feature_importances_
std = np.std([tree.feature_importances_ for tree in forest.estimators_],
axis=0)
indices = np.argsort(importances)[::-1]
# Print the feature ranking

print("Feature ranking:")
for f in range(X.shape[1]):
print("%d. feature %d (%f)" % (f + 1, indices[f], importances[indices[f]]))
# Plot the feature importances of the forest

plt.figure()
plt.title("Feature importances")
plt.bar(range(X.shape[1]), importances[indices],
color="r", yerr=std[indices], align="center")
plt.xticks(range(X.shape[1]), indices)
plt.xlim([-1, X.shape[1]])
plt.show()
5.12.4 IsolationForest example
An example using sklearn.ensemble.IsolationForest for anomaly detection.

The IsolationForest ‘isolates’ observations by randomly selecting a feature and then randomly selecting a split value
between the maximum and minimum values of the selected feature.
Since recursive partitioning can be represented by a tree structure, the number of splittings required to isolate a sample
is equivalent to the path length from the root node to the terminating node.
This path length, averaged over a forest of such random trees, is a measure of normality and our decision function.
Random partitioning produces noticeable shorter paths for anomalies. Hence, when a forest of random trees collec-
tively produce shorter path lengths for particular samples, they are highly likely to be anomalies.

print(__doc__)
import numpy as np
from sklearn.ensemble import IsolationForest
# Generate train data

X = 0.3 * rng.randn(100, 2)
X_train = np.r_[X + 2, X - 2]
# Generate some regular novel observations
X = 0.3 * rng.randn(20, 2)
X_test = np.r_[X + 2, X - 2]
# Generate some abnormal novel observations
X_outliers = rng.uniform(low=-4, high=4, size=(20, 2))
# fit the model

clf = IsolationForest(behaviour='new', max_samples=100,
random_state=rng, contamination='auto')
clf.fit(X_train)
y_pred_train = clf.predict(X_train)
y_pred_test = clf.predict(X_test)
y_pred_outliers = clf.predict(X_outliers)
# plot the line, the samples, and the nearest vectors to the plane

xx, yy = np.meshgrid(np.linspace(-5, 5, 50), np.linspace(-5, 5, 50))

plt.title("IsolationForest")
plt.contourf(xx, yy, Z, cmap=plt.cm.Blues_r)
b1 = plt.scatter(X_train[:, 0], X_train[:, 1], c='white',

b2 = plt.scatter(X_test[:, 0], X_test[:, 1], c='green',
c = plt.scatter(X_outliers[:, 0], X_outliers[:, 1], c='red',
plt.axis('tight')
plt.xlim((-5, 5))
plt.ylim((-5, 5))
plt.legend([b1, b2, c],
["training observations",
"new regular observations", "new abnormal observations"],
loc="upper left")
plt.show()
5.12.5 Plot the decision boundaries of a VotingClassifier
Plot the decision boundaries of a VotingClassifier for two features of the Iris dataset.
Plot the class probabilities of the first sample in a toy dataset predicted by three different classifiers and averaged by
the VotingClassifier.
First, three exemplary classifiers are initialized (DecisionTreeClassifier, KNeighborsClassifier, and SVC) and used to
initialize a soft-voting VotingClassifier with weights [2, 1, 2], which means that the predicted probabilities of the
DecisionTreeClassifier and SVC count 5 times as much as the weights of the KNeighborsClassifier classifier when the
averaged probability is calculated.

print(__doc__)
from itertools import product
import numpy as np

from sklearn.ensemble import VotingClassifier
# Loading some example data

X = iris.data[:, [0, 2]]
y = iris.target
# Training classifiers
clf1 = DecisionTreeClassifier(max_depth=4)
clf2 = KNeighborsClassifier(n_neighbors=7)
clf3 = SVC(gamma=.1, kernel='rbf', probability=True)
eclf = VotingClassifier(estimators=[('dt', clf1), ('knn', clf2),

('svc', clf3)],
voting='soft', weights=[2, 1, 2])
clf1.fit(X, y)
clf2.fit(X, y)
clf3.fit(X, y)
eclf.fit(X, y)
# Plotting decision regions

x_min, x_max = X[:, 0].min() - 1, X[:, 0].max() + 1
y_min, y_max = X[:, 1].min() - 1, X[:, 1].max() + 1
xx, yy = np.meshgrid(np.arange(x_min, x_max, 0.1),
np.arange(y_min, y_max, 0.1))
f, axarr = plt.subplots(2, 2, sharex='col', sharey='row', figsize=(10, 8))
for idx, clf, tt in zip(product([0, 1], [0, 1]),

[clf1, clf2, clf3, eclf],
['Decision Tree (depth=4)', 'KNN (k=7)',
'Kernel SVM', 'Soft Voting']):
Z = clf.predict(np.c_[xx.ravel(), yy.ravel()])
axarr[idx[0], idx[1]].contourf(xx, yy, Z, alpha=0.4)

axarr[idx[0], idx[1]].scatter(X[:, 0], X[:, 1], c=y,
axarr[idx[0], idx[1]].set_title(tt)
plt.show()
5.12.6 Comparing random forests and the multi-output meta estimator
An example to compare multi-output regression with random forest and the multioutput.MultiOutputRegressor meta-
estimator.
This example illustrates the use of the multioutput.MultiOutputRegressor meta-estimator to perform multi-output re-
gression. A random forest regressor is used, which supports multi-output regression natively, so the results can be
compared.
The random forest regressor will only ever predict values within the range of observations or closer to zero for each of
the targets. As a result the predictions are biased towards the centre of the circle.
Using a single underlying feature the model learns both the x and y coordinate as output.

print(__doc__)
# Author: Tim Head <betatim@gmail.com>

#
import numpy as np
from sklearn.ensemble import RandomForestRegressor
from sklearn.multioutput import MultiOutputRegressor
# Create a random dataset

X = np.sort(200 * rng.rand(600, 1) - 100, axis=0)
y = np.array([np.pi * np.sin(X).ravel(), np.pi * np.cos(X).ravel()]).T
y += (0.5 - rng.rand(*y.shape))

X, y, train_size=400, test_size=200, random_state=4)
max_depth = 30
regr_multirf = MultiOutputRegressor(RandomForestRegressor(n_estimators=100,
max_depth=max_depth,
random_state=0))

regr_multirf.fit(X_train, y_train)
regr_rf = RandomForestRegressor(n_estimators=100, max_depth=max_depth,

random_state=2)
regr_rf.fit(X_train, y_train)
# Predict on new data

y_multirf = regr_multirf.predict(X_test)
y_rf = regr_rf.predict(X_test)
# Plot the results

plt.figure()
s = 50
a = 0.4
plt.scatter(y_test[:, 0], y_test[:, 1], edgecolor='k',
c="navy", s=s, marker="s", alpha=a, label="Data")
plt.scatter(y_multirf[:, 0], y_multirf[:, 1], edgecolor='k',
c="cornflowerblue", s=s, alpha=a,
label="Multi RF score=%.2f" % regr_multirf.score(X_test, y_test))
plt.scatter(y_rf[:, 0], y_rf[:, 1], edgecolor='k',
c="c", s=s, marker="^", alpha=a,
label="RF score=%.2f" % regr_rf.score(X_test, y_test))
plt.xlim([-6, 6])
plt.ylim([-6, 6])
plt.xlabel("target 1")
plt.ylabel("target 2")
plt.title("Comparing random forests and the multi-output meta estimator")
plt.legend()
plt.show()
5.12.7 Prediction Intervals for Gradient Boosting Regression
This example shows how quantile regression can be used to create prediction intervals.

import numpy as np
from sklearn.ensemble import GradientBoostingRegressor
np.random.seed(1)
def f(x):
"""The function to predict."""
return x * np.sin(x)
#----------------------------------------------------------------------
# First the noiseless case
X = np.atleast_2d(np.random.uniform(0, 10.0, size=100)).T
X = X.astype(np.float32)
# Observations
y = f(X).ravel()
dy = 1.5 + 1.0 * np.random.random(y.shape)

noise = np.random.normal(0, dy)
y += noise
y = y.astype(np.float32)
# Mesh the input space for evaluations of the real function, the prediction and

# its MSE
xx = np.atleast_2d(np.linspace(0, 10, 1000)).T
xx = xx.astype(np.float32)
alpha = 0.95
clf = GradientBoostingRegressor(loss='quantile', alpha=alpha,

n_estimators=250, max_depth=3,
learning_rate=.1, min_samples_leaf=9,
min_samples_split=9)
clf.fit(X, y)
# Make the prediction on the meshed x-axis

y_upper = clf.predict(xx)
clf.set_params(alpha=1.0 - alpha)
clf.fit(X, y)

y_lower = clf.predict(xx)
clf.set_params(loss='ls')
clf.fit(X, y)

y_pred = clf.predict(xx)
# Plot the function, the prediction and the 90% confidence interval based on
# the MSE
fig = plt.figure()
plt.plot(xx, f(xx), 'g:', label=r'$f(x) = x\,\sin(x)$')
plt.plot(X, y, 'b.', markersize=10, label=u'Observations')
plt.plot(xx, y_pred, 'r-', label=u'Prediction')
plt.plot(xx, y_upper, 'k-')
plt.plot(xx, y_lower, 'k-')
plt.fill(np.concatenate([xx, xx[::-1]]),
np.concatenate([y_upper, y_lower[::-1]]),
alpha=.5, fc='b', ec='None', label='90% prediction interval')
plt.xlabel('$x$')
plt.ylabel('$f(x)$')
plt.ylim(-10, 20)
plt.show()
5.12.8 Gradient Boosting regularization
Illustration of the effect of different regularization strategies for Gradient Boosting. The example is taken from Hastie
et al 20091 .
1 T. Hastie, R. Tibshirani and J. Friedman, “Elements of Statistical Learning Ed. 2”, Springer, 2009.

The loss function used is binomial deviance. Regularization via shrinkage (learning_rate < 1.0) improves
performance considerably. In combination with shrinkage, stochastic gradient boosting (subsample < 1.0) can
produce more accurate models by reducing the variance via bagging. Subsampling without shrinkage usually does
poorly. Another strategy to reduce the variance is by subsampling the features analogous to the random splits in
Random Forests (via the max_features parameter).
print(__doc__)
# Author: Peter Prettenhofer <peter.prettenhofer@gmail.com>

#
import numpy as np
from sklearn import ensemble

X, y = datasets.make_hastie_10_2(n_samples=12000, random_state=1)
# map labels from {-1, 1} to {0, 1}

labels, y = np.unique(y, return_inverse=True)
X_train, X_test = X[:2000], X[2000:]

y_train, y_test = y[:2000], y[2000:]
original_params = {'n_estimators': 1000, 'max_leaf_nodes': 4, 'max_depth': None,

˓→'random_state': 2,
'min_samples_split': 5}
plt.figure()
for label, color, setting in [('No shrinkage', 'orange',

{'learning_rate': 1.0, 'subsample': 1.0}),
('learning_rate=0.1', 'turquoise',
('subsample=0.5', 'blue',
('learning_rate=0.1, subsample=0.5', 'gray',
('learning_rate=0.1, max_features=2', 'magenta',
{'learning_rate': 0.1, 'max_features': 2})]:
params = dict(original_params)
params.update(setting)
clf = ensemble.GradientBoostingClassifier(**params)
# compute test set deviance

test_deviance = np.zeros((params['n_estimators'],), dtype=np.float64)
for i, y_pred in enumerate(clf.staged_decision_function(X_test)):

# clf.loss_ assumes that y_test[i] in {0, 1}
test_deviance[i] = clf.loss_(y_test, y_pred)
plt.plot((np.arange(test_deviance.shape[0]) + 1)[::5], test_deviance[::5],

'-', color=color, label=label)
plt.xlabel('Boosting Iterations')
plt.ylabel('Test Set Deviance')
plt.show()
5.12.9 Plot class probabilities calculated by the VotingClassifier
Plot the class probabilities of the first sample in a toy dataset predicted by three different classifiers and averaged by
the VotingClassifier.
First, three examplary classifiers are initialized (LogisticRegression, GaussianNB, and RandomForestClassifier) and
used to initialize a soft-voting VotingClassifier with weights [1, 1, 5], which means that the predicted probabilities of
the RandomForestClassifier count 5 times as much as the weights of the other classifiers when the averaged probability
is calculated.
To visualize the probability weighting, we fit each classifier on the training set and plot the predicted class probabilities

for the first sample in this example dataset.
print(__doc__)
import numpy as np

from sklearn.ensemble import VotingClassifier
clf1 = LogisticRegression(solver='lbfgs', max_iter=1000, random_state=123)

clf2 = RandomForestClassifier(n_estimators=100, random_state=123)
clf3 = GaussianNB()
X = np.array([[-1.0, -1.0], [-1.2, -1.4], [-3.4, -2.2], [1.1, 1.2]])
y = np.array([1, 1, 2, 2])
eclf = VotingClassifier(estimators=[('lr', clf1), ('rf', clf2), ('gnb', clf3)],

voting='soft',
weights=[1, 1, 5])
# predict class probabilities for all classifiers

probas = [c.fit(X, y).predict_proba(X) for c in (clf1, clf2, clf3, eclf)]
# get class probabilities for the first sample in the dataset

class1_1 = [pr[0, 0] for pr in probas]

class2_1 = [pr[0, 1] for pr in probas]
# plotting
N = 4 # number of groups
ind = np.arange(N) # group positions
width = 0.35 # bar width
fig, ax = plt.subplots()
# bars for classifier 1-3

p1 = ax.bar(ind, np.hstack(([class1_1[:-1], [0]])), width,
color='green', edgecolor='k')
p2 = ax.bar(ind + width, np.hstack(([class2_1[:-1], [0]])), width,
color='lightgreen', edgecolor='k')
# bars for VotingClassifier

p3 = ax.bar(ind, [0, 0, 0, class1_1[-1]], width,
color='blue', edgecolor='k')
p4 = ax.bar(ind + width, [0, 0, 0, class2_1[-1]], width,
color='steelblue', edgecolor='k')
# plot annotations
plt.axvline(2.8, color='k', linestyle='dashed')
ax.set_xticks(ind + width)
ax.set_xticklabels(['LogisticRegression\nweight 1',
'GaussianNB\nweight 1',
'RandomForestClassifier\nweight 5',
'VotingClassifier\n(average probabilities)'],
rotation=40,
ha='right')
plt.ylim([0, 1])
plt.title('Class probabilities for sample 1 by different classifiers')
plt.legend([p1[0], p2[0]], ['class 1', 'class 2'], loc='upper left')
plt.tight_layout()
plt.show()
5.12.10 Gradient Boosting regression
Demonstrate Gradient Boosting on the Boston housing dataset.

This example fits a Gradient Boosting model with least squares loss and 500 regression trees of depth 4.

Out:
MSE: 6.5493
print(__doc__)

#
import numpy as np

# #############################################################################
# Load data
boston = datasets.load_boston()
X, y = shuffle(boston.data, boston.target, random_state=13)
offset = int(X.shape[0] * 0.9)
X_train, y_train = X[:offset], y[:offset]
X_test, y_test = X[offset:], y[offset:]
# #############################################################################
params = {'n_estimators': 500, 'max_depth': 4, 'min_samples_split': 2,
'learning_rate': 0.01, 'loss': 'ls'}

clf = ensemble.GradientBoostingRegressor(**params)
mse = mean_squared_error(y_test, clf.predict(X_test))
print("MSE: %.4f" % mse)
# #############################################################################
# Plot training deviance
# compute test set deviance

test_score = np.zeros((params['n_estimators'],), dtype=np.float64)
for i, y_pred in enumerate(clf.staged_predict(X_test)):

test_score[i] = clf.loss_(y_test, y_pred)
plt.title('Deviance')
plt.plot(np.arange(params['n_estimators']) + 1, clf.train_score_, 'b-',
label='Training Set Deviance')
plt.plot(np.arange(params['n_estimators']) + 1, test_score, 'r-',
label='Test Set Deviance')
plt.legend(loc='upper right')
plt.xlabel('Boosting Iterations')
plt.ylabel('Deviance')
# #############################################################################
# Plot feature importance
feature_importance = clf.feature_importances_
# make importances relative to max importance
feature_importance = 100.0 * (feature_importance / feature_importance.max())
sorted_idx = np.argsort(feature_importance)
pos = np.arange(sorted_idx.shape[0]) + .5
plt.barh(pos, feature_importance[sorted_idx], align='center')
plt.yticks(pos, boston.feature_names[sorted_idx])
plt.xlabel('Relative Importance')
plt.title('Variable Importance')
plt.show()
5.12.11 Two-class AdaBoost
This example fits an AdaBoosted decision stump on a non-linearly separable classification dataset composed of two
“Gaussian quantiles” clusters (see sklearn.datasets.make_gaussian_quantiles) and plots the decision
boundary and decision scores. The distributions of decision scores are shown separately for samples of class A and B.
The predicted class label for each sample is determined by the sign of the decision score. Samples with decision scores
greater than zero are classified as B, and are otherwise classified as A. The magnitude of a decision score determines
the degree of likeness with the predicted class label. Additionally, a new dataset could be constructed containing a
desired purity of class B, for example, by only selecting samples with a decision score above some value.

print(__doc__)

#
import numpy as np
from sklearn.ensemble import AdaBoostClassifier

# Construct dataset
X1, y1 = make_gaussian_quantiles(cov=2.,
n_samples=200, n_features=2,
n_classes=2, random_state=1)
X2, y2 = make_gaussian_quantiles(mean=(3, 3), cov=1.5,
n_samples=300, n_features=2,
X = np.concatenate((X1, X2))
y = np.concatenate((y1, - y2 + 1))
# Create and fit an AdaBoosted decision tree

bdt = AdaBoostClassifier(DecisionTreeClassifier(max_depth=1),
algorithm="SAMME",
n_estimators=200)
bdt.fit(X, y)
plot_colors = "br"
plot_step = 0.02
class_names = "AB"

# Plot the decision boundaries

plt.subplot(121)
xx, yy = np.meshgrid(np.arange(x_min, x_max, plot_step),
np.arange(y_min, y_max, plot_step))
Z = bdt.predict(np.c_[xx.ravel(), yy.ravel()])
cs = plt.contourf(xx, yy, Z, cmap=plt.cm.Paired)
plt.axis("tight")

for i, n, c in zip(range(2), class_names, plot_colors):
idx = np.where(y == i)
plt.scatter(X[idx, 0], X[idx, 1],
c=c, cmap=plt.cm.Paired,
s=20, edgecolor='k',
label="Class %s" % n)
plt.xlabel('x')
plt.ylabel('y')
plt.title('Decision Boundary')
# Plot the two-class decision scores

twoclass_output = bdt.decision_function(X)
plot_range = (twoclass_output.min(), twoclass_output.max())
plt.subplot(122)
for i, n, c in zip(range(2), class_names, plot_colors):
plt.hist(twoclass_output[y == i],
bins=10,
range=plot_range,
facecolor=c,
label='Class %s' % n,
alpha=.5,
edgecolor='k')
x1, x2, y1, y2 = plt.axis()
plt.axis((x1, x2, y1, y2 * 1.2))
plt.ylabel('Samples')
plt.xlabel('Score')
plt.title('Decision Scores')
plt.tight_layout()
plt.subplots_adjust(wspace=0.35)
plt.show()

5.12.12 OOB Errors for Random Forests
The RandomForestClassifier is trained using bootstrap aggregation, where each new tree is fit from a boot-
strap sample of the training observations 𝑧𝑖 = (𝑥𝑖 , 𝑦𝑖 ). The out-of-bag (OOB) error is the average error for each 𝑧𝑖
calculated using predictions from the trees that do not contain 𝑧𝑖 in their respective bootstrap sample. This allows the
RandomForestClassifier to be fit and validated whilst being trained1 .
The example below demonstrates how the OOB error can be measured at the addition of each new tree during train-
ing. The resulting plot allows a practitioner to approximate a suitable value of n_estimators at which the error
stabilizes.
from collections import OrderedDict

# Author: Kian Ho <hui.kian.ho@gmail.com>

# Gilles Louppe <g.louppe@gmail.com>
#
# License: BSD 3 Clause
print(__doc__)
1 T. Hastie, R. Tibshirani and J. Friedman, “Elements of Statistical Learning Ed. 2”, p592-593, Springer, 2009.

RANDOM_STATE = 123
# Generate a binary classification dataset.

X, y = make_classification(n_samples=500, n_features=25,
n_clusters_per_class=1, n_informative=15,
random_state=RANDOM_STATE)
# NOTE: Setting the `warm_start` construction parameter to `True` disables

# support for parallelized ensembles but is necessary for tracking the OOB
# error trajectory during training.
ensemble_clfs = [
("RandomForestClassifier, max_features='sqrt'",
RandomForestClassifier(n_estimators=100,
warm_start=True, oob_score=True,
max_features="sqrt",
random_state=RANDOM_STATE)),
("RandomForestClassifier, max_features='log2'",
warm_start=True, max_features='log2',
oob_score=True,
random_state=RANDOM_STATE)),
("RandomForestClassifier, max_features=None",
warm_start=True, max_features=None,
oob_score=True,
random_state=RANDOM_STATE))
]
# Map a classifier name to a list of (<n_estimators>, <error rate>) pairs.

error_rate = OrderedDict((label, []) for label, _ in ensemble_clfs)
# Range of `n_estimators` values to explore.

min_estimators = 15
max_estimators = 175
for label, clf in ensemble_clfs:

for i in range(min_estimators, max_estimators + 1):
clf.set_params(n_estimators=i)
clf.fit(X, y)
# Record the OOB error for each `n_estimators=i` setting.

oob_error = 1 - clf.oob_score_
error_rate[label].append((i, oob_error))
# Generate the "OOB error rate" vs. "n_estimators" plot.

for label, clf_err in error_rate.items():
xs, ys = zip(*clf_err)
plt.plot(xs, ys, label=label)
plt.xlim(min_estimators, max_estimators)
plt.xlabel("n_estimators")
plt.ylabel("OOB error rate")
plt.show()

5.12.13 Hashing feature transformation using Totally Random Trees
RandomTreesEmbedding provides a way to map data to a very high-dimensional, sparse representation, which might
be beneficial for classification. The mapping is completely unsupervised and very efficient.
This example visualizes the partitions given by several trees and shows how the transformation can also be used for
non-linear dimensionality reduction or non-linear classification.
Points that are neighboring often share the same leaf of a tree and therefore share large parts of their hashed repre-
sentation. This allows to separate two concentric circles simply based on the principal components of the transformed
data with truncated SVD.
In high-dimensional spaces, linear classifiers often achieve excellent accuracy. For sparse binary data, BernoulliNB is
particularly well-suited. The bottom row compares the decision boundary obtained by BernoulliNB in the transformed
space with an ExtraTreesClassifier forests learned on the original data.

import numpy as np

from sklearn.ensemble import RandomTreesEmbedding, ExtraTreesClassifier
from sklearn.naive_bayes import BernoulliNB
# make a synthetic dataset

X, y = make_circles(factor=0.5, random_state=0, noise=0.05)
# use RandomTreesEmbedding to transform data

hasher = RandomTreesEmbedding(n_estimators=10, random_state=0, max_depth=3)
X_transformed = hasher.fit_transform(X)
# Visualize result after dimensionality reduction using truncated SVD

svd = TruncatedSVD(n_components=2)
X_reduced = svd.fit_transform(X_transformed)
# Learn a Naive Bayes classifier on the transformed data

nb = BernoulliNB()
nb.fit(X_transformed, y)
# Learn an ExtraTreesClassifier for comparison

trees = ExtraTreesClassifier(max_depth=3, n_estimators=10, random_state=0)
trees.fit(X, y)
# scatter plot of original and reduced data

ax.scatter(X[:, 0], X[:, 1], c=y, s=50, edgecolor='k')
ax.set_title("Original Data (2d)")
ax.set_xticks(())
ax.set_yticks(())
ax.scatter(X_reduced[:, 0], X_reduced[:, 1], c=y, s=50, edgecolor='k')
ax.set_title("Truncated SVD reduction (2d) of transformed data (%dd)" %
X_transformed.shape[1])
ax.set_xticks(())
ax.set_yticks(())
# Plot the decision in original space. For that, we will assign a color
# to each point in the mesh [x_min, x_max]x[y_min, y_max].
h = .01
# transform grid using RandomTreesEmbedding

transformed_grid = hasher.transform(np.c_[xx.ravel(), yy.ravel()])
y_grid_pred = nb.predict_proba(transformed_grid)[:, 1]
ax.set_title("Naive Bayes on Transformed data")

ax.pcolormesh(xx, yy, y_grid_pred.reshape(xx.shape))

ax.set_ylim(-1.4, 1.4)
ax.set_xlim(-1.4, 1.4)
ax.set_xticks(())
ax.set_yticks(())
# transform grid using ExtraTreesClassifier

y_grid_pred = trees.predict_proba(np.c_[xx.ravel(), yy.ravel()])[:, 1]
ax.set_title("ExtraTrees predictions")
ax.pcolormesh(xx, yy, y_grid_pred.reshape(xx.shape))
ax.set_ylim(-1.4, 1.4)
ax.set_xlim(-1.4, 1.4)
ax.set_xticks(())
ax.set_yticks(())
plt.tight_layout()
plt.show()
5.12.14 Partial Dependence Plots
Partial dependence plots show the dependence between the target function2 and a set of ‘target’ features, marginalizing
over the values of all other features (the complement features). Due to the limits of human perception the size of the
target feature set must be small (usually, one or two) thus the target features are usually chosen among the most
important features (see feature_importances_).
This example shows how to obtain partial dependence plots from a GradientBoostingRegressor trained on
the California housing dataset. The example is taken from1 .
The plot shows four one-way and one two-way partial dependence plots. The target variables for the one-way PDP
are: median income (MedInc), avg. occupants per household (AvgOccup), median house age (HouseAge), and avg.
rooms per household (AveRooms).
We can clearly see that the median house price shows a linear relationship with the median income (top left) and that
the house price drops when the avg. occupants per household increases (top middle). The top right plot shows that the
house age in a district does not have a strong influence on the (median) house price; so does the average rooms per
household. The tick marks on the x-axis represent the deciles of the feature values in the training data.
Partial dependence plots with two target features enable us to visualize interactions among them. The two-way partial
dependence plot shows the dependence of median house price on joint values of house age and avg. occupants per
household. We can clearly see an interaction between the two features: For an avg. occupancy greater than two, the
house price is nearly independent of the house age, whereas for values less than two there is a strong dependence on
age.
2 For classification you can think of it as the regression score before the link function.


•
Out:
Training GBRT...
done.
Convenience plot with ``partial_dependence_plots``
Custom 3d plot via ``partial_dependence``

print(__doc__)
import numpy as np

from sklearn.ensemble import GradientBoostingRegressor
from sklearn.ensemble.partial_dependence import plot_partial_dependence
from sklearn.ensemble.partial_dependence import partial_dependence
from sklearn.datasets.california_housing import fetch_california_housing

def main():
cal_housing = fetch_california_housing()
# split 80/20 train-test

X_train, X_test, y_train, y_test = train_test_split(cal_housing.data,
cal_housing.target,
test_size=0.2,
random_state=1)
names = cal_housing.feature_names
print("Training GBRT...")
clf = GradientBoostingRegressor(n_estimators=100, max_depth=4,
learning_rate=0.1, loss='huber',
random_state=1)
print(" done.")
print('Convenience plot with ``partial_dependence_plots``')
features = [0, 5, 1, 2, (5, 1)]

fig, axs = plot_partial_dependence(clf, X_train, features,
feature_names=names,
n_jobs=3, grid_resolution=50)
fig.suptitle('Partial dependence of house value on nonlocation features\n'
'for the California housing dataset')
plt.subplots_adjust(top=0.9) # tight_layout causes overlap with suptitle
print('Custom 3d plot via ``partial_dependence``')

fig = plt.figure()
target_feature = (1, 5)
pdp, axes = partial_dependence(clf, target_feature,
X=X_train, grid_resolution=50)
XX, YY = np.meshgrid(axes[0], axes[1])
Z = pdp[0].reshape(list(map(np.size, axes))).T
ax = Axes3D(fig)
surf = ax.plot_surface(XX, YY, Z, rstride=1, cstride=1,
cmap=plt.cm.BuPu, edgecolor='k')
ax.set_xlabel(names[target_feature[0]])
ax.set_ylabel(names[target_feature[1]])
ax.set_zlabel('Partial dependence')
# pretty init view
ax.view_init(elev=22, azim=122)
plt.colorbar(surf)
plt.suptitle('Partial dependence of house value on median\n'
'age and average occupancy')
plt.subplots_adjust(top=0.9)
plt.show()
# Needed on Windows because plot_partial_dependence uses multiprocessing

if __name__ == '__main__':
main()

5.12.15 Discrete versus Real AdaBoost
This example is based on Figure 10.2 from Hastie et al 20091 and illustrates the difference in performance between
the discrete SAMME2 boosting algorithm and real SAMME.R boosting algorithm. Both algorithms are evaluated on
a binary classification task where the target Y is a non-linear function of 10 input features.
Discrete SAMME AdaBoost adapts based on errors in predicted class labels whereas real SAMME.R uses the pre-
dicted class probabilities.
print(__doc__)
# Author: Peter Prettenhofer <peter.prettenhofer@gmail.com>,

# Noel Dawe <noel.dawe@gmail.com>
#
2
10. Zhu, H. Zou, S. Rosset, T. Hastie, “Multi-class AdaBoost”, 2009.

import numpy as np

from sklearn.metrics import zero_one_loss
n_estimators = 400
# A learning rate of 1. may not be optimal for both SAMME and SAMME.R
learning_rate = 1.
X, y = datasets.make_hastie_10_2(n_samples=12000, random_state=1)
X_test, y_test = X[2000:], y[2000:]

X_train, y_train = X[:2000], y[:2000]
dt_stump = DecisionTreeClassifier(max_depth=1, min_samples_leaf=1)

dt_stump.fit(X_train, y_train)
dt_stump_err = 1.0 - dt_stump.score(X_test, y_test)
dt = DecisionTreeClassifier(max_depth=9, min_samples_leaf=1)
dt.fit(X_train, y_train)
dt_err = 1.0 - dt.score(X_test, y_test)
ada_discrete = AdaBoostClassifier(
base_estimator=dt_stump,
learning_rate=learning_rate,
n_estimators=n_estimators,
algorithm="SAMME")
ada_discrete.fit(X_train, y_train)
ada_real = AdaBoostClassifier(
base_estimator=dt_stump,
learning_rate=learning_rate,
n_estimators=n_estimators,
algorithm="SAMME.R")
ada_real.fit(X_train, y_train)
fig = plt.figure()
ax = fig.add_subplot(111)
ax.plot([1, n_estimators], [dt_stump_err] * 2, 'k-',

label='Decision Stump Error')
ax.plot([1, n_estimators], [dt_err] * 2, 'k--',
label='Decision Tree Error')
ada_discrete_err = np.zeros((n_estimators,))
for i, y_pred in enumerate(ada_discrete.staged_predict(X_test)):
ada_discrete_err[i] = zero_one_loss(y_pred, y_test)
ada_discrete_err_train = np.zeros((n_estimators,))
for i, y_pred in enumerate(ada_discrete.staged_predict(X_train)):
ada_discrete_err_train[i] = zero_one_loss(y_pred, y_train)
ada_real_err = np.zeros((n_estimators,))
for i, y_pred in enumerate(ada_real.staged_predict(X_test)):

ada_real_err[i] = zero_one_loss(y_pred, y_test)
ada_real_err_train = np.zeros((n_estimators,))
for i, y_pred in enumerate(ada_real.staged_predict(X_train)):
ada_real_err_train[i] = zero_one_loss(y_pred, y_train)
ax.plot(np.arange(n_estimators) + 1, ada_discrete_err,
label='Discrete AdaBoost Test Error',
color='red')
ax.plot(np.arange(n_estimators) + 1, ada_discrete_err_train,
label='Discrete AdaBoost Train Error',
color='blue')
ax.plot(np.arange(n_estimators) + 1, ada_real_err,
label='Real AdaBoost Test Error',
color='orange')
ax.plot(np.arange(n_estimators) + 1, ada_real_err_train,
label='Real AdaBoost Train Error',
color='green')
ax.set_ylim((0.0, 0.5))
ax.set_xlabel('n_estimators')
ax.set_ylabel('error rate')
leg = ax.legend(loc='upper right', fancybox=True)

leg.get_frame().set_alpha(0.7)
plt.show()
5.12.16 Multi-class AdaBoosted Decision Trees
This example reproduces Figure 1 of Zhu et al1 and shows how boosting can improve prediction accuracy on a multi-
class problem. The classification dataset is constructed by taking a ten-dimensional standard normal distribution and
defining three classes separated by nested concentric ten-dimensional spheres such that roughly equal numbers of
samples are in each class (quantiles of the 𝜒2 distribution).
The performance of the SAMME and SAMME.R1 algorithms are compared. SAMME.R uses the probability estimates
to update the additive model, while SAMME uses the classifications only. As the example illustrates, the SAMME.R
algorithm typically converges faster than SAMME, achieving a lower test error with fewer boosting iterations. The
error of each algorithm on the test set after each boosting iteration is shown on the left, the classification error on the
test set of each tree is shown in the middle, and the boost weight of each tree is shown on the right. All trees have a
weight of one in the SAMME.R algorithm and therefore are not shown.
1
10. Zhu, H. Zou, S. Rosset, T. Hastie, “Multi-class AdaBoost”, 2009.

print(__doc__)

#
from sklearn.externals.six.moves import zip

X, y = make_gaussian_quantiles(n_samples=13000, n_features=10,
n_split = 3000
X_train, X_test = X[:n_split], X[n_split:]

y_train, y_test = y[:n_split], y[n_split:]
bdt_real = AdaBoostClassifier(
n_estimators=600,
learning_rate=1)
bdt_discrete = AdaBoostClassifier(
n_estimators=600,
learning_rate=1.5,
algorithm="SAMME")
bdt_real.fit(X_train, y_train)
bdt_discrete.fit(X_train, y_train)
real_test_errors = []
discrete_test_errors = []
for real_test_predict, discrete_train_predict in zip(

bdt_real.staged_predict(X_test), bdt_discrete.staged_predict(X_test)):

real_test_errors.append(
1. - accuracy_score(real_test_predict, y_test))
discrete_test_errors.append(
1. - accuracy_score(discrete_train_predict, y_test))
n_trees_discrete = len(bdt_discrete)
n_trees_real = len(bdt_real)
# Boosting might terminate early, but the following arrays are always
# n_estimators long. We crop them to the actual number of trees here:
discrete_estimator_errors = bdt_discrete.estimator_errors_[:n_trees_discrete]
real_estimator_errors = bdt_real.estimator_errors_[:n_trees_real]
discrete_estimator_weights = bdt_discrete.estimator_weights_[:n_trees_discrete]
plt.subplot(131)
plt.plot(range(1, n_trees_discrete + 1),
discrete_test_errors, c='black', label='SAMME')
plt.plot(range(1, n_trees_real + 1),
real_test_errors, c='black',
linestyle='dashed', label='SAMME.R')
plt.legend()
plt.ylim(0.18, 0.62)
plt.ylabel('Test Error')
plt.xlabel('Number of Trees')
plt.subplot(132)
plt.plot(range(1, n_trees_discrete + 1), discrete_estimator_errors,
"b", label='SAMME', alpha=.5)
plt.plot(range(1, n_trees_real + 1), real_estimator_errors,
"r", label='SAMME.R', alpha=.5)
plt.legend()
plt.ylabel('Error')
plt.ylim((.2,
max(real_estimator_errors.max(),
discrete_estimator_errors.max()) * 1.2))
plt.xlim((-20, len(bdt_discrete) + 20))
plt.subplot(133)
plt.plot(range(1, n_trees_discrete + 1), discrete_estimator_weights,
"b", label='SAMME')
plt.legend()
plt.ylabel('Weight')
plt.ylim((0, discrete_estimator_weights.max() * 1.2))
plt.xlim((-20, n_trees_discrete + 20))
# prevent overlapping y-axis labels

plt.subplots_adjust(wspace=0.25)
plt.show()

5.12.17 Early stopping of Gradient Boosting
Gradient boosting is an ensembling technique where several weak learners (regression trees) are combined to yield a
powerful single model, in an iterative fashion.
Early stopping support in Gradient Boosting enables us to find the least number of iterations which is sufficient to
build a model that generalizes well to unseen data.
The concept of early stopping is simple. We specify a validation_fraction which denotes the fraction of the
whole dataset that will be kept aside from training to assess the validation loss of the model. The gradient boosting
model is trained using the training set and evaluated using the validation set. When each additional stage of regression
tree is added, the validation set is used to score the model. This is continued until the scores of the model in the last
n_iter_no_change stages do not improve by atleast tol. After that the model is considered to have converged and
further addition of stages is “stopped early”.
The number of stages of the final model is available at the attribute n_estimators_.
This example illustrates how the early stopping can used in the sklearn.ensemble.
GradientBoostingClassifier model to achieve almost the same accuracy as compared to a model
built without early stopping using many fewer estimators. This can significantly reduce training time, memory usage
and prediction latency.
# Authors: Vighnesh Birodkar <vighneshbirodkar@nyu.edu>

# Raghav RV <rvraghav93@gmail.com>
import time
import numpy as np

print(__doc__)
data_list = [datasets.load_iris(), datasets.load_digits()]

data_list = [(d.data, d.target) for d in data_list]
data_list += [datasets.make_hastie_10_2()]
names = ['Iris Data', 'Digits Data', 'Hastie Data']
n_gb = []
score_gb = []
time_gb = []
n_gbes = []
score_gbes = []
time_gbes = []
n_estimators = 500
for X, y in data_list:
random_state=0)
# We specify that if the scores don't improve by atleast 0.01 for the last
# 10 stages, stop fitting additional stages
gbes = ensemble.GradientBoostingClassifier(n_estimators=n_estimators,
validation_fraction=0.2,

n_iter_no_change=5, tol=0.01,
random_state=0)
gb = ensemble.GradientBoostingClassifier(n_estimators=n_estimators,
random_state=0)
start = time.time()
gb.fit(X_train, y_train)
time_gb.append(time.time() - start)
start = time.time()
gbes.fit(X_train, y_train)
time_gbes.append(time.time() - start)
score_gb.append(gb.score(X_test, y_test))
score_gbes.append(gbes.score(X_test, y_test))
n_gb.append(gb.n_estimators_)
n_gbes.append(gbes.n_estimators_)
bar_width = 0.2
n = len(data_list)
index = np.arange(0, n * bar_width, bar_width) * 2.5
index = index[0:n]
Compare scores with and without early stopping
bar1 = plt.bar(index, score_gb, bar_width, label='Without early stopping',

color='crimson')
bar2 = plt.bar(index + bar_width, score_gbes, bar_width,
label='With early stopping', color='coral')
plt.xticks(index + bar_width, names)

plt.yticks(np.arange(0, 1.3, 0.1))
def autolabel(rects, n_estimators):

"""
Attach a text label above each bar displaying n_estimators of each model
"""
for i, rect in enumerate(rects):
plt.text(rect.get_x() + rect.get_width() / 2.,
1.05 * rect.get_height(), 'n_est=%d' % n_estimators[i],
ha='center', va='bottom')
autolabel(bar1, n_gb)
autolabel(bar2, n_gbes)
plt.ylim([0, 1.3])
plt.grid(True)
plt.xlabel('Datasets')
plt.ylabel('Test score')

plt.show()
Compare fit times with and without early stopping
bar1 = plt.bar(index, time_gb, bar_width, label='Without early stopping',

color='crimson')
bar2 = plt.bar(index + bar_width, time_gbes, bar_width,
label='With early stopping', color='coral')
max_y = np.amax(np.maximum(time_gb, time_gbes))
plt.xticks(index + bar_width, names)

plt.yticks(np.linspace(0, 1.3 * max_y, 13))
autolabel(bar1, n_gb)
autolabel(bar2, n_gbes)
plt.ylim([0, 1.3 * max_y])

plt.grid(True)
plt.xlabel('Datasets')
plt.ylabel('Fit Time')
plt.show()

5.12.18 Feature transformations with ensembles of trees
Transform your features into a higher dimensional, sparse space. Then train a linear model on these features.
First fit an ensemble of trees (totally random trees, a random forest, or gradient boosted trees) on the training set. Then
each leaf of each tree in the ensemble is assigned a fixed arbitrary feature index in a new feature space. These leaf
indices are then encoded in a one-hot fashion.
Each sample goes through the decisions of each tree of the ensemble and ends up in one leaf per tree. The sample is
encoded by setting feature values for these leaves to 1 and the other feature values to 0.
The resulting transformer has then learned a supervised, sparse, high-dimensional categorical embedding of the data.


# Author: Tim Head <betatim@gmail.com>

#
import numpy as np
np.random.seed(10)

from sklearn.ensemble import (RandomTreesEmbedding, RandomForestClassifier,
GradientBoostingClassifier)
from sklearn.preprocessing import OneHotEncoder
from sklearn.metrics import roc_curve
from sklearn.pipeline import make_pipeline
n_estimator = 10
X, y = make_classification(n_samples=80000)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5)
# It is important to train the ensemble of trees on a different subset

# of the training data than the linear regression model to avoid
# overfitting, in particular if the total number of leaves is
# similar to the number of training samples
X_train, X_train_lr, y_train, y_train_lr = train_test_split(

X_train, y_train, test_size=0.5)
# Unsupervised transformation based on totally random trees

rt = RandomTreesEmbedding(max_depth=3, n_estimators=n_estimator,
random_state=0)
rt_lm = LogisticRegression(solver='lbfgs', max_iter=1000)

pipeline = make_pipeline(rt, rt_lm)
pipeline.fit(X_train, y_train)
y_pred_rt = pipeline.predict_proba(X_test)[:, 1]
fpr_rt_lm, tpr_rt_lm, _ = roc_curve(y_test, y_pred_rt)
# Supervised transformation based on random forests

rf = RandomForestClassifier(max_depth=3, n_estimators=n_estimator)
rf_enc = OneHotEncoder(categories='auto')
rf_lm = LogisticRegression(solver='lbfgs', max_iter=1000)
rf.fit(X_train, y_train)
rf_enc.fit(rf.apply(X_train))
rf_lm.fit(rf_enc.transform(rf.apply(X_train_lr)), y_train_lr)
y_pred_rf_lm = rf_lm.predict_proba(rf_enc.transform(rf.apply(X_test)))[:, 1]
fpr_rf_lm, tpr_rf_lm, _ = roc_curve(y_test, y_pred_rf_lm)
# Supervised transformation based on gradient boosted trees

grd = GradientBoostingClassifier(n_estimators=n_estimator)
grd_enc = OneHotEncoder(categories='auto')
grd_lm = LogisticRegression(solver='lbfgs', max_iter=1000)
grd.fit(X_train, y_train)
grd_enc.fit(grd.apply(X_train)[:, :, 0])
grd_lm.fit(grd_enc.transform(grd.apply(X_train_lr)[:, :, 0]), y_train_lr)
y_pred_grd_lm = grd_lm.predict_proba(
grd_enc.transform(grd.apply(X_test)[:, :, 0]))[:, 1]
fpr_grd_lm, tpr_grd_lm, _ = roc_curve(y_test, y_pred_grd_lm)
# The gradient boosted model by itself

y_pred_grd = grd.predict_proba(X_test)[:, 1]
fpr_grd, tpr_grd, _ = roc_curve(y_test, y_pred_grd)
# The random forest model by itself

y_pred_rf = rf.predict_proba(X_test)[:, 1]
fpr_rf, tpr_rf, _ = roc_curve(y_test, y_pred_rf)
plt.figure(1)
plt.plot([0, 1], [0, 1], 'k--')
plt.plot(fpr_rt_lm, tpr_rt_lm, label='RT + LR')
plt.plot(fpr_rf, tpr_rf, label='RF')
plt.plot(fpr_rf_lm, tpr_rf_lm, label='RF + LR')
plt.plot(fpr_grd, tpr_grd, label='GBT')
plt.plot(fpr_grd_lm, tpr_grd_lm, label='GBT + LR')
plt.xlabel('False positive rate')
plt.ylabel('True positive rate')
plt.title('ROC curve')
plt.show()
plt.figure(2)
plt.xlim(0, 0.2)

plt.ylim(0.8, 1)
plt.plot([0, 1], [0, 1], 'k--')
plt.plot(fpr_rt_lm, tpr_rt_lm, label='RT + LR')
plt.plot(fpr_rf, tpr_rf, label='RF')
plt.plot(fpr_rf_lm, tpr_rf_lm, label='RF + LR')
plt.plot(fpr_grd, tpr_grd, label='GBT')
plt.plot(fpr_grd_lm, tpr_grd_lm, label='GBT + LR')
plt.xlabel('False positive rate')
plt.ylabel('True positive rate')
plt.title('ROC curve (zoomed in at top left)')
plt.show()
5.12.19 Gradient Boosting Out-of-Bag estimates
Out-of-bag (OOB) estimates can be a useful heuristic to estimate the “optimal” number of boosting iterations. OOB
estimates are almost identical to cross-validation estimates but they can be computed on-the-fly without the need for
repeated model fitting. OOB estimates are only available for Stochastic Gradient Boosting (i.e. subsample < 1.
0), the estimates are derived from the improvement in loss based on the examples not included in the bootstrap sample
(the so-called out-of-bag examples). The OOB estimator is a pessimistic estimator of the true test loss, but remains a
fairly good approximation for a small number of trees.
The figure shows the cumulative sum of the negative OOB improvements as a function of the boosting iteration. As
you can see, it tracks the test loss for the first hundred iterations but then diverges in a pessimistic way. The figure
also shows the performance of 3-fold cross validation which usually gives a better estimate of the test loss but is
computationally more demanding.

Out:
Accuracy: 0.6840
print(__doc__)

#
import numpy as np

# Generate data (adapted from G. Ridgeway's gbm example)

n_samples = 1000

x1 = random_state.uniform(size=n_samples)
x2 = random_state.uniform(size=n_samples)
x3 = random_state.randint(0, 4, size=n_samples)
p = 1 / (1.0 + np.exp(-(np.sin(3 * x1) - 4 * x2 + x3)))

y = random_state.binomial(1, p, size=n_samples)
X = np.c_[x1, x2, x3]
random_state=9)
# Fit classifier with out-of-bag estimates

params = {'n_estimators': 1200, 'max_depth': 3, 'subsample': 0.5,
'learning_rate': 0.01, 'min_samples_leaf': 1, 'random_state': 3}
clf = ensemble.GradientBoostingClassifier(**params)
acc = clf.score(X_test, y_test)
print("Accuracy: {:.4f}".format(acc))
n_estimators = params['n_estimators']
x = np.arange(n_estimators) + 1
def heldout_score(clf, X_test, y_test):

"""compute deviance scores on ``X_test`` and ``y_test``. """
score = np.zeros((n_estimators,), dtype=np.float64)
for i, y_pred in enumerate(clf.staged_decision_function(X_test)):
score[i] = clf.loss_(y_test, y_pred)
return score
def cv_estimate(n_splits=None):
cv = KFold(n_splits=n_splits)
cv_clf = ensemble.GradientBoostingClassifier(**params)
val_scores = np.zeros((n_estimators,), dtype=np.float64)
for train, test in cv.split(X_train, y_train):
cv_clf.fit(X_train[train], y_train[train])
val_scores += heldout_score(cv_clf, X_train[test], y_train[test])
val_scores /= n_splits
return val_scores
# Estimate best n_estimator using cross-validation

cv_score = cv_estimate(3)
# Compute best n_estimator for test data

test_score = heldout_score(clf, X_test, y_test)
# negative cumulative sum of oob improvements

cumsum = -np.cumsum(clf.oob_improvement_)
# min loss according to OOB

oob_best_iter = x[np.argmin(cumsum)]
# min loss according to test (normalize such that first loss is 0)

test_score -= test_score[0]
test_best_iter = x[np.argmin(test_score)]
# min loss according to cv (normalize such that first loss is 0)

cv_score -= cv_score[0]
cv_best_iter = x[np.argmin(cv_score)]
# color brew for the three curves

oob_color = list(map(lambda x: x / 256.0, (190, 174, 212)))
test_color = list(map(lambda x: x / 256.0, (127, 201, 127)))
cv_color = list(map(lambda x: x / 256.0, (253, 192, 134)))
# plot curves and vertical lines for best iterations

plt.plot(x, cumsum, label='OOB loss', color=oob_color)
plt.plot(x, test_score, label='Test loss', color=test_color)
plt.plot(x, cv_score, label='CV loss', color=cv_color)
plt.axvline(x=oob_best_iter, color=oob_color)
plt.axvline(x=test_best_iter, color=test_color)
plt.axvline(x=cv_best_iter, color=cv_color)
# add three vertical lines to xticks

xticks = plt.xticks()
xticks_pos = np.array(xticks[0].tolist() +
[oob_best_iter, cv_best_iter, test_best_iter])
xticks_label = np.array(list(map(lambda t: int(t), xticks[0])) +
['OOB', 'CV', 'Test'])
ind = np.argsort(xticks_pos)
xticks_pos = xticks_pos[ind]
xticks_label = xticks_label[ind]
plt.xticks(xticks_pos, xticks_label)
plt.ylabel('normalized loss')
plt.xlabel('number of iterations')
plt.show()
5.12.20 Single estimator versus bagging: bias-variance decomposition
This example illustrates and compares the bias-variance decomposition of the expected mean squared error of a single
estimator against a bagging ensemble.
In regression, the expected mean squared error of an estimator can be decomposed in terms of bias, variance and
noise. On average over datasets of the regression problem, the bias term measures the average amount by which the
predictions of the estimator differ from the predictions of the best possible estimator for the problem (i.e., the Bayes
model). The variance term measures the variability of the predictions of the estimator when fit over different instances
LS of the problem. Finally, the noise measures the irreducible part of the error which is due the variability in the data.
The upper left figure illustrates the predictions (in dark red) of a single decision tree trained over a random dataset LS
(the blue dots) of a toy 1d regression problem. It also illustrates the predictions (in light red) of other single decision
trees trained over other (and different) randomly drawn instances LS of the problem. Intuitively, the variance term

here corresponds to the width of the beam of predictions (in light red) of the individual estimators. The larger the
variance, the more sensitive are the predictions for x to small changes in the training set. The bias term corresponds to
the difference between the average prediction of the estimator (in cyan) and the best possible model (in dark blue). On
this problem, we can thus observe that the bias is quite low (both the cyan and the blue curves are close to each other)
while the variance is large (the red beam is rather wide).
The lower left figure plots the pointwise decomposition of the expected mean squared error of a single decision tree.
It confirms that the bias term (in blue) is low while the variance is large (in green). It also illustrates the noise part of
the error which, as expected, appears to be constant and around 0.01.
The right figures correspond to the same plots but using instead a bagging ensemble of decision trees. In both figures,
we can observe that the bias term is larger than in the previous case. In the upper right figure, the difference between
the average prediction (in cyan) and the best possible model is larger (e.g., notice the offset around x=2). In the lower
right figure, the bias curve is also slightly higher than in the lower left figure. In terms of variance however, the beam
of predictions is narrower, which suggests that the variance is lower. Indeed, as the lower right figure confirms, the
variance term (in green) is lower than for single decision trees. Overall, the bias- variance decomposition is therefore
no longer the same. The tradeoff is better for bagging: averaging several decision trees fit on bootstrap copies of the
dataset slightly increases the bias term but allows for a larger reduction of the variance, which results in a lower overall
mean squared error (compare the red curves int the lower figures). The script output also confirms this intuition. The
total error of the bagging ensemble is lower than the total error of a single decision tree, and this difference indeed
mainly stems from a reduced variance.
For further details on bias-variance decomposition, see section 7.3 of1 .
1 T. Hastie, R. Tibshirani and J. Friedman, “Elements of Statistical Learning”, Springer, 2009.

References
Out:
Tree: 0.0255 (error) = 0.0003 (bias^2) + 0.0152 (var) + 0.0098 (noise)

Bagging(Tree): 0.0196 (error) = 0.0004 (bias^2) + 0.0092 (var) + 0.0098 (noise)
print(__doc__)
# Author: Gilles Louppe <g.louppe@gmail.com>

import numpy as np
from sklearn.ensemble import BaggingRegressor


# Settings
n_repeat = 50 # Number of iterations for computing expectations
n_train = 50 # Size of the training set
n_test = 1000 # Size of the test set
noise = 0.1 # Standard deviation of the noise
np.random.seed(0)
# Change this for exploring the bias-variance decomposition of other

# estimators. This should work well for estimators with high variance (e.g.,
# decision trees or KNN), but poorly for estimators with low variance (e.g.,
# linear models).
estimators = [("Tree", DecisionTreeRegressor()),
("Bagging(Tree)", BaggingRegressor(DecisionTreeRegressor()))]
n_estimators = len(estimators)
# Generate data
def f(x):
x = x.ravel()
return np.exp(-x ** 2) + 1.5 * np.exp(-(x - 2) ** 2)
def generate(n_samples, noise, n_repeat=1):

X = np.random.rand(n_samples) * 10 - 5
X = np.sort(X)
if n_repeat == 1:
y = f(X) + np.random.normal(0.0, noise, n_samples)
else:
y = np.zeros((n_samples, n_repeat))
for i in range(n_repeat):
y[:, i] = f(X) + np.random.normal(0.0, noise, n_samples)
X = X.reshape((n_samples, 1))
return X, y
X_train = []
y_train = []
X, y = generate(n_samples=n_train, noise=noise)
X_train.append(X)
y_train.append(y)
X_test, y_test = generate(n_samples=n_test, noise=noise, n_repeat=n_repeat)
# Loop over estimators to compare

for n, (name, estimator) in enumerate(estimators):
# Compute predictions
y_predict = np.zeros((n_test, n_repeat))

estimator.fit(X_train[i], y_train[i])
y_predict[:, i] = estimator.predict(X_test)
# Bias^2 + Variance + Noise decomposition of the mean squared error

y_error = np.zeros(n_test)
for j in range(n_repeat):
y_error += (y_test[:, j] - y_predict[:, i]) ** 2
y_error /= (n_repeat * n_repeat)
y_noise = np.var(y_test, axis=1)

y_bias = (f(X_test) - np.mean(y_predict, axis=1)) ** 2
y_var = np.var(y_predict, axis=1)
print("{0}: {1:.4f} (error) = {2:.4f} (bias^2) "

" + {3:.4f} (var) + {4:.4f} (noise)".format(name,
np.mean(y_error),
np.mean(y_bias),
np.mean(y_var),
np.mean(y_noise)))
# Plot figures
plt.subplot(2, n_estimators, n + 1)
plt.plot(X_test, f(X_test), "b", label="$f(x)$")
plt.plot(X_train[0], y_train[0], ".b", label="LS ~ $y = f(x)+noise$")
if i == 0:
plt.plot(X_test, y_predict[:, i], "r", label=r"$\^y(x)$")
else:
plt.plot(X_test, y_predict[:, i], "r", alpha=0.05)
plt.plot(X_test, np.mean(y_predict, axis=1), "c",

label=r"$\mathbb{E}_{LS} \^y(x)$")
plt.xlim([-5, 5])
plt.title(name)
if n == n_estimators - 1:
plt.legend(loc=(1.1, .5))
plt.subplot(2, n_estimators, n_estimators + n + 1)

plt.plot(X_test, y_error, "r", label="$error(x)$")
plt.plot(X_test, y_bias, "b", label="$bias^2(x)$"),
plt.plot(X_test, y_var, "g", label="$variance(x)$"),
plt.plot(X_test, y_noise, "c", label="$noise(x)$")
plt.xlim([-5, 5])
plt.ylim([0, 0.1])
if n == n_estimators - 1:
plt.legend(loc=(1.1, .5))
plt.subplots_adjust(right=.75)

plt.show()
5.12.21 Plot the decision surfaces of ensembles of trees on the iris dataset
Plot the decision surfaces of forests of randomized trees trained on pairs of features of the iris dataset.
This plot compares the decision surfaces learned by a decision tree classifier (first column), by a random forest classi-
fier (second column), by an extra- trees classifier (third column) and by an AdaBoost classifier (fourth column).
In the first row, the classifiers are built using the sepal width and the sepal length features only, on the second row
using the petal length and sepal length only, and on the third row using the petal width and the petal length only.
In descending order of quality, when trained (outside of this example) on all 4 features using 30 estimators and scored
using 10 fold cross validation, we see:
ExtraTreesClassifier() # 0.95 score

RandomForestClassifier() # 0.94 score
AdaBoost(DecisionTree(max_depth=3)) # 0.94 score
DecisionTree(max_depth=None) # 0.94 score
Increasing max_depth for AdaBoost lowers the standard deviation of the scores (but the average score does not im-
prove).
See the console’s output for further details about each model.
In this example you might try to:
1. vary the max_depth for the DecisionTreeClassifier and AdaBoostClassifier,
perhaps try max_depth=3 for the DecisionTreeClassifier or max_depth=None for
AdaBoostClassifier
2. vary n_estimators
It is worth noting that RandomForests and ExtraTrees can be fitted in parallel on many cores as each tree is built
independently of the others. AdaBoost’s samples are built sequentially and so do not use multiple cores.

Out:
DecisionTree with features [0, 1] has a score of 0.9266666666666666

RandomForest with 30 estimators with features [0, 1] has a score of 0.9266666666666666
ExtraTrees with 30 estimators with features [0, 1] has a score of 0.9266666666666666
AdaBoost with 30 estimators with features [0, 1] has a score of 0.84
print(__doc__)
import numpy as np


from sklearn.ensemble import (RandomForestClassifier, ExtraTreesClassifier,
AdaBoostClassifier)
# Parameters
n_classes = 3
n_estimators = 30
cmap = plt.cm.RdYlBu
plot_step = 0.02 # fine step width for decision surface contours
plot_step_coarser = 0.5 # step widths for coarse classifier guesses
RANDOM_SEED = 13 # fix the seed on each iteration
# Load data
iris = load_iris()
plot_idx = 1
models = [DecisionTreeClassifier(max_depth=None),
RandomForestClassifier(n_estimators=n_estimators),
ExtraTreesClassifier(n_estimators=n_estimators),
AdaBoostClassifier(DecisionTreeClassifier(max_depth=3),
n_estimators=n_estimators)]
for pair in ([0, 1], [0, 2], [2, 3]):

for model in models:
# We only take the two corresponding features
X = iris.data[:, pair]
y = iris.target
# Shuffle
idx = np.arange(X.shape[0])
np.random.seed(RANDOM_SEED)
np.random.shuffle(idx)
X = X[idx]
y = y[idx]
# Standardize
mean = X.mean(axis=0)
std = X.std(axis=0)
X = (X - mean) / std
# Train
model.fit(X, y)
scores = model.score(X, y)
# Create a title for each column and the console by using str() and
# slicing away useless parts of the string
model_title = str(type(model)).split(
".")[-1][:-2][:-len("Classifier")]
model_details = model_title
if hasattr(model, "estimators_"):
model_details += " with {} estimators".format(
len(model.estimators_))
print(model_details + " with features", pair,
"has a score of", scores)

plt.subplot(3, 4, plot_idx)
if plot_idx <= len(models):
# Add a title at the top of each column
plt.title(model_title, fontsize=9)
# Now plot the decision boundary using a fine mesh as input to a

# filled contour plot
# Plot either a single DecisionTreeClassifier or alpha blend the

# decision surfaces of the ensemble of classifiers
if isinstance(model, DecisionTreeClassifier):
Z = model.predict(np.c_[xx.ravel(), yy.ravel()])
cs = plt.contourf(xx, yy, Z, cmap=cmap)
else:
# Choose alpha blend level with respect to the number
# of estimators
# that are in use (noting that AdaBoost can use fewer estimators
# than its maximum if it achieves a good enough fit early on)
estimator_alpha = 1.0 / len(model.estimators_)
for tree in model.estimators_:
Z = tree.predict(np.c_[xx.ravel(), yy.ravel()])
cs = plt.contourf(xx, yy, Z, alpha=estimator_alpha, cmap=cmap)
# Build a coarser grid to plot a set of ensemble classifications

# to show how these are different to what we see in the decision
# surfaces. These points are regularly space and do not have a
# black outline
xx_coarser, yy_coarser = np.meshgrid(
np.arange(x_min, x_max, plot_step_coarser),
np.arange(y_min, y_max, plot_step_coarser))
Z_points_coarser = model.predict(np.c_[xx_coarser.ravel(),
yy_coarser.ravel()]
).reshape(xx_coarser.shape)
cs_points = plt.scatter(xx_coarser, yy_coarser, s=15,
c=Z_points_coarser, cmap=cmap,
edgecolors="none")
# Plot the training points, these are clustered together and have a
# black outline
plt.scatter(X[:, 0], X[:, 1], c=y,
cmap=ListedColormap(['r', 'y', 'b']),
edgecolor='k', s=20)
plot_idx += 1 # move on to the next plot in sequence
plt.suptitle("Classifiers on feature subsets of the Iris dataset", fontsize=12)

plt.axis("tight")
plt.tight_layout(h_pad=0.2, w_pad=0.2, pad=2.5)
plt.show()

5.13 Tutorial exercises
Exercises for the tutorials
5.13.1 Digits Classification Exercise
A tutorial exercise regarding the use of classification techniques on the Digits dataset.
This exercise is used in the Classification part of the Supervised learning: predicting an output variable from high-
dimensional observations section of the A tutorial on statistical-learning for scientific data processing.
Out:
KNN score: 0.961111

LogisticRegression score: 0.933333
print(__doc__)
from sklearn import datasets, neighbors, linear_model
X_digits = digits.data / digits.data.max()
n_samples = len(X_digits)
X_train = X_digits[:int(.9 * n_samples)]

y_train = y_digits[:int(.9 * n_samples)]
X_test = X_digits[int(.9 * n_samples):]
y_test = y_digits[int(.9 * n_samples):]
knn = neighbors.KNeighborsClassifier()
logistic = linear_model.LogisticRegression(solver='lbfgs', max_iter=1000,
multi_class='multinomial')
print('KNN score: %f' % knn.fit(X_train, y_train).score(X_test, y_test))

print('LogisticRegression score: %f'
% logistic.fit(X_train, y_train).score(X_test, y_test))

5.13.2 Cross-validation on Digits Dataset Exercise
A tutorial exercise using Cross-validation with an SVM on the Digits dataset.

This exercise is used in the Cross-validation generators part of the Model selection: choosing estimators and their
parameters section of the A tutorial on statistical-learning for scientific data processing.
print(__doc__)
import numpy as np
X = digits.data
y = digits.target
svc = svm.SVC(kernel='linear')
C_s = np.logspace(-10, 0, 10)
scores = list()
scores_std = list()
for C in C_s:
svc.C = C
this_scores = cross_val_score(svc, X, y, cv=5, n_jobs=1)
scores.append(np.mean(this_scores))
scores_std.append(np.std(this_scores))
# Do the plotting
plt.clf()
plt.semilogx(C_s, scores)
plt.semilogx(C_s, np.array(scores) + np.array(scores_std), 'b--')
plt.semilogx(C_s, np.array(scores) - np.array(scores_std), 'b--')
locs, labels = plt.yticks()
plt.yticks(locs, list(map(lambda x: "%g" % x, locs)))
5.13. Tutorial exercises 1031

plt.ylabel('CV score')
plt.xlabel('Parameter C')
plt.ylim(0, 1.1)
plt.show()
5.13.3 SVM Exercise
A tutorial exercise for using different SVM kernels.

This exercise is used in the Using kernels part of the Supervised learning: predicting an output variable from high-
dimensional observations section of the A tutorial on statistical-learning for scientific data processing.


print(__doc__)
import numpy as np
X = iris.data
y = iris.target
X = X[y != 0, :2]
y = y[y != 0]
n_sample = len(X)
np.random.seed(0)
order = np.random.permutation(n_sample)
X = X[order]
y = y[order].astype(np.float)
X_train = X[:int(.9 * n_sample)]

y_train = y[:int(.9 * n_sample)]
X_test = X[int(.9 * n_sample):]
y_test = y[int(.9 * n_sample):]
# fit the model

for fig_num, kernel in enumerate(('linear', 'rbf', 'poly')):

clf = svm.SVC(kernel=kernel, gamma=10)
plt.figure(fig_num)
plt.clf()
plt.scatter(X[:, 0], X[:, 1], c=y, zorder=10, cmap=plt.cm.Paired,
# Circle out the test data

plt.scatter(X_test[:, 0], X_test[:, 1], s=80, facecolors='none',
zorder=10, edgecolor='k')
plt.axis('tight')
x_min = X[:, 0].min()
x_max = X[:, 0].max()
y_min = X[:, 1].min()
y_max = X[:, 1].max()
XX, YY = np.mgrid[x_min:x_max:200j, y_min:y_max:200j]

Z = clf.decision_function(np.c_[XX.ravel(), YY.ravel()])

Z = Z.reshape(XX.shape)
plt.pcolormesh(XX, YY, Z > 0, cmap=plt.cm.Paired)
plt.contour(XX, YY, Z, colors=['k', 'k', 'k'],
linestyles=['--', '-', '--'], levels=[-.5, 0, .5])
plt.title(kernel)
plt.show()
5.13.4 Cross-validation on diabetes Dataset Exercise
A tutorial exercise which uses cross-validation with linear models.

This exercise is used in the Cross-validated estimators part of the Model selection: choosing estimators and their
parameters section of the A tutorial on statistical-learning for scientific data processing.

Out:
Answer to the bonus question: how much can you trust the selection of alpha?
Alpha parameters maximising the generalization score on different

subsets of the data:
[fold 0] alpha: 0.05968, score: 0.54209
[fold 1] alpha: 0.04520, score: 0.15523
[fold 2] alpha: 0.07880, score: 0.45193
Answer: Not very much since we obtained different alphas for different
subsets of the data and moreover, the scores for these alphas differ
quite substantially.

print(__doc__)
import numpy as np


X = diabetes.data[:150]
y = diabetes.target[:150]
lasso = Lasso(random_state=0)
alphas = np.logspace(-4, -0.5, 30)
tuned_parameters = [{'alpha': alphas}]

n_folds = 5
clf = GridSearchCV(lasso, tuned_parameters, cv=n_folds, refit=False)

clf.fit(X, y)
scores = clf.cv_results_['mean_test_score']
scores_std = clf.cv_results_['std_test_score']
plt.figure().set_size_inches(8, 6)
plt.semilogx(alphas, scores)
# plot error lines showing +/- std. errors of the scores

std_error = scores_std / np.sqrt(n_folds)
plt.semilogx(alphas, scores + std_error, 'b--')

plt.semilogx(alphas, scores - std_error, 'b--')
# alpha=0.2 controls the translucency of the fill color

plt.fill_between(alphas, scores + std_error, scores - std_error, alpha=0.2)
plt.ylabel('CV score +/- std error')

plt.xlabel('alpha')
plt.axhline(np.max(scores), linestyle='--', color='.5')
plt.xlim([alphas[0], alphas[-1]])
# #############################################################################
# Bonus: how much can you trust the selection of alpha?
# To answer this question we use the LassoCV object that sets its alpha
# parameter automatically from the data by internal cross-validation (i.e. it
# performs cross-validation on the training data it receives).
# We use external cross-validation to see how much the automatically obtained
# alphas differ across different cross-validation folds.
lasso_cv = LassoCV(alphas=alphas, cv=5, random_state=0)
k_fold = KFold(3)
print("Answer to the bonus question:",

"how much can you trust the selection of alpha?")
print()
print("Alpha parameters maximising the generalization score on different")
print("subsets of the data:")
for k, (train, test) in enumerate(k_fold.split(X, y)):
lasso_cv.fit(X[train], y[train])
print("[fold {0}] alpha: {1:.5f}, score: {2:.5f}".
format(k, lasso_cv.alpha_, lasso_cv.score(X[test], y[test])))
print()
print("Answer: Not very much since we obtained different alphas for different")

print("subsets of the data and moreover, the scores for these alphas differ")
print("quite substantially.")
plt.show()
5.14 Feature Selection
Examples concerning the sklearn.feature_selection module.
5.14.1 Recursive feature elimination
A recursive feature elimination example showing the relevance of pixels in a digit classification task.
Note: See also Recursive feature elimination with cross-validation

print(__doc__)

from sklearn.feature_selection import RFE
# Load the digits dataset

X = digits.images.reshape((len(digits.images), -1))
y = digits.target
# Create the RFE object and rank each pixel

svc = SVC(kernel="linear", C=1)
rfe = RFE(estimator=svc, n_features_to_select=1, step=1)
rfe.fit(X, y)
ranking = rfe.ranking_.reshape(digits.images[0].shape)
# Plot pixel ranking

plt.matshow(ranking, cmap=plt.cm.Blues)
plt.colorbar()
plt.title("Ranking of pixels with RFE")
plt.show()
5.14.2 Comparison of F-test and mutual information
This example illustrates the differences between univariate F-test statistics and mutual information.
We consider 3 features x_1, x_2, x_3 distributed uniformly over [0, 1], the target depends on them as follows:
y = x_1 + sin(6 * pi * x_2) + 0.1 * N(0, 1), that is the third features is completely irrelevant.
The code below plots the dependency of y against individual x_i and normalized values of univariate F-tests statistics
and mutual information.
As F-test captures only linear dependency, it rates x_1 as the most discriminative feature. On the other hand, mutual
information can capture any kind of dependency between variables and it rates x_2 as the most discriminative feature,
which probably agrees better with our intuitive perception for this example. Both methods correctly marks x_3 as
irrelevant.
5.14. Feature Selection 1039

print(__doc__)
import numpy as np
from sklearn.feature_selection import f_regression, mutual_info_regression
np.random.seed(0)
X = np.random.rand(1000, 3)
y = X[:, 0] + np.sin(6 * np.pi * X[:, 1]) + 0.1 * np.random.randn(1000)
f_test, _ = f_regression(X, y)
f_test /= np.max(f_test)
mi = mutual_info_regression(X, y)
mi /= np.max(mi)
for i in range(3):
plt.scatter(X[:, i], y, edgecolor='black', s=20)
plt.xlabel("$x_{}$".format(i + 1), fontsize=14)
if i == 0:
plt.ylabel("$y$", fontsize=14)
plt.title("F-test={:.2f}, MI={:.2f}".format(f_test[i], mi[i]),
fontsize=16)
plt.show()
5.14.3 Pipeline Anova SVM
Simple usage of Pipeline that runs successively a univariate feature selection with anova and then a C-SVM of the
selected features.
Out:

0 0.75 0.50 0.60 6

1 0.60 1.00 0.75 6
2 0.67 0.80 0.73 5
3 1.00 0.62 0.77 8
micro avg 0.72 0.72 0.72 25

macro avg 0.75 0.73 0.71 25
weighted avg 0.78 0.72 0.72 25

from sklearn.datasets import samples_generator
from sklearn.feature_selection import SelectKBest, f_regression
print(__doc__)

X, y = samples_generator.make_classification(
n_features=20, n_informative=3, n_redundant=0, n_classes=4,
# ANOVA SVM-C
# 1) anova filter, take 3 best ranked features
anova_filter = SelectKBest(f_regression, k=3)
# 2) svm
clf = svm.SVC(kernel='linear')
anova_svm = make_pipeline(anova_filter, clf)

anova_svm.fit(X_train, y_train)
y_pred = anova_svm.predict(X_test)
print(classification_report(y_test, y_pred))
5.14.4 Recursive feature elimination with cross-validation
A recursive feature elimination example with automatic tuning of the number of features selected with cross-validation.

Out:
Optimal number of features : 3
print(__doc__)

from sklearn.model_selection import StratifiedKFold
from sklearn.feature_selection import RFECV
# Build a classification task using 3 informative features

X, y = make_classification(n_samples=1000, n_features=25, n_informative=3,
n_redundant=2, n_repeated=0, n_classes=8,
n_clusters_per_class=1, random_state=0)
# Create the RFE object and compute a cross-validated score.

svc = SVC(kernel="linear")
# The "accuracy" scoring is proportional to the number of correct
# classifications

rfecv = RFECV(estimator=svc, step=1, cv=StratifiedKFold(2),

scoring='accuracy')
rfecv.fit(X, y)
print("Optimal number of features : %d" % rfecv.n_features_)
# Plot number of features VS. cross-validation scores

plt.figure()
plt.xlabel("Number of features selected")
plt.ylabel("Cross validation score (nb of correct classifications)")
plt.plot(range(1, len(rfecv.grid_scores_) + 1), rfecv.grid_scores_)
plt.show()
5.14.5 Feature selection using SelectFromModel and LassoCV
Use SelectFromModel meta-transformer along with Lasso to select the best couple of features from the Boston dataset.
# Author: Manoj Kumar <mks542@nyu.edu>


print(__doc__)

import numpy as np

from sklearn.feature_selection import SelectFromModel
# Load the boston dataset.

boston = load_boston()
X, y = boston['data'], boston['target']
# We use the base estimator LassoCV since the L1 norm promotes sparsity of features.
clf = LassoCV(cv=5)
# Set a minimum threshold of 0.25

sfm = SelectFromModel(clf, threshold=0.25)
sfm.fit(X, y)
n_features = sfm.transform(X).shape[1]
# Reset the threshold till the number of features equals two.

# Note that the attribute can be set directly instead of repeatedly
# fitting the metatransformer.
while n_features > 2:
sfm.threshold += 0.1
X_transform = sfm.transform(X)
n_features = X_transform.shape[1]
# Plot the selected two features from X.

plt.title(
"Features selected from Boston using SelectFromModel with "
"threshold %0.3f." % sfm.threshold)
feature1 = X_transform[:, 0]
feature2 = X_transform[:, 1]
plt.plot(feature1, feature2, 'r.')
plt.xlabel("Feature number 1")
plt.ylabel("Feature number 2")
plt.ylim([np.min(feature2), np.max(feature2)])
plt.show()
5.14.6 Test with permutations the significance of a classification score
In order to test if a classification score is significative a technique in repeating the classification procedure after ran-
domizing, permuting, the labels. The p-value is then given by the percentage of runs for which the score obtained is
greater than the classification score obtained in the first place.

Out:
Classification score 0.5133333333333333 (pvalue : 0.009900990099009901)

print(__doc__)
import numpy as np

from sklearn.model_selection import permutation_test_score
# #############################################################################
# Loading a dataset

X = iris.data
y = iris.target
n_classes = np.unique(y).size
# Some noisy data not correlated

random = np.random.RandomState(seed=0)
E = random.normal(size=(len(X), 2200))
# Add noisy data to the informative features for make the task harder
X = np.c_[X, E]
svm = SVC(kernel='linear')
cv = StratifiedKFold(2)
score, permutation_scores, pvalue = permutation_test_score(

svm, X, y, scoring="accuracy", cv=cv, n_permutations=100, n_jobs=1)
print("Classification score %s (pvalue : %s)" % (score, pvalue))
# #############################################################################
# View histogram of permutation scores
plt.hist(permutation_scores, 20, label='Permutation scores',
edgecolor='black')
ylim = plt.ylim()
# BUG: vlines(..., linestyle='--') fails on older versions of matplotlib
# plt.vlines(score, ylim[0], ylim[1], linestyle='--',
# color='g', linewidth=3, label='Classification Score'
# ' (pvalue %s)' % pvalue)
# plt.vlines(1.0 / n_classes, ylim[0], ylim[1], linestyle='--',
# color='k', linewidth=3, label='Luck')
plt.plot(2 * [score], ylim, '--g', linewidth=3,
label='Classification Score'
' (pvalue %s)' % pvalue)
plt.plot(2 * [1. / n_classes], ylim, '--k', linewidth=3, label='Luck')
plt.ylim(ylim)
plt.legend()
plt.xlabel('Score')
plt.show()
5.14.7 Univariate Feature Selection
An example showing univariate feature selection.

Noisy (non informative) features are added to the iris data and univariate feature selection is applied. For each feature,
we plot the p-values for the univariate feature selection and the corresponding weights of an SVM. We can see that
univariate feature selection selects the informative features and that these have larger SVM weights.
In the total set of features, only the 4 first ones are significant. We can see that they have the highest score with
univariate feature selection. The SVM assigns a large weight to one of these features, but also Selects many of the
non-informative features. Applying univariate feature selection before the SVM increases the SVM weight attributed

to the significant features, and will thus improve classification.
print(__doc__)
import numpy as np

from sklearn.feature_selection import SelectPercentile, f_classif
# #############################################################################
# Import some data to play with
# The iris dataset

# Some noisy data not correlated

E = np.random.uniform(0, 0.1, size=(len(iris.data), 20))
# Add the noisy data to the informative features

X = np.hstack((iris.data, E))
y = iris.target
plt.figure(1)
plt.clf()

X_indices = np.arange(X.shape[-1])
# #############################################################################
# Univariate feature selection with F-test for feature scoring
# We use the default selection function: the 10% most significant features
selector = SelectPercentile(f_classif, percentile=10)
selector.fit(X, y)
scores = -np.log10(selector.pvalues_)
scores /= scores.max()
plt.bar(X_indices - .45, scores, width=.2,
label=r'Univariate score ($-Log(p_{value})$)', color='darkorange',
edgecolor='black')
# #############################################################################
# Compare to the weights of an SVM
clf = svm.SVC(kernel='linear')
clf.fit(X, y)
svm_weights = (clf.coef_ ** 2).sum(axis=0)

svm_weights /= svm_weights.max()
plt.bar(X_indices - .25, svm_weights, width=.2, label='SVM weight',

color='navy', edgecolor='black')
clf_selected = svm.SVC(kernel='linear')
clf_selected.fit(selector.transform(X), y)
svm_weights_selected = (clf_selected.coef_ ** 2).sum(axis=0)

svm_weights_selected /= svm_weights_selected.max()
plt.bar(X_indices[selector.get_support()] - .05, svm_weights_selected,

width=.2, label='SVM weights after selection', color='c',
edgecolor='black')
plt.title("Comparing feature selection")

plt.xlabel('Feature number')
plt.yticks(())
plt.axis('tight')
plt.show()
5.15 Gaussian Process for Machine Learning
Examples concerning the sklearn.gaussian_process module.

5.15.1 Illustration of Gaussian process classification (GPC) on the XOR dataset
This example illustrates GPC on XOR data. Compared are a stationary, isotropic kernel (RBF) and a non-stationary
kernel (DotProduct). On this particular dataset, the DotProduct kernel obtains considerably better results because the
class-boundaries are linear and coincide with the coordinate axes. In general, stationary kernels often obtain better
results.
print(__doc__)

#
import numpy as np

from sklearn.gaussian_process.kernels import RBF, DotProduct

X = rng.randn(200, 2)
Y = np.logical_xor(X[:, 0] > 0, X[:, 1] > 0)
# fit the model

kernels = [1.0 * RBF(length_scale=1.0), 1.0 * DotProduct(sigma_0=1.0)**2]
for i, kernel in enumerate(kernels):
clf = GaussianProcessClassifier(kernel=kernel, warm_start=True).fit(X, Y)
# plot the decision function for each datapoint on the grid

Z = clf.predict_proba(np.vstack((xx.ravel(), yy.ravel())).T)[:, 1]
5.15. Gaussian Process for Machine Learning 1049

image = plt.imshow(Z, interpolation='nearest',
extent=(xx.min(), xx.max(), yy.min(), yy.max()),
aspect='auto', origin='lower', cmap=plt.cm.PuOr_r)
contours = plt.contour(xx, yy, Z, levels=[0.5], linewidths=2,
colors=['k'])
plt.scatter(X[:, 0], X[:, 1], s=30, c=Y, cmap=plt.cm.Paired,
plt.xticks(())
plt.yticks(())
plt.axis([-3, 3, -3, 3])
plt.colorbar(image)
plt.title("%s\n Log-Marginal-Likelihood:%.3f"
% (clf.kernel_, clf.log_marginal_likelihood(clf.kernel_.theta)),
fontsize=12)
plt.tight_layout()
plt.show()
5.15.2 Gaussian process classification (GPC) on iris dataset
This example illustrates the predicted probability of GPC for an isotropic and anisotropic RBF kernel on a two-
dimensional version for the iris-dataset. The anisotropic RBF kernel obtains slightly higher log-marginal-likelihood
by assigning different length-scales to the two feature dimensions.
print(__doc__)
import numpy as np



y = np.array(iris.target, dtype=int)
kernel = 1.0 * RBF([1.0])

gpc_rbf_isotropic = GaussianProcessClassifier(kernel=kernel).fit(X, y)
kernel = 1.0 * RBF([1.0, 1.0])
gpc_rbf_anisotropic = GaussianProcessClassifier(kernel=kernel).fit(X, y)
# create a mesh to plot in

titles = ["Isotropic RBF", "Anisotropic RBF"]

for i, clf in enumerate((gpc_rbf_isotropic, gpc_rbf_anisotropic)):
# Plot the predicted probabilities. For that, we will assign a color to
# each point in the mesh [x_min, m_max]x[y_min, y_max].
Z = clf.predict_proba(np.c_[xx.ravel(), yy.ravel()])

Z = Z.reshape((xx.shape[0], xx.shape[1], 3))
plt.imshow(Z, extent=(x_min, x_max, y_min, y_max), origin="lower")

plt.scatter(X[:, 0], X[:, 1], c=np.array(["r", "g", "b"])[y],
plt.xlim(xx.min(), xx.max())
plt.ylim(yy.min(), yy.max())
plt.xticks(())
plt.yticks(())
plt.title("%s, LML: %.3f" %
(titles[i], clf.log_marginal_likelihood(clf.kernel_.theta)))
plt.tight_layout()
plt.show()

5.15.3 Comparison of kernel ridge and Gaussian process regression
Both kernel ridge regression (KRR) and Gaussian process regression (GPR) learn a target function by employing
internally the “kernel trick”. KRR learns a linear function in the space induced by the respective kernel which corre-
sponds to a non-linear function in the original space. The linear function in the kernel space is chosen based on the
mean-squared error loss with ridge regularization. GPR uses the kernel to define the covariance of a prior distribution
over the target functions and uses the observed training data to define a likelihood function. Based on Bayes theorem,
a (Gaussian) posterior distribution over target functions is defined, whose mean is used for prediction.
A major difference is that GPR can choose the kernel’s hyperparameters based on gradient-ascent on the marginal
likelihood function while KRR needs to perform a grid search on a cross-validated loss function (mean-squared error
loss). A further difference is that GPR learns a generative, probabilistic model of the target function and can thus
provide meaningful confidence intervals and posterior samples along with the predictions while KRR only provides
predictions.
This example illustrates both methods on an artificial dataset, which consists of a sinusoidal target function and strong
noise. The figure compares the learned model of KRR and GPR based on a ExpSineSquared kernel, which is suited
for learning periodic functions. The kernel’s hyperparameters control the smoothness (l) and periodicity of the kernel
(p). Moreover, the noise level of the data is learned explicitly by GPR by an additional WhiteKernel component in the
kernel and by the regularization parameter alpha of KRR.
The figure shows that both methods learn reasonable models of the target function. GPR correctly identifies the peri-
odicity of the function to be roughly 2*pi (6.28), while KRR chooses the doubled periodicity 4*pi. Besides that, GPR
provides reasonable confidence bounds on the prediction which are not available for KRR. A major difference between
the two methods is the time required for fitting and predicting: while fitting KRR is fast in principle, the grid-search
for hyperparameter optimization scales exponentially with the number of hyperparameters (“curse of dimensional-
ity”). The gradient-based optimization of the parameters in GPR does not suffer from this exponential scaling and is
thus considerable faster on this example with 3-dimensional hyperparameter space. The time for predicting is similar;
however, generating the variance of the predictive distribution of GPR takes considerable longer than just predicting
the mean.
Out:
Time for KRR fitting: 4.309

Time for GPR fitting: 0.387
Time for KRR prediction: 0.008

Time for GPR prediction: 0.010

Time for GPR prediction with standard-deviation: 0.061
print(__doc__)

import time
import numpy as np
from sklearn.kernel_ridge import KernelRidge

from sklearn.gaussian_process import GaussianProcessRegressor
from sklearn.gaussian_process.kernels import WhiteKernel, ExpSineSquared

X = 15 * rng.rand(100, 1)
y += 3 * (0.5 - rng.rand(X.shape[0])) # add noise
# Fit KernelRidge with parameter selection based on 5-fold cross validation

param_grid = {"alpha": [1e0, 1e-1, 1e-2, 1e-3],
"kernel": [ExpSineSquared(l, p)
for l in np.logspace(-2, 2, 10)
for p in np.logspace(0, 2, 10)]}
kr = GridSearchCV(KernelRidge(), cv=5, param_grid=param_grid)
stime = time.time()
kr.fit(X, y)
print("Time for KRR fitting: %.3f" % (time.time() - stime))
gp_kernel = ExpSineSquared(1.0, 5.0, periodicity_bounds=(1e-2, 1e1)) \

+ WhiteKernel(1e-1)
gpr = GaussianProcessRegressor(kernel=gp_kernel)
stime = time.time()
gpr.fit(X, y)
print("Time for GPR fitting: %.3f" % (time.time() - stime))
# Predict using kernel ridge

X_plot = np.linspace(0, 20, 10000)[:, None]
stime = time.time()
y_kr = kr.predict(X_plot)
print("Time for KRR prediction: %.3f" % (time.time() - stime))
# Predict using gaussian process regressor

stime = time.time()
y_gpr = gpr.predict(X_plot, return_std=False)

print("Time for GPR prediction: %.3f" % (time.time() - stime))
stime = time.time()
y_gpr, y_std = gpr.predict(X_plot, return_std=True)
print("Time for GPR prediction with standard-deviation: %.3f"
% (time.time() - stime))
# Plot results
lw = 2
plt.scatter(X, y, c='k', label='data')
plt.plot(X_plot, np.sin(X_plot), color='navy', lw=lw, label='True')
plt.plot(X_plot, y_kr, color='turquoise', lw=lw,
label='KRR (%s)' % kr.best_params_)
plt.plot(X_plot, y_gpr, color='darkorange', lw=lw,
label='GPR (%s)' % gpr.kernel_)
plt.fill_between(X_plot[:, 0], y_gpr - y_std, y_gpr + y_std, color='darkorange',
alpha=0.2)
plt.xlabel('data')
plt.ylabel('target')
plt.xlim(0, 20)
plt.ylim(-4, 4)
plt.title('GPR versus Kernel Ridge')
plt.legend(loc="best", scatterpoints=1, prop={'size': 8})
plt.show()
5.15.4 Illustration of prior and posterior Gaussian process for different kernels
This example illustrates the prior and posterior of a GPR with different kernels. Mean, standard deviation, and 10
samples are shown for both prior and posterior.





print(__doc__)

#
import numpy as np

from sklearn.gaussian_process.kernels import (RBF, Matern, RationalQuadratic,
ExpSineSquared, DotProduct,
ConstantKernel)
kernels = [1.0 * RBF(length_scale=1.0, length_scale_bounds=(1e-1, 10.0)),

1.0 * RationalQuadratic(length_scale=1.0, alpha=0.1),

1.0 * ExpSineSquared(length_scale=1.0, periodicity=3.0,
length_scale_bounds=(0.1, 10.0),
periodicity_bounds=(1.0, 10.0)),
ConstantKernel(0.1, (0.01, 10.0))
* (DotProduct(sigma_0=1.0, sigma_0_bounds=(0.1, 10.0)) ** 2),
1.0 * Matern(length_scale=1.0, length_scale_bounds=(1e-1, 10.0),
nu=1.5)]
for fig_index, kernel in enumerate(kernels):

# Specify Gaussian Process
gp = GaussianProcessRegressor(kernel=kernel)
# Plot prior
plt.figure(fig_index, figsize=(8, 8))
X_ = np.linspace(0, 5, 100)
y_mean, y_std = gp.predict(X_[:, np.newaxis], return_std=True)
plt.plot(X_, y_mean, 'k', lw=3, zorder=9)
plt.fill_between(X_, y_mean - y_std, y_mean + y_std,
alpha=0.2, color='k')
y_samples = gp.sample_y(X_[:, np.newaxis], 10)
plt.plot(X_, y_samples, lw=1)
plt.xlim(0, 5)
plt.ylim(-3, 3)
plt.title("Prior (kernel: %s)" % kernel, fontsize=12)
# Generate data and fit GP

X = rng.uniform(0, 5, 10)[:, np.newaxis]
y = np.sin((X[:, 0] - 2.5) ** 2)
gp.fit(X, y)
# Plot posterior
y_mean, y_std = gp.predict(X_[:, np.newaxis], return_std=True)
plt.fill_between(X_, y_mean - y_std, y_mean + y_std,
y_samples = gp.sample_y(X_[:, np.newaxis], 10)

plt.plot(X_, y_samples, lw=1)
plt.scatter(X[:, 0], y, c='r', s=50, zorder=10, edgecolors=(0, 0, 0))
plt.xlim(0, 5)
plt.ylim(-3, 3)
plt.title("Posterior (kernel: %s)\n Log-Likelihood: %.3f"
% (gp.kernel_, gp.log_marginal_likelihood(gp.kernel_.theta)),
fontsize=12)
plt.tight_layout()
plt.show()

5.15.5 Iso-probability lines for Gaussian Processes classification (GPC)
A two-dimensional classification example showing iso-probability lines for the predicted probabilities.
Out:
Learned kernel: 0.0256**2 * DotProduct(sigma_0=5.72) ** 2
print(__doc__)
# Author: Vincent Dubourg <vincent.dubourg@gmail.com>

# Adapted to GaussianProcessClassifier:
import numpy as np


from sklearn.gaussian_process.kernels import DotProduct, ConstantKernel as C
# A few constants
lim = 8
def g(x):
"""The function to predict (classification will then consist in predicting
whether g(x) <= 0 or not)"""
return 5. - x[:, 1] - .5 * x[:, 0] ** 2.
# Design of experiments
X = np.array([[-4.61611719, -6.00099547],
[4.10469096, 5.32782448],
[0.00000000, -0.50000000],
[-6.17289014, -4.6984743],
[1.3109306, -6.93271427],
[-5.03823144, 3.10584743],
[-2.87600388, 6.74310541],
[5.21301203, 4.26386883]])
# Observations
y = np.array(g(X) > 0, dtype=int)
# Instantiate and fit Gaussian Process Model

kernel = C(0.1, (1e-5, np.inf)) * DotProduct(sigma_0=0.1) ** 2
gp = GaussianProcessClassifier(kernel=kernel)
gp.fit(X, y)
print("Learned kernel: %s " % gp.kernel_)
# Evaluate real function and the predicted probability

res = 50
x1, x2 = np.meshgrid(np.linspace(- lim, lim, res),
np.linspace(- lim, lim, res))
xx = np.vstack([x1.reshape(x1.size), x2.reshape(x2.size)]).T
y_true = g(xx)
y_prob = gp.predict_proba(xx)[:, 1]
y_true = y_true.reshape((res, res))
y_prob = y_prob.reshape((res, res))
# Plot the probabilistic classification iso-values

fig = plt.figure(1)
ax = fig.gca()
ax.axes.set_aspect('equal')
plt.xticks([])
plt.yticks([])
ax.set_xticklabels([])
ax.set_yticklabels([])
plt.xlabel('$x_1$')
plt.ylabel('$x_2$')
cax = plt.imshow(y_prob, cmap=cm.gray_r, alpha=0.8,

extent=(-lim, lim, -lim, lim))
norm = plt.matplotlib.colors.Normalize(vmin=0., vmax=0.9)
cb = plt.colorbar(cax, ticks=[0., 0.2, 0.4, 0.6, 0.8, 1.], norm=norm)
cb.set_label(r'${\rm \mathbb{P}}\left[\widehat{G}(\mathbf{x}) \leq 0\right]$')
plt.clim(0, 1)

plt.plot(X[y <= 0, 0], X[y <= 0, 1], 'r.', markersize=12)
plt.plot(X[y > 0, 0], X[y > 0, 1], 'b.', markersize=12)
plt.contour(x1, x2, y_true, [0.], colors='k', linestyles='dashdot')
cs = plt.contour(x1, x2, y_prob, [0.666], colors='b',

linestyles='solid')
plt.clabel(cs, fontsize=11)
cs = plt.contour(x1, x2, y_prob, [0.5], colors='k',

linestyles='dashed')
cs = plt.contour(x1, x2, y_prob, [0.334], colors='r',

linestyles='solid')
plt.show()
5.15.6 Probabilistic predictions with Gaussian process classification (GPC)
This example illustrates the predicted probability of GPC for an RBF kernel with different choices of the hyperparam-
eters. The first figure shows the predicted probability of GPC with arbitrarily chosen hyperparameters and with the
hyperparameters corresponding to the maximum log-marginal-likelihood (LML).
While the hyperparameters chosen by optimizing LML have a considerable larger LML, they perform slightly worse
according to the log-loss on test data. The figure shows that this is because they exhibit a steep change of the class
probabilities at the class boundaries (which is good) but have predicted probabilities close to 0.5 far away from the
class boundaries (which is bad) This undesirable effect is caused by the Laplace approximation used internally by
GPC.
The second figure shows the log-marginal-likelihood for different choices of the kernel’s hyperparameters, highlighting
the two choices of the hyperparameters used in the first figure by black dots.


•
Out:
Log Marginal Likelihood (initial): -17.598

Log Marginal Likelihood (optimized): -3.875
Accuracy: 1.000 (initial) 1.000 (optimized)
Log-loss: 0.214 (initial) 0.319 (optimized)
print(__doc__)

#
import numpy as np
from sklearn.metrics.classification import accuracy_score, log_loss


# Generate data
train_size = 50
y = np.array(X[:, 0] > 2.5, dtype=int)
# Specify Gaussian Processes with fixed and optimized hyperparameters

gp_fix = GaussianProcessClassifier(kernel=1.0 * RBF(length_scale=1.0),
optimizer=None)
gp_fix.fit(X[:train_size], y[:train_size])
gp_opt = GaussianProcessClassifier(kernel=1.0 * RBF(length_scale=1.0))

gp_opt.fit(X[:train_size], y[:train_size])
print("Log Marginal Likelihood (initial): %.3f"

% gp_fix.log_marginal_likelihood(gp_fix.kernel_.theta))
print("Log Marginal Likelihood (optimized): %.3f"
% gp_opt.log_marginal_likelihood(gp_opt.kernel_.theta))
print("Accuracy: %.3f (initial) %.3f (optimized)"

% (accuracy_score(y[:train_size], gp_fix.predict(X[:train_size])),
accuracy_score(y[:train_size], gp_opt.predict(X[:train_size]))))
print("Log-loss: %.3f (initial) %.3f (optimized)"
% (log_loss(y[:train_size], gp_fix.predict_proba(X[:train_size])[:, 1]),
log_loss(y[:train_size], gp_opt.predict_proba(X[:train_size])[:, 1])))
# Plot posteriors
plt.figure(0)
plt.scatter(X[:train_size, 0], y[:train_size], c='k', label="Train data",
plt.scatter(X[train_size:, 0], y[train_size:], c='g', label="Test data",
plt.plot(X_, gp_fix.predict_proba(X_[:, np.newaxis])[:, 1], 'r',
label="Initial kernel: %s" % gp_fix.kernel_)
plt.plot(X_, gp_opt.predict_proba(X_[:, np.newaxis])[:, 1], 'b',
label="Optimized kernel: %s" % gp_opt.kernel_)
plt.xlabel("Feature")
plt.ylabel("Class 1 probability")
plt.xlim(0, 5)
plt.ylim(-0.25, 1.5)
# Plot LML landscape

plt.figure(1)
theta0 = np.logspace(0, 8, 30)
theta1 = np.logspace(-1, 1, 29)
Theta0, Theta1 = np.meshgrid(theta0, theta1)
LML = [[gp_opt.log_marginal_likelihood(np.log([Theta0[i, j], Theta1[i, j]]))
for i in range(Theta0.shape[0])] for j in range(Theta0.shape[1])]
LML = np.array(LML).T
plt.plot(np.exp(gp_fix.kernel_.theta)[0], np.exp(gp_fix.kernel_.theta)[1],
'ko', zorder=10)
plt.plot(np.exp(gp_opt.kernel_.theta)[0], np.exp(gp_opt.kernel_.theta)[1],
'ko', zorder=10)
plt.pcolor(Theta0, Theta1, LML)
plt.xscale("log")

plt.yscale("log")
plt.colorbar()
plt.xlabel("Magnitude")
plt.ylabel("Length-scale")
plt.title("Log-marginal-likelihood")
plt.show()
5.15.7 Gaussian process regression (GPR) with noise-level estimation
This example illustrates that GPR with a sum-kernel including a WhiteKernel can estimate the noise level of data.
An illustration of the log-marginal-likelihood (LML) landscape shows that there exist two local maxima of LML. The
first corresponds to a model with a high noise level and a large length scale, which explains all variations in the data
by noise. The second one has a smaller noise level and shorter length scale, which explains most of the variation by
the noise-free functional relationship. The second model has a higher likelihood; however, depending on the initial
value for the hyperparameters, the gradient-based optimization might also converge to the high-noise solution. It is
thus important to repeat the optimization several times for different initializations.


print(__doc__)

#
import numpy as np

from matplotlib.colors import LogNorm

from sklearn.gaussian_process.kernels import RBF, WhiteKernel
y = 0.5 * np.sin(3 * X[:, 0]) + rng.normal(0, 0.5, X.shape[0])
# First run
plt.figure(0)
kernel = 1.0 * RBF(length_scale=100.0, length_scale_bounds=(1e-2, 1e3)) \
+ WhiteKernel(noise_level=1, noise_level_bounds=(1e-10, 1e+1))
gp = GaussianProcessRegressor(kernel=kernel,
alpha=0.0).fit(X, y)
y_mean, y_cov = gp.predict(X_[:, np.newaxis], return_cov=True)


plt.fill_between(X_, y_mean - np.sqrt(np.diag(y_cov)),
y_mean + np.sqrt(np.diag(y_cov)),
plt.plot(X_, 0.5*np.sin(3*X_), 'r', lw=3, zorder=9)
plt.title("Initial: %s\nOptimum: %s\nLog-Marginal-Likelihood: %s"
% (kernel, gp.kernel_,
gp.log_marginal_likelihood(gp.kernel_.theta)))
plt.tight_layout()
# Second run
plt.figure(1)
kernel = 1.0 * RBF(length_scale=1.0, length_scale_bounds=(1e-2, 1e3)) \
+ WhiteKernel(noise_level=1e-5, noise_level_bounds=(1e-10, 1e+1))
gp = GaussianProcessRegressor(kernel=kernel,
alpha=0.0).fit(X, y)
y_mean, y_cov = gp.predict(X_[:, np.newaxis], return_cov=True)
plt.fill_between(X_, y_mean - np.sqrt(np.diag(y_cov)),
y_mean + np.sqrt(np.diag(y_cov)),
plt.plot(X_, 0.5*np.sin(3*X_), 'r', lw=3, zorder=9)
plt.title("Initial: %s\nOptimum: %s\nLog-Marginal-Likelihood: %s"
% (kernel, gp.kernel_,
gp.log_marginal_likelihood(gp.kernel_.theta)))
plt.tight_layout()
# Plot LML landscape

plt.figure(2)
Theta0, Theta1 = np.meshgrid(theta0, theta1)
LML = [[gp.log_marginal_likelihood(np.log([0.36, Theta0[i, j], Theta1[i, j]]))
for i in range(Theta0.shape[0])] for j in range(Theta0.shape[1])]
LML = np.array(LML).T
vmin, vmax = (-LML).min(), (-LML).max()

vmax = 50
level = np.around(np.logspace(np.log10(vmin), np.log10(vmax), 50), decimals=1)
plt.contour(Theta0, Theta1, -LML,
levels=level, norm=LogNorm(vmin=vmin, vmax=vmax))
plt.colorbar()
plt.xscale("log")
plt.yscale("log")
plt.xlabel("Length-scale")
plt.ylabel("Noise-level")
plt.title("Log-marginal-likelihood")
plt.tight_layout()
plt.show()

5.15.8 Gaussian Processes regression: basic introductory example
A simple one-dimensional regression example computed in two different ways:

1. A noise-free case
2. A noisy case with known noise-level per datapoint
In both cases, the kernel’s parameters are estimated using the maximum likelihood principle.
The figures illustrate the interpolating property of the Gaussian Process model as well as its probabilistic nature in the
form of a pointwise 95% confidence interval.
Note that the parameter alpha is applied as a Tikhonov regularization of the assumed covariance between the training
points.

print(__doc__)
# Author: Vincent Dubourg <vincent.dubourg@gmail.com>

# Jake Vanderplas <vanderplas@astro.washington.edu>
# Jan Hendrik Metzen <jhm@informatik.uni-bremen.de>s
import numpy as np

from sklearn.gaussian_process.kernels import RBF, ConstantKernel as C
np.random.seed(1)
def f(x):
"""The function to predict."""
# ----------------------------------------------------------------------
# First the noiseless case
X = np.atleast_2d([1., 3., 5., 6., 7., 8.]).T
# Observations
y = f(X).ravel()

# Mesh the input space for evaluations of the real function, the prediction and
# its MSE
x = np.atleast_2d(np.linspace(0, 10, 1000)).T
# Instantiate a Gaussian Process model

kernel = C(1.0, (1e-3, 1e3)) * RBF(10, (1e-2, 1e2))
gp = GaussianProcessRegressor(kernel=kernel, n_restarts_optimizer=9)
# Fit to data using Maximum Likelihood Estimation of the parameters

gp.fit(X, y)
# Make the prediction on the meshed x-axis (ask for MSE as well)
y_pred, sigma = gp.predict(x, return_std=True)
# the MSE
plt.figure()
plt.plot(x, f(x), 'r:', label=r'$f(x) = x\,\sin(x)$')
plt.plot(X, y, 'r.', markersize=10, label=u'Observations')
plt.plot(x, y_pred, 'b-', label=u'Prediction')
plt.fill(np.concatenate([x, x[::-1]]),
np.concatenate([y_pred - 1.9600 * sigma,
(y_pred + 1.9600 * sigma)[::-1]]),
alpha=.5, fc='b', ec='None', label='95% confidence interval')
plt.xlabel('$x$')
plt.ylim(-10, 20)
# ----------------------------------------------------------------------
# now the noisy case
X = np.linspace(0.1, 9.9, 20)
X = np.atleast_2d(X).T
# Observations and noise

y = f(X).ravel()
dy = 0.5 + 1.0 * np.random.random(y.shape)
noise = np.random.normal(0, dy)
y += noise
# Instantiate a Gaussian Process model

gp = GaussianProcessRegressor(kernel=kernel, alpha=dy ** 2,
n_restarts_optimizer=10)
# Fit to data using Maximum Likelihood Estimation of the parameters

gp.fit(X, y)
# Make the prediction on the meshed x-axis (ask for MSE as well)
y_pred, sigma = gp.predict(x, return_std=True)
# the MSE
plt.figure()
plt.plot(x, f(x), 'r:', label=r'$f(x) = x\,\sin(x)$')
plt.errorbar(X.ravel(), y, dy, fmt='r.', markersize=10, label=u'Observations')
plt.plot(x, y_pred, 'b-', label=u'Prediction')
plt.fill(np.concatenate([x, x[::-1]]),
np.concatenate([y_pred - 1.9600 * sigma,

(y_pred + 1.9600 * sigma)[::-1]]),

alpha=.5, fc='b', ec='None', label='95% confidence interval')
plt.xlabel('$x$')
plt.ylim(-10, 20)
plt.show()
5.15.9 Gaussian process regression (GPR) on Mauna Loa CO2 data.
This example is based on Section 5.4.3 of “Gaussian Processes for Machine Learning” [RW2006]. It illustrates an
example of complex kernel engineering and hyperparameter optimization using gradient ascent on the log-marginal-
likelihood. The data consists of the monthly average atmospheric CO2 concentrations (in parts per million by volume
(ppmv)) collected at the Mauna Loa Observatory in Hawaii, between 1958 and 2001. The objective is to model the
CO2 concentration as a function of the time t.
The kernel is composed of several terms that are responsible for explaining different properties of the signal:
• a long term, smooth rising trend is to be explained by an RBF kernel. The RBF kernel with a large length-scale
enforces this component to be smooth; it is not enforced that the trend is rising which leaves this choice to the
GP. The specific length-scale and the amplitude are free hyperparameters.
• a seasonal component, which is to be explained by the periodic ExpSineSquared kernel with a fixed periodicity
of 1 year. The length-scale of this periodic component, controlling its smoothness, is a free parameter. In order
to allow decaying away from exact periodicity, the product with an RBF kernel is taken. The length-scale of this
RBF component controls the decay time and is a further free parameter.
• smaller, medium term irregularities are to be explained by a RationalQuadratic kernel component, whose length-
scale and alpha parameter, which determines the diffuseness of the length-scales, are to be determined. Ac-
cording to [RW2006], these irregularities can better be explained by a RationalQuadratic than an RBF kernel
component, probably because it can accommodate several length-scales.
• a “noise” term, consisting of an RBF kernel contribution, which shall explain the correlated noise components
such as local weather phenomena, and a WhiteKernel contribution for the white noise. The relative amplitudes
and the RBF’s length scale are further free parameters.
Maximizing the log-marginal-likelihood after subtracting the target’s mean yields the following kernel with an LML
of -83.214:
34.4**2 * RBF(length_scale=41.8)
+ 3.27**2 * RBF(length_scale=180) * ExpSineSquared(length_scale=1.44,
periodicity=1)
+ 0.446**2 * RationalQuadratic(alpha=17.7, length_scale=0.957)
+ 0.197**2 * RBF(length_scale=0.138) + WhiteKernel(noise_level=0.0336)
Thus, most of the target signal (34.4ppm) is explained by a long-term rising trend (length-scale 41.8 years). The
periodic component has an amplitude of 3.27ppm, a decay time of 180 years and a length-scale of 1.44. The long
decay time indicates that we have a locally very close to periodic seasonal component. The correlated noise has an
amplitude of 0.197ppm with a length scale of 0.138 years and a white-noise contribution of 0.197ppm. Thus, the

overall noise level is very small, indicating that the data can be very well explained by the model. The figure shows
also that the model makes very confident predictions until around 2015.
Out:
GPML kernel: 66**2 * RBF(length_scale=67) + 2.4**2 * RBF(length_scale=90) *

˓→ExpSineSquared(length_scale=1.3, periodicity=1) + 0.66**2 *
˓→RationalQuadratic(alpha=0.78, length_scale=1.2) + 0.18**2 * RBF(length_scale=0.134)
˓→+ WhiteKernel(noise_level=0.0361)
Log-marginal-likelihood: -117.023
Learned kernel: 44.8**2 * RBF(length_scale=51.6) + 2.64**2 * RBF(length_scale=91.5) *

˓→ExpSineSquared(length_scale=1.48, periodicity=1) + 0.536**2 *
˓→RationalQuadratic(alpha=2.89, length_scale=0.968) + 0.188**2 * RBF(length_scale=0.
˓→122) + WhiteKernel(noise_level=0.0367)
Log-marginal-likelihood: -115.050

#

from __future__ import division, print_function
import numpy as np

from sklearn.datasets import fetch_openml
from sklearn.gaussian_process.kernels \
import RBF, WhiteKernel, RationalQuadratic, ExpSineSquared
print(__doc__)
def load_mauna_loa_atmospheric_co2():
ml_data = fetch_openml(data_id=41187)
months = []
ppmv_sums = []
counts = []
y = ml_data.data[:, 0]
m = ml_data.data[:, 1]
month_float = y + (m - 1) / 12
ppmvs = ml_data.target
for month, ppmv in zip(month_float, ppmvs):

if not months or month != months[-1]:
months.append(month)
ppmv_sums.append(ppmv)
counts.append(1)
else:
# aggregate monthly sum to produce average
ppmv_sums[-1] += ppmv
counts[-1] += 1
months = np.asarray(months).reshape(-1, 1)
avg_ppmvs = np.asarray(ppmv_sums) / counts
return months, avg_ppmvs
X, y = load_mauna_loa_atmospheric_co2()
# Kernel with parameters given in GPML book

k1 = 66.0**2 * RBF(length_scale=67.0) # long term smooth rising trend
k2 = 2.4**2 * RBF(length_scale=90.0) \
* ExpSineSquared(length_scale=1.3, periodicity=1.0) # seasonal component
# medium term irregularity
k3 = 0.66**2 \
* RationalQuadratic(length_scale=1.2, alpha=0.78)
+ WhiteKernel(noise_level=0.19**2) # noise terms
kernel_gpml = k1 + k2 + k3 + k4
gp = GaussianProcessRegressor(kernel=kernel_gpml, alpha=0,
optimizer=None, normalize_y=True)
gp.fit(X, y)
print("GPML kernel: %s" % gp.kernel_)

print("Log-marginal-likelihood: %.3f"

% gp.log_marginal_likelihood(gp.kernel_.theta))
# Kernel with optimized parameters

k1 = 50.0**2 * RBF(length_scale=50.0) # long term smooth rising trend
* ExpSineSquared(length_scale=1.0, periodicity=1.0,
periodicity_bounds="fixed") # seasonal component
# medium term irregularities
k3 = 0.5**2 * RationalQuadratic(length_scale=1.0, alpha=1.0)
+ WhiteKernel(noise_level=0.1**2,
noise_level_bounds=(1e-3, np.inf)) # noise terms
kernel = k1 + k2 + k3 + k4
gp = GaussianProcessRegressor(kernel=kernel, alpha=0,
normalize_y=True)
gp.fit(X, y)
print("\nLearned kernel: %s" % gp.kernel_)

print("Log-marginal-likelihood: %.3f"
% gp.log_marginal_likelihood(gp.kernel_.theta))
X_ = np.linspace(X.min(), X.max() + 30, 1000)[:, np.newaxis]

y_pred, y_std = gp.predict(X_, return_std=True)
# Illustration
plt.scatter(X, y, c='k')
plt.plot(X_, y_pred)
plt.fill_between(X_[:, 0], y_pred - y_std, y_pred + y_std,
plt.xlim(X_.min(), X_.max())
plt.xlabel("Year")
plt.ylabel(r"CO$_2$ in ppm")
plt.title(r"Atmospheric CO$_2$ concentration at Mauna Loa")
plt.tight_layout()
plt.show()
5.16 Generalized Linear Models
Examples concerning the sklearn.linear_model module.
5.16.1 Lasso path using LARS
Computes Lasso Path along the regularization parameter using the LARS algorithm on the diabetes dataset. Each
color represents a different feature of the coefficient vector, and this is displayed as a function of the regularization
parameter.
5.16. Generalized Linear Models 1077

Out:
Computing regularization path using the LARS ...

.
print(__doc__)
# Author: Fabian Pedregosa <fabian.pedregosa@inria.fr>

# Alexandre Gramfort <alexandre.gramfort@inria.fr>
import numpy as np
from sklearn import linear_model

X = diabetes.data
y = diabetes.target

print("Computing regularization path using the LARS ...")

_, _, coefs = linear_model.lars_path(X, y, method='lasso', verbose=True)
xx = np.sum(np.abs(coefs.T), axis=1)
xx /= xx[-1]
plt.plot(xx, coefs.T)
ymin, ymax = plt.ylim()
plt.vlines(xx, ymin, ymax, linestyle='dashed')
plt.xlabel('|coef| / max|coef|')
plt.ylabel('Coefficients')
plt.title('LASSO Path')
plt.axis('tight')
plt.show()
5.16.2 Plot Ridge coefficients as a function of the regularization
Shows the effect of collinearity in the coefficients of an estimator.

Ridge Regression is the estimator used in this example. Each color represents a different feature of the coefficient
vector, and this is displayed as a function of the regularization parameter.
This example also shows the usefulness of applying Ridge regression to highly ill-conditioned matrices. For such
matrices, a slight change in the target variable can cause huge variances in the calculated weights. In such cases, it is
useful to set a certain regularization (alpha) to reduce this variation (noise).
When alpha is very large, the regularization effect dominates the squared loss function and the coefficients tend to
zero. At the end of the path, as alpha tends toward zero and the solution tends towards the ordinary least squares,
coefficients exhibit big oscillations. In practise it is necessary to tune alpha in such a way that a balance is maintained
between both.

# Author: Fabian Pedregosa -- <fabian.pedregosa@inria.fr>

print(__doc__)
import numpy as np
# X is the 10x10 Hilbert matrix

X = 1. / (np.arange(1, 11) + np.arange(0, 10)[:, np.newaxis])
y = np.ones(10)
# #############################################################################
# Compute paths
n_alphas = 200
alphas = np.logspace(-10, -2, n_alphas)
coefs = []
for a in alphas:
ridge = linear_model.Ridge(alpha=a, fit_intercept=False)
ridge.fit(X, y)
coefs.append(ridge.coef_)
# #############################################################################

# Display results
ax = plt.gca()
ax.plot(alphas, coefs)
ax.set_xscale('log')
ax.set_xlim(ax.get_xlim()[::-1]) # reverse axis
plt.xlabel('alpha')
plt.ylabel('weights')
plt.title('Ridge coefficients as a function of the regularization')
plt.axis('tight')
plt.show()
5.16.3 SGD: Maximum margin separating hyperplane
Plot the maximum margin separating hyperplane within a two-class separable dataset using a linear Support Vector
Machines classifier trained using SGD.

print(__doc__)
import numpy as np
# we create 50 separable points

X, Y = make_blobs(n_samples=50, centers=2, random_state=0, cluster_std=0.60)
# fit the model

clf = SGDClassifier(loss="hinge", alpha=0.01, max_iter=200, fit_intercept=True)
clf.fit(X, Y)
# plot the line, the points, and the nearest vectors to the plane
xx = np.linspace(-1, 5, 10)
yy = np.linspace(-1, 5, 10)
X1, X2 = np.meshgrid(xx, yy)

Z = np.empty(X1.shape)
for (i, j), val in np.ndenumerate(X1):
x1 = val
x2 = X2[i, j]
p = clf.decision_function([[x1, x2]])
Z[i, j] = p[0]
levels = [-1.0, 0.0, 1.0]
linestyles = ['dashed', 'solid', 'dashed']
colors = 'k'
plt.contour(X1, X2, Z, levels, colors=colors, linestyles=linestyles)
plt.scatter(X[:, 0], X[:, 1], c=Y, cmap=plt.cm.Paired,
edgecolor='black', s=20)
plt.axis('tight')
plt.show()
5.16.4 SGD: convex loss functions
A plot that compares the various convex loss functions supported by sklearn.linear_model.
SGDClassifier .

print(__doc__)
import numpy as np
def modified_huber_loss(y_true, y_pred):

z = y_pred * y_true
loss = -4 * z
loss[z >= -1] = (1 - z[z >= -1]) ** 2
loss[z >= 1.] = 0
return loss
xmin, xmax = -4, 4

xx = np.linspace(xmin, xmax, 100)
lw = 2
plt.plot([xmin, 0, 0, xmax], [1, 1, 0, 0], color='gold', lw=lw,
label="Zero-one loss")
plt.plot(xx, np.where(xx < 1, 1 - xx, 0), color='teal', lw=lw,
label="Hinge loss")
plt.plot(xx, -np.minimum(xx, 0), color='yellowgreen', lw=lw,
label="Perceptron loss")
plt.plot(xx, np.log2(1 + np.exp(-xx)), color='cornflowerblue', lw=lw,
label="Log loss")
plt.plot(xx, np.where(xx < 1, 1 - xx, 0) ** 2, color='orange', lw=lw,

label="Squared hinge loss")

plt.plot(xx, modified_huber_loss(xx, 1), color='darkorchid', lw=lw,
linestyle='--', label="Modified Huber loss")
plt.ylim((0, 8))
plt.xlabel(r"Decision function $f(x)$")
plt.ylabel("$L(y=1, f(x))$")
plt.show()
5.16.5 Ordinary Least Squares and Ridge Regression Variance
Due to the few points in each dimension and the straight line that linear regression uses to follow these points as well
as it can, noise on the observations will cause great variance as shown in the first plot. Every line’s slope can vary
quite a bit for each prediction due to the noise induced in the observations.
Ridge regression is basically minimizing a penalised version of the least-squared function. The penalising shrinks the
value of the regression coefficients. Despite the few data points in each dimension, the slope of the prediction is much
more stable and the variance in the line itself is greatly reduced, in comparison to that of the standard linear regression

•
print(__doc__)

import numpy as np
X_train = np.c_[.5, 1].T

y_train = [.5, 1]
X_test = np.c_[0, 2].T
np.random.seed(0)
classifiers = dict(ols=linear_model.LinearRegression(),
ridge=linear_model.Ridge(alpha=.1))
for name, clf in classifiers.items():

for _ in range(6):
this_X = .1 * np.random.normal(size=(2, 1)) + X_train
clf.fit(this_X, y_train)
ax.plot(X_test, clf.predict(X_test), color='gray')

ax.scatter(this_X, y_train, s=3, c='gray', marker='o', zorder=10)
ax.plot(X_test, clf.predict(X_test), linewidth=2, color='blue')
ax.scatter(X_train, y_train, s=30, c='red', marker='+', zorder=10)
ax.set_title(name)
ax.set_xlim(0, 2)

ax.set_ylim((0, 1.6))
ax.set_xlabel('X')
ax.set_ylabel('y')
fig.tight_layout()
plt.show()
5.16.6 Plot Ridge coefficients as a function of the L2 regularization
Ridge Regression is the estimator used in this example. Each color in the left plot represents one different dimension
of the coefficient vector, and this is displayed as a function of the regularization parameter. The right plot shows
how exact the solution is. This example illustrates how a well defined solution is found by Ridge regression and
how regularization affects the coefficients and their values. The plot on the right shows how the difference of the
coefficients from the estimator changes as a function of regularization.
In this example the dependent variable Y is set as a function of the input features: y = X*w + c. The coefficient vector
w is randomly sampled from a normal distribution, whereas the bias term c is set to a constant.
As alpha tends toward zero the coefficients found by Ridge regression stabilize towards the randomly sampled vector
w. For big alpha (strong regularisation) the coefficients are smaller (eventually converging at 0) leading to a simpler
and biased solution. These dependencies can be observed on the left plot.
The right plot shows the mean squared error between the coefficients found by the model and the chosen vector w.
Less regularised models retrieve the exact coefficients (error is equal to 0), stronger regularised models increase the
error.
Please note that in this example the data is non-noisy, hence it is possible to extract the exact coefficients.
# Author: Kornel Kielczewski -- <kornel.k@plusnet.pl>
print(__doc__)

import numpy as np


clf = Ridge()
X, y, w = make_regression(n_samples=10, n_features=10, coef=True,

random_state=1, bias=3.5)
coefs = []
errors = []
alphas = np.logspace(-6, 6, 200)
# Train the model with different regularisation strengths

for a in alphas:
clf.set_params(alpha=a)
clf.fit(X, y)
coefs.append(clf.coef_)
errors.append(mean_squared_error(clf.coef_, w))
# Display results
plt.subplot(121)
ax = plt.gca()
ax.plot(alphas, coefs)
plt.xlabel('alpha')
plt.ylabel('weights')
plt.title('Ridge coefficients as a function of the regularization')
plt.axis('tight')
plt.subplot(122)
ax = plt.gca()
ax.plot(alphas, errors)
plt.xlabel('alpha')
plt.ylabel('error')
plt.title('Coefficient error as a function of the regularization')
plt.axis('tight')
plt.show()
5.16.7 SGD: Penalties
Contours of where the penalty is equal to 1 for the three penalties L1, L2 and elastic-net.
All of the above are supported by sklearn.linear_model.stochastic_gradient.

print(__doc__)
import numpy as np
l1_color = "navy"
l2_color = "c"
elastic_net_color = "darkorange"
line = np.linspace(-1.5, 1.5, 1001)

xx, yy = np.meshgrid(line, line)
l2 = xx ** 2 + yy ** 2
l1 = np.abs(xx) + np.abs(yy)
rho = 0.5

elastic_net = rho * l1 + (1 - rho) * l2
plt.figure(figsize=(10, 10), dpi=100)

ax = plt.gca()
elastic_net_contour = plt.contour(xx, yy, elastic_net, levels=[1],

colors=elastic_net_color)
l2_contour = plt.contour(xx, yy, l2, levels=[1], colors=l2_color)
l1_contour = plt.contour(xx, yy, l1, levels=[1], colors=l1_color)
ax.set_aspect("equal")
ax.spines['left'].set_position('center')
ax.spines['right'].set_color('none')
ax.spines['bottom'].set_position('center')
ax.spines['top'].set_color('none')
plt.clabel(elastic_net_contour, inline=1, fontsize=18,

fmt={1.0: 'elastic-net'}, manual=[(-1, -1)])
plt.clabel(l2_contour, inline=1, fontsize=18,
fmt={1.0: 'L2'}, manual=[(-1, -1)])
plt.clabel(l1_contour, inline=1, fontsize=18,
fmt={1.0: 'L1'}, manual=[(-1, -1)])
plt.tight_layout()
plt.show()
5.16.8 Regularization path of L1- Logistic Regression
Train l1-penalized logistic regression models on a binary classification problem derived from the Iris dataset.
The models are ordered from strongest regularized to least regularized. The 4 coefficients of the models are collected
and plotted as a “regularization path”: on the left-hand side of the figure (strong regularizers), all the coefficients are
exactly 0. When regularization gets progressively looser, coefficients can get non-zero values one after the other.
Here we choose the SAGA solver because it can efficiently optimize for the Logistic Regression loss with a non-
smooth, sparsity inducing l1 penalty.
Also note that we set a low value for the tolerance to make sure that the model has converged before collecting the
coefficients.
We also use warm_start=True which means that the coefficients of the models are reused to initialize the next model
fit to speed-up the computation of the full-path.

Out:
Computing regularization path ...

This took 3.311s
print(__doc__)


import numpy as np

from sklearn.svm import l1_min_c
X = iris.data
y = iris.target

X = X[y != 2]
y = y[y != 2]
X /= X.max() # Normalize X to speed-up convergence
# #############################################################################
# Demo path functions
cs = l1_min_c(X, y, loss='log') * np.logspace(0, 7, 16)
print("Computing regularization path ...")

start = time()
clf = linear_model.LogisticRegression(penalty='l1', solver='saga',
tol=1e-6, max_iter=int(1e6),
warm_start=True)
coefs_ = []
for c in cs:
clf.set_params(C=c)
clf.fit(X, y)
coefs_.append(clf.coef_.ravel().copy())
print("This took %0.3fs" % (time() - start))
coefs_ = np.array(coefs_)
plt.plot(np.log10(cs), coefs_, marker='o')
plt.xlabel('log(C)')
plt.ylabel('Coefficients')
plt.title('Logistic Regression Path')
plt.axis('tight')
plt.show()
5.16.9 Polynomial interpolation
This example demonstrates how to approximate a function with a polynomial of degree n_degree by using ridge
regression. Concretely, from n_samples 1d points, it suffices to build the Vandermonde matrix, which is n_samples x
n_degree+1 and has the following form:
[[1, x_1, x_1 ** 2, x_1 ** 3, . . . ], [1, x_2, x_2 ** 2, x_2 ** 3, . . . ], . . . ]
Intuitively, this matrix can be interpreted as a matrix of pseudo features (the points raised to some power). The matrix
is akin to (but different from) the matrix induced by a polynomial kernel.
This example shows that you can do non-linear regression with a linear model, using a pipeline to add non-linear
features. Kernel methods extend this idea and can induce very high (even infinite) dimensional feature spaces.

print(__doc__)
# Author: Mathieu Blondel

# Jake Vanderplas
import numpy as np

from sklearn.preprocessing import PolynomialFeatures
def f(x):
""" function to approximate by polynomial interpolation"""
# generate points used to plot

x_plot = np.linspace(0, 10, 100)
# generate points and keep a subset of them

x = np.linspace(0, 10, 100)
rng.shuffle(x)

x = np.sort(x[:20])
y = f(x)
# create matrix versions of these arrays

X = x[:, np.newaxis]
X_plot = x_plot[:, np.newaxis]
colors = ['teal', 'yellowgreen', 'gold']

lw = 2
plt.plot(x_plot, f(x_plot), color='cornflowerblue', linewidth=lw,
label="ground truth")
plt.scatter(x, y, color='navy', s=30, marker='o', label="training points")
for count, degree in enumerate([3, 4, 5]):

model = make_pipeline(PolynomialFeatures(degree), Ridge())
model.fit(X, y)
y_plot = model.predict(X_plot)
plt.plot(x_plot, y_plot, color=colors[count], linewidth=lw,
label="degree %d" % degree)
plt.legend(loc='lower left')
plt.show()
5.16.10 Logistic function
Shown in the plot is how the logistic regression would, in this synthetic dataset, classify values as either 0 or 1, i.e.
class one or two, using the logistic curve.

print(__doc__)
# Code source: Gael Varoquaux

import numpy as np
# General a toy dataset:s it's just a straight line with some Gaussian noise:
xmin, xmax = -5, 5
n_samples = 100
np.random.seed(0)
X = np.random.normal(size=n_samples)
y = (X > 0).astype(np.float)
X[X > 0] *= 4
X += .3 * np.random.normal(size=n_samples)
X = X[:, np.newaxis]
# Fit the classifier

clf = linear_model.LogisticRegression(C=1e5, solver='lbfgs')
clf.fit(X, y)
# and plot the result

plt.clf()
plt.scatter(X.ravel(), y, color='black', zorder=20)
X_test = np.linspace(-5, 10, 300)
def model(x):
return 1 / (1 + np.exp(-x))
loss = model(X_test * clf.coef_ + clf.intercept_).ravel()

plt.plot(X_test, loss, color='red', linewidth=3)
ols = linear_model.LinearRegression()
ols.fit(X, y)
plt.plot(X_test, ols.coef_ * X_test + ols.intercept_, linewidth=1)
plt.axhline(.5, color='.5')
plt.ylabel('y')
plt.xlabel('X')
plt.xticks(range(-5, 10))
plt.yticks([0, 0.5, 1])
plt.ylim(-.25, 1.25)
plt.xlim(-4, 10)
plt.legend(('Logistic Regression Model', 'Linear Regression Model'),
loc="lower right", fontsize='small')
plt.tight_layout()
plt.show()

5.16.11 SGD: Weighted samples
Plot decision function of a weighted dataset, where the size of points is proportional to its weight.
print(__doc__)
import numpy as np
# we create 20 points
np.random.seed(0)
X = np.r_[np.random.randn(10, 2) + [1, 1], np.random.randn(10, 2)]
y = [1] * 10 + [-1] * 10
sample_weight = 100 * np.abs(np.random.randn(20))
# and assign a bigger weight to the last 10 samples
sample_weight[:10] *= 10
# plot the weighted data points

plt.figure()

plt.scatter(X[:, 0], X[:, 1], c=y, s=sample_weight, alpha=0.9,

cmap=plt.cm.bone, edgecolor='black')
# fit the unweighted model

clf = linear_model.SGDClassifier(alpha=0.01, max_iter=100)
clf.fit(X, y)
no_weights = plt.contour(xx, yy, Z, levels=[0], linestyles=['solid'])
# fit the weighted model

clf = linear_model.SGDClassifier(alpha=0.01, max_iter=100)
clf.fit(X, y, sample_weight=sample_weight)
samples_weights = plt.contour(xx, yy, Z, levels=[0], linestyles=['dashed'])
plt.legend([no_weights.collections[0], samples_weights.collections[0]],
["no weights", "with weights"], loc="lower left")
plt.xticks(())
plt.yticks(())
plt.show()
5.16.12 Logistic Regression 3-class Classifier
Show below is a logistic-regression classifiers decision boundaries on the first two dimensions (sepal length and width)
of the iris dataset. The datapoints are colored according to their labels.

print(__doc__)

import numpy as np

Y = iris.target
logreg = LogisticRegression(C=1e5, solver='lbfgs', multi_class='multinomial')
# Create an instance of Logistic Regression Classifier and fit the data.

logreg.fit(X, Y)
Z = logreg.predict(np.c_[xx.ravel(), yy.ravel()])

plt.pcolormesh(xx, yy, Z, cmap=plt.cm.Paired)

plt.scatter(X[:, 0], X[:, 1], c=Y, edgecolors='k', cmap=plt.cm.Paired)
plt.xticks(())
plt.yticks(())
plt.show()
5.16.13 Linear Regression Example
This example uses the only the first feature of the diabetes dataset, in order to illustrate a two-dimensional plot of
this regression technique. The straight line can be seen in the plot, showing how linear regression attempts to draw a

straight line that will best minimize the residual sum of squares between the observed responses in the dataset, and the
responses predicted by the linear approximation.
The coefficients, the residual sum of squares and the variance score are also calculated.
Out:
Coefficients:
[938.23786125]
Mean squared error: 2548.07
Variance score: 0.47
print(__doc__)
# Code source: Jaques Grobler


import numpy as np
from sklearn import datasets, linear_model

from sklearn.metrics import mean_squared_error, r2_score
# Load the diabetes dataset

# Use only one feature

diabetes_X = diabetes.data[:, np.newaxis, 2]
# Split the data into training/testing sets

diabetes_X_train = diabetes_X[:-20]
diabetes_X_test = diabetes_X[-20:]
# Split the targets into training/testing sets

diabetes_y_train = diabetes.target[:-20]
diabetes_y_test = diabetes.target[-20:]
# Create linear regression object

regr = linear_model.LinearRegression()
# Train the model using the training sets

regr.fit(diabetes_X_train, diabetes_y_train)
# Make predictions using the testing set

diabetes_y_pred = regr.predict(diabetes_X_test)
# The coefficients
print('Coefficients: \n', regr.coef_)
# The mean squared error
print("Mean squared error: %.2f"
% mean_squared_error(diabetes_y_test, diabetes_y_pred))
# Explained variance score: 1 is perfect prediction
print('Variance score: %.2f' % r2_score(diabetes_y_test, diabetes_y_pred))
# Plot outputs
plt.scatter(diabetes_X_test, diabetes_y_test, color='black')
plt.plot(diabetes_X_test, diabetes_y_pred, color='blue', linewidth=3)
plt.xticks(())
plt.yticks(())
plt.show()
5.16.14 Robust linear model estimation using RANSAC
In this example we see how to robustly fit a linear model to faulty data using the RANSAC algorithm.

Out:
Estimated coefficients (true, linear regression, RANSAC):

82.1903908407869 [54.17236387] [82.08533159]
import numpy as np
from sklearn import linear_model, datasets
n_samples = 1000
n_outliers = 50
X, y, coef = datasets.make_regression(n_samples=n_samples, n_features=1,

n_informative=1, noise=10,
coef=True, random_state=0)
# Add outlier data

np.random.seed(0)

X[:n_outliers] = 3 + 0.5 * np.random.normal(size=(n_outliers, 1))

y[:n_outliers] = -3 + 10 * np.random.normal(size=n_outliers)
# Fit line using all data

lr = linear_model.LinearRegression()
lr.fit(X, y)
# Robustly fit linear model with RANSAC algorithm

ransac = linear_model.RANSACRegressor()
ransac.fit(X, y)
inlier_mask = ransac.inlier_mask_
outlier_mask = np.logical_not(inlier_mask)
# Predict data of estimated models

line_X = np.arange(X.min(), X.max())[:, np.newaxis]
line_y = lr.predict(line_X)
line_y_ransac = ransac.predict(line_X)
# Compare estimated coefficients

print("Estimated coefficients (true, linear regression, RANSAC):")
print(coef, lr.coef_, ransac.estimator_.coef_)
lw = 2
plt.scatter(X[inlier_mask], y[inlier_mask], color='yellowgreen', marker='.',
label='Inliers')
plt.scatter(X[outlier_mask], y[outlier_mask], color='gold', marker='.',
label='Outliers')
plt.plot(line_X, line_y, color='navy', linewidth=lw, label='Linear regressor')
plt.plot(line_X, line_y_ransac, color='cornflowerblue', linewidth=lw,
label='RANSAC regressor')
plt.legend(loc='lower right')
plt.xlabel("Input")
plt.ylabel("Response")
plt.show()
5.16.15 Sparsity Example: Fitting only features 1 and 2
Features 1 and 2 of the diabetes-dataset are fitted and plotted below. It illustrates that although feature 2 has a strong
coefficient on the full model, it does not give us much regarding y when compared to just feature 1


•
print(__doc__)


import numpy as np
from sklearn import datasets, linear_model
indices = (0, 1)
X_train = diabetes.data[:-20, indices]

X_test = diabetes.data[-20:, indices]
y_train = diabetes.target[:-20]
y_test = diabetes.target[-20:]
ols = linear_model.LinearRegression()
ols.fit(X_train, y_train)
# #############################################################################
# Plot the figure
def plot_figs(fig_num, elev, azim, X_train, clf):
fig = plt.figure(fig_num, figsize=(4, 3))
plt.clf()
ax = Axes3D(fig, elev=elev, azim=azim)
ax.scatter(X_train[:, 0], X_train[:, 1], y_train, c='k', marker='+')

ax.plot_surface(np.array([[-.1, -.1], [.15, .15]]),
np.array([[-.1, .15], [-.1, .15]]),
clf.predict(np.array([[-.1, -.1, .15, .15],
[-.1, .15, -.1, .15]]).T
).reshape((2, 2)),

alpha=.5)
ax.set_xlabel('X_1')
ax.set_ylabel('X_2')
ax.set_zlabel('Y')
#Generate the three different figures from different views

elev = 43.5
azim = -110
plot_figs(1, elev, azim, X_train, ols)
elev = -.5
azim = 0
elev = -.5
azim = 90
plt.show()
5.16.16 Lasso on dense and sparse data
We show that linear_model.Lasso provides the same results for dense and sparse data and that in the case of sparse
data the speed is improved.
Out:
--- Dense matrices
Sparse Lasso done in 0.198772s
Dense Lasso done in 0.115297s
Distance between coefficients : 9.043732562018544e-14
--- Sparse matrices
Matrix density : 0.6263000000000001 %
Sparse Lasso done in 0.250710s
Dense Lasso done in 1.030017s
Distance between coefficients : 7.344760355532163e-12
print(__doc__)


from sklearn.datasets.samples_generator import make_regression

# #############################################################################
# The two Lasso implementations on Dense data
print("--- Dense matrices")
X, y = make_regression(n_samples=200, n_features=5000, random_state=0)

X_sp = sparse.coo_matrix(X)
alpha = 1
sparse_lasso = Lasso(alpha=alpha, fit_intercept=False, max_iter=1000)
dense_lasso = Lasso(alpha=alpha, fit_intercept=False, max_iter=1000)
t0 = time()
sparse_lasso.fit(X_sp, y)
print("Sparse Lasso done in %fs" % (time() - t0))
t0 = time()
dense_lasso.fit(X, y)
print("Dense Lasso done in %fs" % (time() - t0))
print("Distance between coefficients : %s"

% linalg.norm(sparse_lasso.coef_ - dense_lasso.coef_))
# #############################################################################
# The two Lasso implementations on Sparse data
print("--- Sparse matrices")
Xs = X.copy()
Xs[Xs < 2.5] = 0.0
Xs = sparse.coo_matrix(Xs)
Xs = Xs.tocsc()
print("Matrix density : %s %%" % (Xs.nnz / float(X.size) * 100))
alpha = 0.1
sparse_lasso = Lasso(alpha=alpha, fit_intercept=False, max_iter=10000)
dense_lasso = Lasso(alpha=alpha, fit_intercept=False, max_iter=10000)
t0 = time()
sparse_lasso.fit(Xs, y)
print("Sparse Lasso done in %fs" % (time() - t0))
t0 = time()
dense_lasso.fit(Xs.toarray(), y)
print("Dense Lasso done in %fs" % (time() - t0))
print("Distance between coefficients : %s"

% linalg.norm(sparse_lasso.coef_ - dense_lasso.coef_))

5.16.17 HuberRegressor vs Ridge on dataset with strong outliers
Fit Ridge and HuberRegressor on a dataset with outliers.

The example shows that the predictions in ridge are strongly influenced by the outliers present in the dataset. The
Huber regressor is less influenced by the outliers since the model uses the linear loss for these. As the parameter
epsilon is increased for the Huber regressor, the decision function approaches that of the ridge.
# Authors: Manoj Kumar mks542@nyu.edu

print(__doc__)
import numpy as np

from sklearn.linear_model import HuberRegressor, Ridge
# Generate toy data.

X, y = make_regression(n_samples=20, n_features=1, random_state=0, noise=4.0,
bias=100.0)
# Add four strong outliers to the dataset.

X_outliers = rng.normal(0, 0.5, size=(4, 1))

y_outliers = rng.normal(0, 2.0, size=4)

X_outliers[:2, :] += X.max() + X.mean() / 4.
X_outliers[2:, :] += X.min() - X.mean() / 4.
y_outliers[:2] += y.min() - y.mean() / 4.
y_outliers[2:] += y.max() + y.mean() / 4.
X = np.vstack((X, X_outliers))
y = np.concatenate((y, y_outliers))
plt.plot(X, y, 'b.')
# Fit the huber regressor over a series of epsilon values.

colors = ['r-', 'b-', 'y-', 'm-']
x = np.linspace(X.min(), X.max(), 7)
epsilon_values = [1.35, 1.5, 1.75, 1.9]
for k, epsilon in enumerate(epsilon_values):
huber = HuberRegressor(fit_intercept=True, alpha=0.0, max_iter=100,
epsilon=epsilon)
huber.fit(X, y)
coef_ = huber.coef_ * x + huber.intercept_
plt.plot(x, coef_, colors[k], label="huber loss, %s" % epsilon)
# Fit a ridge regressor to compare it to huber regressor.

ridge = Ridge(fit_intercept=True, alpha=0.0, random_state=0, normalize=True)
ridge.fit(X, y)
coef_ridge = ridge.coef_
coef_ = ridge.coef_ * x + ridge.intercept_
plt.plot(x, coef_, 'g-', label="ridge regression")
plt.title("Comparison of HuberRegressor vs Ridge")

plt.xlabel("X")
plt.ylabel("y")
plt.legend(loc=0)
plt.show()
5.16.18 Comparing various online solvers
An example showing how different online solvers perform on the hand-written digits dataset.

Out:
training SGD
training ASGD
training Perceptron
training Passive-Aggressive I
training Passive-Aggressive II
training SAG
# Author: Rob Zinkov <rob at zinkov dot com>

import numpy as np

from sklearn.linear_model import SGDClassifier, Perceptron

heldout = [0.95, 0.90, 0.75, 0.50, 0.01]

rounds = 20
X, y = digits.data, digits.target
classifiers = [
("SGD", SGDClassifier(max_iter=100)),
("ASGD", SGDClassifier(average=True, max_iter=100)),
("Perceptron", Perceptron(tol=1e-3)),
("Passive-Aggressive I", PassiveAggressiveClassifier(loss='hinge',
C=1.0, tol=1e-4)),
("Passive-Aggressive II", PassiveAggressiveClassifier(loss='squared_hinge',
C=1.0, tol=1e-4)),
("SAG", LogisticRegression(solver='sag', tol=1e-1, C=1.e4 / X.shape[0]))
]
xx = 1. - np.array(heldout)
for name, clf in classifiers:

print("training %s" % name)
yy = []
for i in heldout:
yy_ = []
for r in range(rounds):
X_train, X_test, y_train, y_test = \
train_test_split(X, y, test_size=i, random_state=rng)
y_pred = clf.predict(X_test)
yy_.append(1 - np.mean(y_pred == y_test))
yy.append(np.mean(yy_))
plt.plot(xx, yy, label=name)
plt.xlabel("Proportion train")
plt.ylabel("Test Error Rate")
plt.show()
5.16.19 Joint feature selection with multi-task Lasso
The multi-task lasso allows to fit multiple regression problems jointly enforcing the selected features to be the same
across tasks. This example simulates sequential measurements, each task is a time instant, and the relevant features
vary in amplitude over time while being the same. The multi-task lasso imposes that features that are selected at one
time point are select for all time point. This makes feature selection by the Lasso more stable.


print(__doc__)


import numpy as np
from sklearn.linear_model import MultiTaskLasso, Lasso
# Generate some 2D coefficients with sine waves with random frequency and phase
n_samples, n_features, n_tasks = 100, 30, 40
n_relevant_features = 5
coef = np.zeros((n_tasks, n_features))
times = np.linspace(0, 2 * np.pi, n_tasks)
for k in range(n_relevant_features):
coef[:, k] = np.sin((1. + rng.randn(1)) * times + 3 * rng.randn(1))
X = rng.randn(n_samples, n_features)
Y = np.dot(X, coef.T) + rng.randn(n_samples, n_tasks)
coef_lasso_ = np.array([Lasso(alpha=0.5).fit(X, y).coef_ for y in Y.T])

coef_multi_task_lasso_ = MultiTaskLasso(alpha=1.).fit(X, Y).coef_
# #############################################################################
# Plot support and time series
plt.spy(coef_lasso_)
plt.xlabel('Feature')
plt.ylabel('Time (or Task)')
plt.text(10, 5, 'Lasso')
plt.spy(coef_multi_task_lasso_)
plt.xlabel('Feature')
plt.ylabel('Time (or Task)')
plt.text(10, 5, 'MultiTaskLasso')
fig.suptitle('Coefficient non-zero location')
feature_to_plot = 0
plt.figure()
lw = 2
plt.plot(coef[:, feature_to_plot], color='seagreen', linewidth=lw,
label='Ground truth')
plt.plot(coef_lasso_[:, feature_to_plot], color='cornflowerblue', linewidth=lw,
label='Lasso')
plt.plot(coef_multi_task_lasso_[:, feature_to_plot], color='gold', linewidth=lw,
label='MultiTaskLasso')
plt.legend(loc='upper center')
plt.axis('tight')
plt.ylim([-1.1, 1.1])
plt.show()

5.16.20 Lasso and Elastic Net for Sparse Signals
Estimates Lasso and Elastic-Net regression models on a manually generated sparse signal corrupted with an additive
noise. Estimated coefficients are compared with the ground-truth.
Out:

r^2 on test data : 0.385982
ElasticNet(alpha=0.1, copy_X=True, fit_intercept=True, l1_ratio=0.7,
random_state=None, selection='cyclic', tol=0.0001, warm_start=False)
r^2 on test data : 0.240498

print(__doc__)
import numpy as np
from sklearn.metrics import r2_score
# #############################################################################
# Generate some sparse data to play with
np.random.seed(42)
n_samples, n_features = 50, 200

X = np.random.randn(n_samples, n_features)
coef = 3 * np.random.randn(n_features)
inds = np.arange(n_features)
np.random.shuffle(inds)
coef[inds[10:]] = 0 # sparsify coef
y = np.dot(X, coef)
# add noise
y += 0.01 * np.random.normal(size=n_samples)
# Split data in train set and test set

n_samples = X.shape[0]
X_train, y_train = X[:n_samples // 2], y[:n_samples // 2]
X_test, y_test = X[n_samples // 2:], y[n_samples // 2:]
# #############################################################################
# Lasso
alpha = 0.1
lasso = Lasso(alpha=alpha)
y_pred_lasso = lasso.fit(X_train, y_train).predict(X_test)

r2_score_lasso = r2_score(y_test, y_pred_lasso)
print(lasso)
print("r^2 on test data : %f" % r2_score_lasso)
# #############################################################################
# ElasticNet
from sklearn.linear_model import ElasticNet
enet = ElasticNet(alpha=alpha, l1_ratio=0.7)
y_pred_enet = enet.fit(X_train, y_train).predict(X_test)

r2_score_enet = r2_score(y_test, y_pred_enet)
print(enet)
print("r^2 on test data : %f" % r2_score_enet)
plt.plot(enet.coef_, color='lightgreen', linewidth=2,

label='Elastic net coefficients')
plt.plot(lasso.coef_, color='gold', linewidth=2,
label='Lasso coefficients')
plt.plot(coef, '--', color='navy', label='original coefficients')
plt.title("Lasso R^2: %f, Elastic Net R^2: %f"
% (r2_score_lasso, r2_score_enet))

plt.show()
5.16.21 Orthogonal Matching Pursuit
Using orthogonal matching pursuit for recovering a sparse signal from a noisy measurement encoded with a dictionary
print(__doc__)

import numpy as np
from sklearn.linear_model import OrthogonalMatchingPursuit
from sklearn.linear_model import OrthogonalMatchingPursuitCV
from sklearn.datasets import make_sparse_coded_signal
n_components, n_features = 512, 100

n_nonzero_coefs = 17
# generate the data

###################
# y = Xw
# |x|_0 = n_nonzero_coefs
y, X, w = make_sparse_coded_signal(n_samples=1,
n_components=n_components,
n_features=n_features,
n_nonzero_coefs=n_nonzero_coefs,
random_state=0)
idx, = w.nonzero()
# distort the clean signal
y_noisy = y + 0.05 * np.random.randn(len(y))
# plot the sparse signal
plt.xlim(0, 512)
plt.title("Sparse signal")
plt.stem(idx, w[idx])
# plot the noise-free reconstruction

omp = OrthogonalMatchingPursuit(n_nonzero_coefs=n_nonzero_coefs)
omp.fit(X, y)
coef = omp.coef_
idx_r, = coef.nonzero()
plt.xlim(0, 512)
plt.title("Recovered signal from noise-free measurements")
plt.stem(idx_r, coef[idx_r])
# plot the noisy reconstruction

omp.fit(X, y_noisy)
coef = omp.coef_
plt.xlim(0, 512)
plt.title("Recovered signal from noisy measurements")
# plot the noisy reconstruction with number of non-zeros set by CV

omp_cv = OrthogonalMatchingPursuitCV(cv=5)
omp_cv.fit(X, y_noisy)
coef = omp_cv.coef_
plt.xlim(0, 512)
plt.title("Recovered signal from noisy measurements with CV")

plt.suptitle('Sparse signal recovery with Orthogonal Matching Pursuit',
fontsize=16)
plt.show()

5.16.22 MNIST classfification using multinomial logistic + L1
Here we fit a multinomial logistic regression with L1 penalty on a subset of the MNIST digits classification task. We
use the SAGA algorithm for this purpose: this a solver that is fast when the number of samples is significantly larger
than the number of features and is able to finely optimize non-smooth objective functions which is the case with the
l1-penalty. Test accuracy reaches > 0.8, while weight vectors remains sparse and therefore more easily interpretable.
Note that this accuracy of this l1-penalized linear model is significantly below what can be reached by an l2-penalized
linear model or a non-linear multi-layer perceptron model on this dataset.

Out:
Sparsity with L1 penalty: 80.89%

Test score with L1 penalty: 0.8309
Example run in 29.180 s
import time
import numpy as np

print(__doc__)
# Author: Arthur Mensch <arthur.mensch@m4x.org>

# Turn down for faster convergence

t0 = time.time()
train_samples = 5000
# Load data from https://www.openml.org/d/554

X, y = fetch_openml('mnist_784', version=1, return_X_y=True)
random_state = check_random_state(0)
permutation = random_state.permutation(X.shape[0])
X = X[permutation]

y = y[permutation]
X = X.reshape((X.shape[0], -1))

X, y, train_size=train_samples, test_size=10000)
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)
# Turn up tolerance for faster convergence

clf = LogisticRegression(C=50. / train_samples,
penalty='l1', solver='saga', tol=0.1)
sparsity = np.mean(clf.coef_ == 0) * 100
# print('Best C % .4f' % clf.C_)
print("Sparsity with L1 penalty: %.2f%%" % sparsity)
print("Test score with L1 penalty: %.4f" % score)
coef = clf.coef_.copy()
scale = np.abs(coef).max()
for i in range(10):
l1_plot = plt.subplot(2, 5, i + 1)
l1_plot.imshow(coef[i].reshape(28, 28), interpolation='nearest',
cmap=plt.cm.RdBu, vmin=-scale, vmax=scale)
l1_plot.set_xticks(())
l1_plot.set_yticks(())
l1_plot.set_xlabel('Class %i' % i)
plt.suptitle('Classification vector for...')
run_time = time.time() - t0
print('Example run in %.3f s' % run_time)
plt.show()
5.16.23 Plot multi-class SGD on the iris dataset
Plot decision surface of multi-class SGD on iris dataset. The hyperplanes corresponding to the three one-versus-all
(OVA) classifiers are represented by the dashed lines.

print(__doc__)
import numpy as np

# we only take the first two features. We could

# avoid this ugly slicing by using a two-dim dataset
X = iris.data[:, :2]
y = iris.target
colors = "bry"
# shuffle
idx = np.arange(X.shape[0])
np.random.seed(13)
np.random.shuffle(idx)
X = X[idx]
y = y[idx]
# standardize
mean = X.mean(axis=0)
std = X.std(axis=0)

X = (X - mean) / std
clf = SGDClassifier(alpha=0.001, max_iter=100).fit(X, y)

cs = plt.contourf(xx, yy, Z, cmap=plt.cm.Paired)
plt.axis('tight')

for i, color in zip(clf.classes_, colors):
plt.scatter(X[idx, 0], X[idx, 1], c=color, label=iris.target_names[i],
cmap=plt.cm.Paired, edgecolor='black', s=20)
plt.title("Decision surface of multi-class SGD")
plt.axis('tight')
# Plot the three one-against-all classifiers

xmin, xmax = plt.xlim()
coef = clf.coef_
intercept = clf.intercept_
def plot_hyperplane(c, color):

def line(x0):
return (-(x0 * coef[c, 0]) - intercept[c]) / coef[c, 1]
plt.plot([xmin, xmax], [line(xmin), line(xmax)],

ls="--", color=color)

plot_hyperplane(i, color)
plt.legend()
plt.show()

5.16.24 L1 Penalty and Sparsity in Logistic Regression
Comparison of the sparsity (percentage of zero coefficients) of solutions when L1 and L2 penalty are used for different
values of C. We can see that large values of C give more freedom to the model. Conversely, smaller values of C
constrain the model more. In the L1 penalty case, this leads to sparser solutions.
We classify 8x8 images of digits into two classes: 0-4 against 5-9. The visualization shows coefficients of the models
for varying C.
Out:
C=1.00
score with L1 penalty: 0.9082
C=0.10
C=0.01

print(__doc__)
# Authors: Alexandre Gramfort <alexandre.gramfort@inria.fr>

import numpy as np

# classify small against large digits

y = (y > 4).astype(np.int)
# Set regularization parameter

for i, C in enumerate((1, 0.1, 0.01)):
# turn down tolerance for short training time
clf_l1_LR = LogisticRegression(C=C, penalty='l1', tol=0.01, solver='saga')
clf_l2_LR = LogisticRegression(C=C, penalty='l2', tol=0.01, solver='saga')
clf_l1_LR.fit(X, y)
clf_l2_LR.fit(X, y)
coef_l1_LR = clf_l1_LR.coef_.ravel()
coef_l2_LR = clf_l2_LR.coef_.ravel()
# coef_l1_LR contains zeros due to the

# L1 sparsity inducing norm
sparsity_l1_LR = np.mean(coef_l1_LR == 0) * 100

sparsity_l2_LR = np.mean(coef_l2_LR == 0) * 100
print("C=%.2f" % C)
print("Sparsity with L1 penalty: %.2f%%" % sparsity_l1_LR)
print("score with L1 penalty: %.4f" % clf_l1_LR.score(X, y))
print("Sparsity with L2 penalty: %.2f%%" % sparsity_l2_LR)
print("score with L2 penalty: %.4f" % clf_l2_LR.score(X, y))
l1_plot = plt.subplot(3, 2, 2 * i + 1)
l2_plot = plt.subplot(3, 2, 2 * (i + 1))
if i == 0:
l1_plot.set_title("L1 penalty")
l2_plot.set_title("L2 penalty")
l1_plot.imshow(np.abs(coef_l1_LR.reshape(8, 8)), interpolation='nearest',

cmap='binary', vmax=1, vmin=0)
l2_plot.imshow(np.abs(coef_l2_LR.reshape(8, 8)), interpolation='nearest',

cmap='binary', vmax=1, vmin=0)

plt.text(-8, 3, "C = %.2f" % C)
plt.show()
5.16.25 Theil-Sen Regression
Computes a Theil-Sen Regression on a synthetic dataset.

See Theil-Sen estimator: generalized-median-based estimator for more information on the regressor.
Compared to the OLS (ordinary least squares) estimator, the Theil-Sen estimator is robust against outliers. It has
a breakdown point of about 29.3% in case of a simple linear regression which means that it can tolerate arbitrary
corrupted data (outliers) of up to 29.3% in the two-dimensional case.
The estimation of the model is done by calculating the slopes and intercepts of a subpopulation of all possible com-
binations of p subsample points. If an intercept is fitted, p must be greater than or equal to n_features + 1. The final
slope and intercept is then defined as the spatial median of these slopes and intercepts.
In certain cases Theil-Sen performs better than RANSAC which is also a robust method. This is illustrated in the second
example below where outliers with respect to the x-axis perturb RANSAC. Tuning the residual_threshold
parameter of RANSAC remedies this but in general a priori knowledge about the data and the nature of the outliers
is needed. Due to the computational complexity of Theil-Sen it is recommended to use it only for small problems in
terms of number of samples and features. For larger problems the max_subpopulation parameter restricts the
magnitude of all possible combinations of p subsample points to a randomly chosen subset and therefore also limits the
runtime. Therefore, Theil-Sen is applicable to larger problems with the drawback of losing some of its mathematical
properties since it then works on a random subset.


# Author: Florian Wilhelm -- <florian.wilhelm@gmail.com>

import time
import numpy as np
from sklearn.linear_model import LinearRegression, TheilSenRegressor
from sklearn.linear_model import RANSACRegressor
print(__doc__)
estimators = [('OLS', LinearRegression()),

('Theil-Sen', TheilSenRegressor(random_state=42)),
('RANSAC', RANSACRegressor(random_state=42)), ]
colors = {'OLS': 'turquoise', 'Theil-Sen': 'gold', 'RANSAC': 'lightgreen'}
lw = 2
# #############################################################################
# Outliers only in the y direction
np.random.seed(0)
n_samples = 200
# Linear model y = 3*x + N(2, 0.1**2)
x = np.random.randn(n_samples)
w = 3.
c = 2.
noise = 0.1 * np.random.randn(n_samples)

y = w * x + c + noise
# 10% outliers
y[-20:] += -20 * x[-20:]
plt.scatter(x, y, color='indigo', marker='x', s=40)

line_x = np.array([-3, 3])
for name, estimator in estimators:
t0 = time.time()
estimator.fit(X, y)
y_pred = estimator.predict(line_x.reshape(2, 1))
plt.plot(line_x, y_pred, color=colors[name], linewidth=lw,
label='%s (fit time: %.2fs)' % (name, elapsed_time))
plt.axis('tight')
plt.title("Corrupt y")
# #############################################################################
# Outliers in the X direction
np.random.seed(0)
# Linear model y = 3*x + N(2, 0.1**2)
x = np.random.randn(n_samples)
noise = 0.1 * np.random.randn(n_samples)
y = 3 * x + 2 + noise
# 10% outliers
x[-20:] = 9.9
y[-20:] += 22
plt.figure()
plt.scatter(x, y, color='indigo', marker='x', s=40)
line_x = np.array([-3, 10])

t0 = time.time()
estimator.fit(X, y)
y_pred = estimator.predict(line_x.reshape(2, 1))
plt.plot(line_x, y_pred, color=colors[name], linewidth=lw,
label='%s (fit time: %.2fs)' % (name, elapsed_time))
plt.axis('tight')
plt.title("Corrupt x")
plt.show()

5.16.26 Plot multinomial and One-vs-Rest Logistic Regression
Plot decision surface of multinomial and One-vs-Rest Logistic Regression. The hyperplanes corresponding to the
three One-vs-Rest (OVR) classifiers are represented by the dashed lines.

•
Out:
training score : 0.995 (multinomial)

training score : 0.976 (ovr)
print(__doc__)
# Authors: Tom Dupre la Tour <tom.dupre-la-tour@m4x.org>
import numpy as np
# make 3-class dataset for classification

centers = [[-5, 0], [0, 1.5], [5, -1]]
X, y = make_blobs(n_samples=1000, centers=centers, random_state=40)
transformation = [[0.4, 0.2], [-0.4, 1.2]]
X = np.dot(X, transformation)
for multi_class in ('multinomial', 'ovr'):

clf = LogisticRegression(solver='sag', max_iter=100, random_state=42,

multi_class=multi_class).fit(X, y)
# print the training scores

print("training score : %.3f (%s)" % (clf.score(X, y), multi_class))

plt.figure()
plt.contourf(xx, yy, Z, cmap=plt.cm.Paired)
plt.title("Decision surface of LogisticRegression (%s)" % multi_class)
plt.axis('tight')

colors = "bry"
plt.scatter(X[idx, 0], X[idx, 1], c=color, cmap=plt.cm.Paired,
edgecolor='black', s=20)
# Plot the three one-against-all classifiers

xmin, xmax = plt.xlim()
coef = clf.coef_
intercept = clf.intercept_
def plot_hyperplane(c, color):

def line(x0):
return (-(x0 * coef[c, 0]) - intercept[c]) / coef[c, 1]
plt.plot([xmin, xmax], [line(xmin), line(xmax)],
ls="--", color=color)

plot_hyperplane(i, color)
plt.show()
5.16.27 Robust linear estimator fitting
Here a sine function is fit with a polynomial of order 3, for values close to zero.
Robust fitting is demoed in different situations:

• No measurement errors, only modelling errors (fitting a sine with a polynomial)

• Measurement errors in X
• Measurement errors in y
The median absolute deviation to non corrupt new data is used to judge the quality of the prediction.
What we can see that:
• RANSAC is good for strong outliers in the y direction
• TheilSen is good for small outliers, both in direction X and y, but has a break point above which it performs
worse than OLS.
• The scores of HuberRegressor may not be compared directly to both TheilSen and RANSAC because it does
not attempt to completely filter the outliers but lessen their effect.



import numpy as np
from sklearn.linear_model import (

LinearRegression, TheilSenRegressor, RANSACRegressor, HuberRegressor)

np.random.seed(42)
X = np.random.normal(size=400)
y = np.sin(X)
# Make sure that it X is 2D
X = X[:, np.newaxis]
X_test = np.random.normal(size=200)
y_test = np.sin(X_test)
X_test = X_test[:, np.newaxis]
y_errors = y.copy()
y_errors[::3] = 3
X_errors = X.copy()
X_errors[::3] = 3
y_errors_large = y.copy()
y_errors_large[::3] = 10
X_errors_large = X.copy()
X_errors_large[::3] = 10
estimators = [('OLS', LinearRegression()),

('Theil-Sen', TheilSenRegressor(random_state=42)),
('RANSAC', RANSACRegressor(random_state=42)),
('HuberRegressor', HuberRegressor())]
colors = {'OLS': 'turquoise', 'Theil-Sen': 'gold', 'RANSAC': 'lightgreen',
˓→'HuberRegressor': 'black'}
linestyle = {'OLS': '-', 'Theil-Sen': '-.', 'RANSAC': '--', 'HuberRegressor': '--'}

lw = 3
x_plot = np.linspace(X.min(), X.max())

for title, this_X, this_y in [
('Modeling Errors Only', X, y),
('Corrupt X, Small Deviants', X_errors, y),
('Corrupt y, Small Deviants', X, y_errors),
('Corrupt X, Large Deviants', X_errors_large, y),
('Corrupt y, Large Deviants', X, y_errors_large)]:
plt.plot(this_X[:, 0], this_y, 'b+')

model = make_pipeline(PolynomialFeatures(3), estimator)
model.fit(this_X, this_y)
mse = mean_squared_error(model.predict(X_test), y_test)
y_plot = model.predict(x_plot[:, np.newaxis])
plt.plot(x_plot, y_plot, color=colors[name], linestyle=linestyle[name],
linewidth=lw, label='%s: error = %.3f' % (name, mse))
legend_title = 'Error of Mean\nAbsolute Deviation\nto Non-corrupt Data'

legend = plt.legend(loc='upper right', frameon=False, title=legend_title,
prop=dict(size='x-small'))

plt.xlim(-4, 10.2)
plt.ylim(-2, 10.2)
plt.title(title)
plt.show()
5.16.28 Lasso and Elastic Net
Lasso and elastic net (L1 and L2 penalisation) implemented using a coordinate descent.
The coefficients can be forced to be positive.


•
Out:
Computing regularization path using the lasso...

Computing regularization path using the positive lasso...
Computing regularization path using the elastic net...
Computing regularization path using the positive elastic net...
print(__doc__)


import numpy as np
from sklearn.linear_model import lasso_path, enet_path

X = diabetes.data
y = diabetes.target

X /= X.std(axis=0) # Standardize data (easier to set the l1_ratio parameter)
# Compute paths
eps = 5e-3 # the smaller it is the longer is the path
print("Computing regularization path using the lasso...")

alphas_lasso, coefs_lasso, _ = lasso_path(X, y, eps, fit_intercept=False)
print("Computing regularization path using the positive lasso...")

alphas_positive_lasso, coefs_positive_lasso, _ = lasso_path(
X, y, eps, positive=True, fit_intercept=False)
print("Computing regularization path using the elastic net...")
alphas_enet, coefs_enet, _ = enet_path(
X, y, eps=eps, l1_ratio=0.8, fit_intercept=False)
print("Computing regularization path using the positive elastic net...")

alphas_positive_enet, coefs_positive_enet, _ = enet_path(
X, y, eps=eps, l1_ratio=0.8, positive=True, fit_intercept=False)
# Display results
plt.figure(1)
colors = cycle(['b', 'r', 'g', 'c', 'k'])
neg_log_alphas_lasso = -np.log10(alphas_lasso)
neg_log_alphas_enet = -np.log10(alphas_enet)
for coef_l, coef_e, c in zip(coefs_lasso, coefs_enet, colors):
l1 = plt.plot(neg_log_alphas_lasso, coef_l, c=c)
l2 = plt.plot(neg_log_alphas_enet, coef_e, linestyle='--', c=c)
plt.xlabel('-Log(alpha)')
plt.ylabel('coefficients')
plt.title('Lasso and Elastic-Net Paths')
plt.legend((l1[-1], l2[-1]), ('Lasso', 'Elastic-Net'), loc='lower left')
plt.axis('tight')
plt.figure(2)
neg_log_alphas_positive_lasso = -np.log10(alphas_positive_lasso)
for coef_l, coef_pl, c in zip(coefs_lasso, coefs_positive_lasso, colors):
l1 = plt.plot(neg_log_alphas_lasso, coef_l, c=c)
l2 = plt.plot(neg_log_alphas_positive_lasso, coef_pl, linestyle='--', c=c)
plt.title('Lasso and positive Lasso')
plt.legend((l1[-1], l2[-1]), ('Lasso', 'positive Lasso'), loc='lower left')
plt.axis('tight')
plt.figure(3)
neg_log_alphas_positive_enet = -np.log10(alphas_positive_enet)
for (coef_e, coef_pe, c) in zip(coefs_enet, coefs_positive_enet, colors):
l1 = plt.plot(neg_log_alphas_enet, coef_e, c=c)
l2 = plt.plot(neg_log_alphas_positive_enet, coef_pe, linestyle='--', c=c)

plt.title('Elastic-Net and positive Elastic-Net')
plt.legend((l1[-1], l2[-1]), ('Elastic-Net', 'positive Elastic-Net'),
loc='lower left')
plt.axis('tight')
plt.show()
5.16.29 Automatic Relevance Determination Regression (ARD)
Fit regression model with Bayesian Ridge Regression.

See Bayesian Ridge Regression for more information on the regressor.
Compared to the OLS (ordinary least squares) estimator, the coefficient weights are slightly shifted toward zeros,
which stabilises them.
The histogram of the estimated weights is very peaked, as a sparsity-inducing prior is implied on the weights.
The estimation of the model is done by iteratively maximizing the marginal log-likelihood of the observations.
We also plot predictions and uncertainties for ARD for one dimensional regression using polynomial feature expan-
sion. Note the uncertainty starts going up on the right side of the plot. This is because these test samples are outside
of the range of the training samples.




•
print(__doc__)
import numpy as np
from sklearn.linear_model import ARDRegression, LinearRegression
# #############################################################################
# Generating simulated data with Gaussian weights
# Parameters of the example

np.random.seed(0)
# Create Gaussian data
X = np.random.randn(n_samples, n_features)
# Create weights with a precision lambda_ of 4.
lambda_ = 4.
w = np.zeros(n_features)
# Only keep 10 weights of interest
relevant_features = np.random.randint(0, n_features, 10)
for i in relevant_features:
w[i] = stats.norm.rvs(loc=0, scale=1. / np.sqrt(lambda_))
# Create noise with a precision alpha of 50.
alpha_ = 50.

noise = stats.norm.rvs(loc=0, scale=1. / np.sqrt(alpha_), size=n_samples)

# Create the target
y = np.dot(X, w) + noise
# #############################################################################
# Fit the ARD Regression
clf = ARDRegression(compute_score=True)
clf.fit(X, y)
ols = LinearRegression()
ols.fit(X, y)
# #############################################################################
# Plot the true weights, the estimated weights, the histogram of the
# weights, and predictions with standard deviations
plt.title("Weights of the model")
plt.plot(clf.coef_, color='darkblue', linestyle='-', linewidth=2,
label="ARD estimate")
plt.plot(ols.coef_, color='yellowgreen', linestyle=':', linewidth=2,
label="OLS estimate")
plt.plot(w, color='orange', linestyle='-', linewidth=2, label="Ground truth")
plt.xlabel("Features")
plt.ylabel("Values of the weights")
plt.legend(loc=1)
plt.title("Histogram of the weights")
plt.hist(clf.coef_, bins=n_features, color='navy', log=True)
plt.scatter(clf.coef_[relevant_features], np.full(len(relevant_features), 5.),
color='gold', marker='o', label="Relevant features")
plt.ylabel("Features")
plt.xlabel("Values of the weights")
plt.legend(loc=1)
plt.title("Marginal log-likelihood")
plt.plot(clf.scores_, color='navy', linewidth=2)
plt.ylabel("Score")
plt.xlabel("Iterations")
# Plotting some predictions for polynomial regression

def f(x, noise_amount):
y = np.sqrt(x) * np.sin(x)
noise = np.random.normal(0, 1, len(x))
return y + noise_amount * noise
degree = 10
X = np.linspace(0, 10, 100)
y = f(X, noise_amount=1)
clf_poly = ARDRegression(threshold_lambda=1e5)
clf_poly.fit(np.vander(X, degree), y)
X_plot = np.linspace(0, 11, 25)

y_plot = f(X_plot, noise_amount=0)
y_mean, y_std = clf_poly.predict(np.vander(X_plot, degree), return_std=True)

plt.errorbar(X_plot, y_mean, y_std, color='navy',
label="Polynomial ARD", linewidth=2)
plt.plot(X_plot, y_plot, color='gold', linewidth=2,
label="Ground Truth")
plt.ylabel("Output y")
plt.xlabel("Feature X")
plt.legend(loc="lower left")
plt.show()
5.16.30 Bayesian Ridge Regression
Computes a Bayesian Ridge Regression on a synthetic dataset.

See Bayesian Ridge Regression for more information on the regressor.
Compared to the OLS (ordinary least squares) estimator, the coefficient weights are slightly shifted toward zeros,
which stabilises them.
As the prior on the weights is a Gaussian prior, the histogram of the estimated weights is Gaussian.
The estimation of the model is done by iteratively maximizing the marginal log-likelihood of the observations.
We also plot predictions and uncertainties for Bayesian Ridge Regression for one dimensional regression using poly-
nomial feature expansion. Note the uncertainty starts going up on the right side of the plot. This is because these test
samples are outside of the range of the training samples.




•
print(__doc__)
import numpy as np
from sklearn.linear_model import BayesianRidge, LinearRegression
# #############################################################################
# Generating simulated data with Gaussian weights
np.random.seed(0)
X = np.random.randn(n_samples, n_features) # Create Gaussian data
# Create weights with a precision lambda_ of 4.
lambda_ = 4.
w = np.zeros(n_features)
# Only keep 10 weights of interest
relevant_features = np.random.randint(0, n_features, 10)
for i in relevant_features:
w[i] = stats.norm.rvs(loc=0, scale=1. / np.sqrt(lambda_))
# Create noise with a precision alpha of 50.
alpha_ = 50.
noise = stats.norm.rvs(loc=0, scale=1. / np.sqrt(alpha_), size=n_samples)
# Create the target
y = np.dot(X, w) + noise

# #############################################################################
# Fit the Bayesian Ridge Regression and an OLS for comparison
clf = BayesianRidge(compute_score=True)
clf.fit(X, y)
ols = LinearRegression()
ols.fit(X, y)
# #############################################################################
# Plot true weights, estimated weights, histogram of the weights, and
# predictions with standard deviations
lw = 2
plt.title("Weights of the model")
plt.plot(clf.coef_, color='lightgreen', linewidth=lw,
label="Bayesian Ridge estimate")
plt.plot(w, color='gold', linewidth=lw, label="Ground truth")
plt.plot(ols.coef_, color='navy', linestyle='--', label="OLS estimate")
plt.xlabel("Features")
plt.ylabel("Values of the weights")
plt.legend(loc="best", prop=dict(size=12))
plt.title("Histogram of the weights")
plt.hist(clf.coef_, bins=n_features, color='gold', log=True,
edgecolor='black')
plt.scatter(clf.coef_[relevant_features], np.full(len(relevant_features), 5.),
color='navy', label="Relevant features")
plt.ylabel("Features")
plt.xlabel("Values of the weights")
plt.title("Marginal log-likelihood")
plt.plot(clf.scores_, color='navy', linewidth=lw)
plt.ylabel("Score")
plt.xlabel("Iterations")
# Plotting some predictions for polynomial regression

def f(x, noise_amount):
y = np.sqrt(x) * np.sin(x)
noise = np.random.normal(0, 1, len(x))
return y + noise_amount * noise
degree = 10
X = np.linspace(0, 10, 100)
y = f(X, noise_amount=0.1)
clf_poly = BayesianRidge()
clf_poly.fit(np.vander(X, degree), y)
X_plot = np.linspace(0, 11, 25)

y_plot = f(X_plot, noise_amount=0)
y_mean, y_std = clf_poly.predict(np.vander(X_plot, degree), return_std=True)
plt.errorbar(X_plot, y_mean, y_std, color='navy',

label="Polynomial Bayesian Ridge Regression", linewidth=lw)

plt.plot(X_plot, y_plot, color='gold', linewidth=lw,
label="Ground Truth")
plt.ylabel("Output y")
plt.xlabel("Feature X")
plt.legend(loc="lower left")
plt.show()
5.16.31 Lasso model selection: Cross-Validation / AIC / BIC
Use the Akaike information criterion (AIC), the Bayes Information criterion (BIC) and cross-validation to select an
optimal value of the regularization parameter alpha of the Lasso estimator.
Results obtained with LassoLarsIC are based on AIC/BIC criteria.
Information-criterion based model selection is very fast, but it relies on a proper estimation of degrees of freedom, are
derived for large samples (asymptotic results) and assume the model is correct, i.e. that the data are actually generated
by this model. They also tend to break when the problem is badly conditioned (more features than samples).
For cross-validation, we use 20-fold with 2 algorithms to compute the Lasso path: coordinate descent, as implemented
by the LassoCV class, and Lars (least angle regression) as implemented by the LassoLarsCV class. Both algorithms
give roughly the same results. They differ with regards to their execution speed and sources of numerical errors.
Lars computes a path solution only for each kink in the path. As a result, it is very efficient when there are only of few
kinks, which is the case if there are few features or samples. Also, it is able to compute the full path without setting
any meta parameter. On the opposite, coordinate descent compute the path points on a pre-specified grid (here we use
the default). Thus it is more efficient if the number of grid points is smaller than the number of kinks in the path. Such
a strategy can be interesting if the number of features is really large and there are enough samples to select a large
amount. In terms of numerical errors, for heavily correlated variables, Lars will accumulate more errors, while the
coordinate descent algorithm will only sample the path on a grid.
Note how the optimal value of alpha varies for each fold. This illustrates why nested-cross validation is necessary
when trying to evaluate the performance of a method for which a parameter is chosen by cross-validation: this choice
of parameter may not be optimal for unseen data.



•
Out:
Computing regularization path using the coordinate descent lasso...

Computing regularization path using the Lars lasso...
print(__doc__)
# Author: Olivier Grisel, Gael Varoquaux, Alexandre Gramfort

import time
import numpy as np
from sklearn.linear_model import LassoCV, LassoLarsCV, LassoLarsIC

X = diabetes.data
y = diabetes.target

X = np.c_[X, rng.randn(X.shape[0], 14)] # add some bad features
# normalize data as done by Lars to allow for comparison

X /= np.sqrt(np.sum(X ** 2, axis=0))
# #############################################################################
# LassoLarsIC: least angle regression with BIC/AIC criterion
model_bic = LassoLarsIC(criterion='bic')
t1 = time.time()
model_bic.fit(X, y)
t_bic = time.time() - t1
alpha_bic_ = model_bic.alpha_
model_aic = LassoLarsIC(criterion='aic')
model_aic.fit(X, y)
alpha_aic_ = model_aic.alpha_
def plot_ic_criterion(model, name, color):

alpha_ = model.alpha_
alphas_ = model.alphas_
criterion_ = model.criterion_
plt.plot(-np.log10(alphas_), criterion_, '--', color=color,
linewidth=3, label='%s criterion' % name)
plt.axvline(-np.log10(alpha_), color=color, linewidth=3,
label='alpha: %s estimate' % name)
plt.xlabel('-log(alpha)')
plt.ylabel('criterion')
plt.figure()
plot_ic_criterion(model_aic, 'AIC', 'b')
plot_ic_criterion(model_bic, 'BIC', 'r')
plt.legend()
plt.title('Information-criterion for model selection (training time %.3fs)'
% t_bic)
# #############################################################################
# LassoCV: coordinate descent
# Compute paths
print("Computing regularization path using the coordinate descent lasso...")
t1 = time.time()
model = LassoCV(cv=20).fit(X, y)
t_lasso_cv = time.time() - t1
# Display results
m_log_alphas = -np.log10(model.alphas_)
plt.figure()
ymin, ymax = 2300, 3800
plt.plot(m_log_alphas, model.mse_path_, ':')
plt.plot(m_log_alphas, model.mse_path_.mean(axis=-1), 'k',
label='Average across the folds', linewidth=2)
plt.axvline(-np.log10(model.alpha_), linestyle='--', color='k',
label='alpha: CV estimate')

plt.legend()
plt.ylabel('Mean square error')
plt.title('Mean square error on each fold: coordinate descent '
'(train time: %.2fs)' % t_lasso_cv)
plt.axis('tight')
# #############################################################################
# LassoLarsCV: least angle regression
# Compute paths
print("Computing regularization path using the Lars lasso...")
t1 = time.time()
model = LassoLarsCV(cv=20).fit(X, y)
t_lasso_lars_cv = time.time() - t1
# Display results
m_log_alphas = -np.log10(model.cv_alphas_)
plt.figure()
plt.plot(m_log_alphas, model.mse_path_, ':')
plt.plot(m_log_alphas, model.mse_path_.mean(axis=-1), 'k',
label='Average across the folds', linewidth=2)
plt.axvline(-np.log10(model.alpha_), linestyle='--', color='k',
label='alpha CV')
plt.legend()
plt.ylabel('Mean square error')
plt.title('Mean square error on each fold: Lars (train time: %.2fs)'
% t_lasso_lars_cv)
plt.axis('tight')
plt.show()
5.16.32 Multiclass sparse logisitic regression on newgroups20
Comparison of multinomial logistic L1 vs one-versus-rest L1 logistic regression to classify documents from the new-
groups20 dataset. Multinomial logistic regression yields more accurate results and is faster to train on the larger scale
dataset.
Here we use the l1 sparsity that trims the weights of not informative features to zero. This is good if the goal is to
extract the strongly discriminative vocabulary of each class. If the goal is to get the best predictive accuracy, it is better
to use the non sparsity-inducing l2 penalty instead.
A more traditional (and possibly better) way to predict on a sparse subset of input features would be to use univariate
feature selection followed by a traditional (l2-penalised) logistic regression model.

Out:
Dataset 20newsgroup, train_samples=9000, n_features=130107, n_classes=20

[model=One versus Rest, solver=saga] Number of epochs: 1
[model=One versus Rest, solver=saga] Number of epochs: 3
Test accuracy for model ovr: 0.7410
% non-zero coefficients for model ovr, per class:
[0.27054655 0.66330021 0.80395367 0.73247404 0.67713497 0.73477984
0.40889422 0.48959702 1.01301237 0.56261385 0.60104376 0.332803
0.7094161 0.85083816 0.56876263 0.65715142 0.64408525 0.81163965
0.44271254 0.41120001]
Run time (3 epochs) for model ovr:2.47
[model=Multinomial, solver=saga] Number of epochs: 1
Test accuracy for model multinomial: 0.7450
% non-zero coefficients for model multinomial, per class:
[0.13296748 0.11759552 0.13296748 0.13988486 0.1268187 0.16140561
0.15218243 0.09069458 0.07762841 0.12143851 0.14910804 0.10837234
0.18830655 0.1245129 0.168323 0.21828188 0.11605832 0.07839701
0.06917383 0.15602543]
Run time (7 epochs) for model multinomial:5.66
Example run in 19.826 s

import time

import numpy as np
from sklearn.datasets import fetch_20newsgroups_vectorized

print(__doc__)
# Author: Arthur Mensch
t0 = time.clock()
# We use SAGA solver

solver = 'saga'
# Turn down for faster run time

n_samples = 10000
# Memorized fetch_rcv1 for faster access

dataset = fetch_20newsgroups_vectorized('all')
X = dataset.data
y = dataset.target
X = X[:n_samples]
y = y[:n_samples]
X_train, X_test, y_train, y_test = train_test_split(X, y,

random_state=42,
stratify=y,
test_size=0.1)
train_samples, n_features = X_train.shape
n_classes = np.unique(y).shape[0]
print('Dataset 20newsgroup, train_samples=%i, n_features=%i, n_classes=%i'

% (train_samples, n_features, n_classes))
models = {'ovr': {'name': 'One versus Rest', 'iters': [1, 3]},

'multinomial': {'name': 'Multinomial', 'iters': [1, 3, 7]}}

# Add initial chance-level values for plotting purpose
accuracies = [1 / n_classes]
times = [0]
densities = [1]
model_params = models[model]
# Small number of epochs for fast runtime

for this_max_iter in model_params['iters']:
print('[model=%s, solver=%s] Number of epochs: %s' %
(model_params['name'], solver, this_max_iter))
lr = LogisticRegression(solver=solver,
multi_class=model,
C=1,
penalty='l1',
fit_intercept=True,
max_iter=this_max_iter,

random_state=42,
)
t1 = time.clock()
lr.fit(X_train, y_train)
train_time = time.clock() - t1
y_pred = lr.predict(X_test)
accuracy = np.sum(y_pred == y_test) / y_test.shape[0]
density = np.mean(lr.coef_ != 0, axis=1) * 100
accuracies.append(accuracy)
densities.append(density)
times.append(train_time)
models[model]['times'] = times
models[model]['densities'] = densities
models[model]['accuracies'] = accuracies
print('Test accuracy for model %s: %.4f' % (model, accuracies[-1]))
print('%% non-zero coefficients for model %s, '
'per class:\n %s' % (model, densities[-1]))
print('Run time (%i epochs) for model %s:'
'%.2f' % (model_params['iters'][-1], model, times[-1]))
fig = plt.figure()

name = models[model]['name']
times = models[model]['times']
accuracies = models[model]['accuracies']
ax.plot(times, accuracies, marker='o',
label='Model: %s' % name)
ax.set_xlabel('Train time (s)')
ax.set_ylabel('Test accuracy')
ax.legend()
fig.suptitle('Multinomial vs One-vs-Rest Logistic L1\n'
'Dataset %s' % '20newsgroups')
fig.tight_layout()
fig.subplots_adjust(top=0.85)
run_time = time.clock() - t0
print('Example run in %.3f s' % run_time)
plt.show()
5.16.33 Early stopping of Stochastic Gradient Descent
Stochastic Gradient Descent is an optimization technique which minimizes a loss function in a stochastic fashion,
performing a gradient descent step sample by sample. In particular, it is a very efficient method to fit linear models.
As a stochastic method, the loss function is not necessarily decreasing at each iteration, and convergence is only
guaranteed in expectation. For this reason, monitoring the convergence on the loss function can be difficult.
Another approach is to monitor convergence on a validation score. In this case, the input data is split into a training set
and a validation set. The model is then fitted on the training set and the stopping criterion is based on the prediction

score computed on the validation set. This enables us to find the least number of iterations which is sufficient to build
a model that generalizes well to unseen data and reduces the chance of over-fitting the training data.
This early stopping strategy is activated if early_stopping=True; otherwise the stopping criterion only uses
the training loss on the entire input data. To better control the early stopping strategy, we can specify a parameter
validation_fraction which set the fraction of the input dataset that we keep aside to compute the validation
score. The optimization will continue until the validation score did not improve by at least tol during the last
n_iter_no_change iterations. The actual number of iterations is available at the attribute n_iter_.
This example illustrates how the early stopping can used in the sklearn.linear_model.SGDClassifier
model to achieve almost the same accuracy as compared to a model built without early stopping. This can significantly
reduce training time. Note that scores differ between the stopping criteria even from early iterations because some of
the training data is held out with the validation stopping criterion.
Out:
No stopping criterion: .................................................

Training loss: .................................................
Validation score: .................................................
# Authors: Tom Dupre la Tour

#

import time
import sys
import pandas as pd
import numpy as np

from sklearn.utils.testing import ignore_warnings
from sklearn.exceptions import ConvergenceWarning
print(__doc__)
def load_mnist(n_samples=None, class_0='0', class_1='8'):

"""Load MNIST, select two classes, shuffle and return only n_samples."""
# Load data from http://openml.org/d/554
mnist = fetch_openml('mnist_784', version=1)
# take only two classes for binary classification

mask = np.logical_or(mnist.target == class_0, mnist.target == class_1)
X, y = shuffle(mnist.data[mask], mnist.target[mask], random_state=42)

if n_samples is not None:
X, y = X[:n_samples], y[:n_samples]
return X, y
@ignore_warnings(category=ConvergenceWarning)
def fit_and_score(estimator, max_iter, X_train, X_test, y_train, y_test):
"""Fit the estimator on the train set and score it on both sets"""
estimator.set_params(max_iter=max_iter)
estimator.set_params(random_state=0)
start = time.time()
fit_time = time.time() - start

n_iter = estimator.n_iter_
train_score = estimator.score(X_train, y_train)
test_score = estimator.score(X_test, y_test)
return fit_time, n_iter, train_score, test_score
# Define the estimators to compare

estimator_dict = {
'No stopping criterion':
linear_model.SGDClassifier(tol=None, n_iter_no_change=3),
'Training loss':
linear_model.SGDClassifier(early_stopping=False, n_iter_no_change=3,
tol=0.1),
'Validation score':
linear_model.SGDClassifier(early_stopping=True, n_iter_no_change=3,
tol=0.0001, validation_fraction=0.2)

# Load the dataset

X, y = load_mnist(n_samples=10000)
random_state=0)
results = []
for estimator_name, estimator in estimator_dict.items():
print(estimator_name + ': ', end='')
for max_iter in range(1, 50):
print('.', end='')
sys.stdout.flush()
fit_time, n_iter, train_score, test_score = fit_and_score(

estimator, max_iter, X_train, X_test, y_train, y_test)
results.append((estimator_name, max_iter, fit_time, n_iter,

train_score, test_score))
print('')
# Transform the results in a pandas dataframe for easy plotting

columns = [
'Stopping criterion', 'max_iter', 'Fit time (sec)', 'n_iter_',
'Train score', 'Test score'
]
results_df = pd.DataFrame(results, columns=columns)
# Define what to plot (x_axis, y_axis)

lines = 'Stopping criterion'
plot_list = [
('max_iter', 'Train score'),
('max_iter', 'Test score'),
('max_iter', 'n_iter_'),
('max_iter', 'Fit time (sec)'),
]
nrows = 2
ncols = int(np.ceil(len(plot_list) / 2.))
fig, axes = plt.subplots(nrows=nrows, ncols=ncols, figsize=(6 * ncols,
4 * nrows))
axes[0, 0].get_shared_y_axes().join(axes[0, 0], axes[0, 1])
for ax, (x_axis, y_axis) in zip(axes.ravel(), plot_list):

for criterion, group_df in results_df.groupby(lines):
group_df.plot(x=x_axis, y=y_axis, label=criterion, ax=ax)
ax.set_title(y_axis)
ax.legend(title=lines)
fig.tight_layout()
plt.show()

5.17 Manifold learning
Examples concerning the sklearn.manifold module.
5.17.1 Swiss Roll reduction with LLE
An illustration of Swiss Roll reduction with locally linear embedding
Out:
Computing LLE embedding

Done. Reconstruction error: 7.43471e-08
# Author: Fabian Pedregosa -- <fabian.pedregosa@inria.fr>


print(__doc__)
# This import is needed to modify the way figure behaves

Axes3D
#----------------------------------------------------------------------
# Locally linear embedding of the swiss roll

X, color = datasets.samples_generator.make_swiss_roll(n_samples=1500)
print("Computing LLE embedding")

X_r, err = manifold.locally_linear_embedding(X, n_neighbors=12,
n_components=2)
print("Done. Reconstruction error: %g" % err)
#----------------------------------------------------------------------
# Plot result
fig = plt.figure()
ax = fig.add_subplot(211, projection='3d')
ax.scatter(X[:, 0], X[:, 1], X[:, 2], c=color, cmap=plt.cm.Spectral)
ax.set_title("Original data")
ax.scatter(X_r[:, 0], X_r[:, 1], c=color, cmap=plt.cm.Spectral)
plt.axis('tight')
plt.xticks([]), plt.yticks([])
plt.title('Projected data')
plt.show()
5.17.2 Multi-dimensional scaling
An illustration of the metric and non-metric MDS on generated noisy data.

The reconstructed points using the metric MDS and non metric MDS are slightly shifted to avoid overlapping.
5.17. Manifold learning 1165

# Author: Nelle Varoquaux <nelle.varoquaux@gmail.com>

# License: BSD
print(__doc__)
import numpy as np

from sklearn import manifold

from sklearn.metrics import euclidean_distances
n_samples = 20
seed = np.random.RandomState(seed=3)
X_true = seed.randint(0, 20, 2 * n_samples).astype(np.float)
X_true = X_true.reshape((n_samples, 2))
# Center the data
X_true -= X_true.mean()
similarities = euclidean_distances(X_true)
# Add noise to the similarities

noise = np.random.rand(n_samples, n_samples)
noise = noise + noise.T
noise[np.arange(noise.shape[0]), np.arange(noise.shape[0])] = 0

similarities += noise
mds = manifold.MDS(n_components=2, max_iter=3000, eps=1e-9, random_state=seed,

dissimilarity="precomputed", n_jobs=1)
pos = mds.fit(similarities).embedding_
nmds = manifold.MDS(n_components=2, metric=False, max_iter=3000, eps=1e-12,

dissimilarity="precomputed", random_state=seed, n_jobs=1,
n_init=1)
npos = nmds.fit_transform(similarities, init=pos)
# Rescale the data

pos *= np.sqrt((X_true ** 2).sum()) / np.sqrt((pos ** 2).sum())
npos *= np.sqrt((X_true ** 2).sum()) / np.sqrt((npos ** 2).sum())
# Rotate the data

clf = PCA(n_components=2)
X_true = clf.fit_transform(X_true)
pos = clf.fit_transform(pos)
npos = clf.fit_transform(npos)
fig = plt.figure(1)
ax = plt.axes([0., 0., 1., 1.])
s = 100
plt.scatter(X_true[:, 0], X_true[:, 1], color='navy', s=s, lw=0,
label='True Position')
plt.scatter(pos[:, 0], pos[:, 1], color='turquoise', s=s, lw=0, label='MDS')
plt.scatter(npos[:, 0], npos[:, 1], color='darkorange', s=s, lw=0, label='NMDS')
plt.legend(scatterpoints=1, loc='best', shadow=False)
similarities = similarities.max() / similarities * 100

similarities[np.isinf(similarities)] = 0
# Plot the edges

start_idx, end_idx = np.where(pos)
# a sequence of (*line0*, *line1*, *line2*), where::
# linen = (x0, y0), (x1, y1), ... (xm, ym)
segments = [[X_true[i, :], X_true[j, :]]
for i in range(len(pos)) for j in range(len(pos))]
values = np.abs(similarities)
lc = LineCollection(segments,
zorder=0, cmap=plt.cm.Blues,
norm=plt.Normalize(0, values.max()))
lc.set_array(similarities.flatten())
lc.set_linewidths(np.full(len(segments), 0.5))
ax.add_collection(lc)
plt.show()

5.17.3 t-SNE: The effect of various perplexity values on the shape
An illustration of t-SNE on the two concentric circles and the S-curve datasets for different perplexity values.
We observe a tendency towards clearer shapes as the preplexity value increases.
The size, the distance and the shape of clusters may vary upon initialization, perplexity values and does not always
convey a meaning.
As shown below, t-SNE for higher perplexities finds meaningful topology of two concentric circles, however the size
and the distance of the circles varies slightly from the original. Contrary to the two circles dataset, the shapes visually
diverge from S-curve topology on the S-curve dataset even for larger perplexity values.
For further details, “How to Use t-SNE Effectively” http://distill.pub/2016/misread-tsne/ provides a good discussion
of the effects of various parameters, as well as interactive plots to explore those effects.
Out:
circles, perplexity=5 in 0.84 sec

S-curve, perplexity=5 in 0.88 sec
uniform grid, perplexity=5 in 0.83 sec
uniform grid, perplexity=30 in 1 sec

# Author: Narine Kokhlikyan <narine@slice.com>

# License: BSD
print(__doc__)
import numpy as np
from matplotlib.ticker import NullFormatter

n_samples = 300
n_components = 2
(fig, subplots) = plt.subplots(3, 5, figsize=(15, 8))
perplexities = [5, 30, 50, 100]
X, y = datasets.make_circles(n_samples=n_samples, factor=.5, noise=.05)
red = y == 0
green = y == 1
ax = subplots[0][0]
ax.scatter(X[red, 0], X[red, 1], c="r")
ax.scatter(X[green, 0], X[green, 1], c="g")
ax.xaxis.set_major_formatter(NullFormatter())
ax.yaxis.set_major_formatter(NullFormatter())
plt.axis('tight')
for i, perplexity in enumerate(perplexities):

ax = subplots[0][i + 1]
t0 = time()
tsne = manifold.TSNE(n_components=n_components, init='random',
random_state=0, perplexity=perplexity)
Y = tsne.fit_transform(X)
t1 = time()
print("circles, perplexity=%d in %.2g sec" % (perplexity, t1 - t0))
ax.set_title("Perplexity=%d" % perplexity)
ax.scatter(Y[red, 0], Y[red, 1], c="r")
ax.scatter(Y[green, 0], Y[green, 1], c="g")
ax.axis('tight')
# Another example using s-curve

X, color = datasets.samples_generator.make_s_curve(n_samples, random_state=0)
ax = subplots[1][0]
ax.scatter(X[:, 0], X[:, 2], c=color)

t0 = time()

t1 = time()
print("S-curve, perplexity=%d in %.2g sec" % (perplexity, t1 - t0))
ax.scatter(Y[:, 0], Y[:, 1], c=color)
ax.axis('tight')
# Another example using a 2D uniform grid

x = np.linspace(0, 1, int(np.sqrt(n_samples)))
xx, yy = np.meshgrid(x, x)
X = np.hstack([
xx.ravel().reshape(-1, 1),
yy.ravel().reshape(-1, 1),
])
color = xx.ravel()
ax = subplots[2][0]
ax.scatter(X[:, 0], X[:, 1], c=color)

t0 = time()
t1 = time()
print("uniform grid, perplexity=%d in %.2g sec" % (perplexity, t1 - t0))
ax.scatter(Y[:, 0], Y[:, 1], c=color)
ax.axis('tight')
plt.show()
5.17.4 Comparison of Manifold Learning methods
An illustration of dimensionality reduction on the S-curve dataset with various manifold learning methods.
For a discussion and comparison of these algorithms, see the manifold module page
For a similar example, where the methods are applied to a sphere dataset, see Manifold Learning methods on a severed
sphere

Note that the purpose of the MDS is to find a low-dimensional representation of the data (here 2D) in which the
distances respect well the distances in the original high-dimensional space, unlike other manifold-learning algorithms,
it does not seeks an isotropic representation of the data in the low-dimensional space.
Out:
standard: 0.11 sec

ltsa: 0.19 sec
hessian: 0.37 sec
modified: 0.22 sec
Isomap: 0.34 sec
MDS: 2.5 sec
SpectralEmbedding: 0.16 sec
t-SNE: 5.8 sec
# Author: Jake Vanderplas -- <vanderplas@astro.washington.edu>
print(__doc__)

# Next line to silence pyflakes. This import is needed.

Axes3D

n_points = 1000
X, color = datasets.samples_generator.make_s_curve(n_points, random_state=0)
n_neighbors = 10
n_components = 2

plt.suptitle("Manifold Learning with %i points, %i neighbors"
% (1000, n_neighbors), fontsize=14)
ax.scatter(X[:, 0], X[:, 1], X[:, 2], c=color, cmap=plt.cm.Spectral)
methods = ['standard', 'ltsa', 'hessian', 'modified']

labels = ['LLE', 'LTSA', 'Hessian LLE', 'Modified LLE']
for i, method in enumerate(methods):

t0 = time()
Y = manifold.LocallyLinearEmbedding(n_neighbors, n_components,
eigen_solver='auto',
method=method).fit_transform(X)
t1 = time()
print("%s: %.2g sec" % (methods[i], t1 - t0))
ax = fig.add_subplot(252 + i)
plt.scatter(Y[:, 0], Y[:, 1], c=color, cmap=plt.cm.Spectral)
plt.title("%s (%.2g sec)" % (labels[i], t1 - t0))
plt.axis('tight')
t0 = time()
Y = manifold.Isomap(n_neighbors, n_components).fit_transform(X)
t1 = time()
print("Isomap: %.2g sec" % (t1 - t0))
plt.title("Isomap (%.2g sec)" % (t1 - t0))
plt.axis('tight')
t0 = time()
mds = manifold.MDS(n_components, max_iter=100, n_init=1)
Y = mds.fit_transform(X)
t1 = time()
print("MDS: %.2g sec" % (t1 - t0))
plt.title("MDS (%.2g sec)" % (t1 - t0))
plt.axis('tight')
t0 = time()

se = manifold.SpectralEmbedding(n_components=n_components,
n_neighbors=n_neighbors)
Y = se.fit_transform(X)
t1 = time()
print("SpectralEmbedding: %.2g sec" % (t1 - t0))
plt.title("SpectralEmbedding (%.2g sec)" % (t1 - t0))
plt.axis('tight')
t0 = time()
tsne = manifold.TSNE(n_components=n_components, init='pca', random_state=0)
t1 = time()
print("t-SNE: %.2g sec" % (t1 - t0))
plt.title("t-SNE (%.2g sec)" % (t1 - t0))
plt.axis('tight')
plt.show()
5.17.5 Manifold Learning methods on a severed sphere
An application of the different Manifold learning techniques on a spherical data-set. Here one can see the use of
dimensionality reduction in order to gain some intuition regarding the manifold learning methods. Regarding the
dataset, the poles are cut from the sphere, as well as a thin slice down its side. This enables the manifold learning
techniques to ‘spread it open’ whilst projecting it onto two dimensions.
For a similar example, where the methods are applied to the S-curve dataset, see Comparison of Manifold Learning
methods
Note that the purpose of the MDS is to find a low-dimensional representation of the data (here 2D) in which the
distances respect well the distances in the original high-dimensional space, unlike other manifold-learning algorithms,
it does not seeks an isotropic representation of the data in the low-dimensional space. Here the manifold problem
matches fairly that of representing a flat map of the Earth, as with map projection

Out:
standard: 0.079 sec
ltsa: 0.12 sec
hessian: 0.29 sec
modified: 0.15 sec
ISO: 0.19 sec
MDS: 1.2 sec
Spectral Embedding: 0.12 sec
t-SNE: 3.5 sec
# Author: Jaques Grobler <jaques.grobler@inria.fr>

print(__doc__)
import numpy as np
from sklearn import manifold

# Next line to silence pyflakes.

Axes3D
# Variables for manifold learning.

n_neighbors = 10
n_samples = 1000
# Create our sphere.

random_state = check_random_state(0)
p = random_state.rand(n_samples) * (2 * np.pi - 0.55)
t = random_state.rand(n_samples) * np.pi
# Sever the poles from the sphere.

indices = ((t < (np.pi - (np.pi / 8))) & (t > ((np.pi / 8))))
colors = p[indices]
x, y, z = np.sin(t[indices]) * np.cos(p[indices]), \
np.sin(t[indices]) * np.sin(p[indices]), \
np.cos(t[indices])
# Plot our dataset.

plt.suptitle("Manifold Learning with %i points, %i neighbors"
% (1000, n_neighbors), fontsize=14)
ax.scatter(x, y, z, c=p[indices], cmap=plt.cm.rainbow)
sphere_data = np.array([x, y, z]).T
# Perform Locally Linear Embedding Manifold learning

methods = ['standard', 'ltsa', 'hessian', 'modified']
labels = ['LLE', 'LTSA', 'Hessian LLE', 'Modified LLE']
for i, method in enumerate(methods):

t0 = time()
trans_data = manifold\
.LocallyLinearEmbedding(n_neighbors, 2,
method=method).fit_transform(sphere_data).T
t1 = time()
print("%s: %.2g sec" % (methods[i], t1 - t0))
ax = fig.add_subplot(252 + i)
plt.scatter(trans_data[0], trans_data[1], c=colors, cmap=plt.cm.rainbow)
plt.title("%s (%.2g sec)" % (labels[i], t1 - t0))
plt.axis('tight')
# Perform Isomap Manifold learning.

t0 = time()
trans_data = manifold.Isomap(n_neighbors, n_components=2)\
.fit_transform(sphere_data).T
t1 = time()
print("%s: %.2g sec" % ('ISO', t1 - t0))
plt.title("%s (%.2g sec)" % ('Isomap', t1 - t0))
plt.axis('tight')

# Perform Multi-dimensional scaling.

t0 = time()
mds = manifold.MDS(2, max_iter=100, n_init=1)
trans_data = mds.fit_transform(sphere_data).T
t1 = time()
print("MDS: %.2g sec" % (t1 - t0))
plt.title("MDS (%.2g sec)" % (t1 - t0))
plt.axis('tight')
# Perform Spectral Embedding.

t0 = time()
se = manifold.SpectralEmbedding(n_components=2,
n_neighbors=n_neighbors)
trans_data = se.fit_transform(sphere_data).T
t1 = time()
print("Spectral Embedding: %.2g sec" % (t1 - t0))
plt.title("Spectral Embedding (%.2g sec)" % (t1 - t0))
plt.axis('tight')
# Perform t-distributed stochastic neighbor embedding.

t0 = time()
tsne = manifold.TSNE(n_components=2, init='pca', random_state=0)
trans_data = tsne.fit_transform(sphere_data).T
t1 = time()
print("t-SNE: %.2g sec" % (t1 - t0))
plt.title("t-SNE (%.2g sec)" % (t1 - t0))
plt.axis('tight')
plt.show()
5.17.6 Manifold learning on handwritten digits: Locally Linear Embedding,

Isomap. . .
An illustration of various embeddings on the digits dataset.

The RandomTreesEmbedding, from the sklearn.ensemble module, is not technically a manifold embedding
method, as it learn a high-dimensional representation on which we apply a dimensionality reduction method. However,
it is often useful to cast a dataset into a representation in which the classes are linearly-separable.
t-SNE will be initialized with the embedding that is generated by PCA in this example, which is not the default setting.
It ensures global stability of the embedding, i.e., the embedding does not depend on random initialization.












•
Out:
Computing random projection

Computing PCA projection
Computing Linear Discriminant Analysis projection
Computing Isomap embedding
Done.
Computing LLE embedding
Done. Reconstruction error: 1.63544e-06
Computing modified LLE embedding
Done. Reconstruction error: 0.360651
Computing Hessian LLE embedding
Computing LTSA embedding
Computing MDS embedding
Done. Stress: 142429908.900591
Computing Totally Random Trees embedding
Computing Spectral embedding
Computing t-SNE embedding

# Authors: Fabian Pedregosa <fabian.pedregosa@inria.fr>

# Gael Varoquaux
print(__doc__)
import numpy as np
from matplotlib import offsetbox
from sklearn import (manifold, datasets, decomposition, ensemble,
discriminant_analysis, random_projection)
X = digits.data
y = digits.target
n_neighbors = 30
#----------------------------------------------------------------------
# Scale and visualize the embedding vectors
def plot_embedding(X, title=None):
x_min, x_max = np.min(X, 0), np.max(X, 0)
X = (X - x_min) / (x_max - x_min)
plt.figure()
for i in range(X.shape[0]):
plt.text(X[i, 0], X[i, 1], str(y[i]),
color=plt.cm.Set1(y[i] / 10.),
fontdict={'weight': 'bold', 'size': 9})
if hasattr(offsetbox, 'AnnotationBbox'):
# only print thumbnails with matplotlib > 1.0
shown_images = np.array([[1., 1.]]) # just something big
dist = np.sum((X[i] - shown_images) ** 2, 1)
if np.min(dist) < 4e-3:
# don't show points that are too close
continue
shown_images = np.r_[shown_images, [X[i]]]
imagebox = offsetbox.AnnotationBbox(
offsetbox.OffsetImage(digits.images[i], cmap=plt.cm.gray_r),
X[i])
ax.add_artist(imagebox)
plt.xticks([]), plt.yticks([])
if title is not None:
plt.title(title)
#----------------------------------------------------------------------
# Plot images of the digits
n_img_per_row = 20
img = np.zeros((10 * n_img_per_row, 10 * n_img_per_row))
for i in range(n_img_per_row):

ix = 10 * i + 1
for j in range(n_img_per_row):
iy = 10 * j + 1
img[ix:ix + 8, iy:iy + 8] = X[i * n_img_per_row + j].reshape((8, 8))
plt.imshow(img, cmap=plt.cm.binary)
plt.xticks([])
plt.yticks([])
plt.title('A selection from the 64-dimensional digits dataset')
#----------------------------------------------------------------------
# Random 2D projection using a random unitary matrix
print("Computing random projection")
rp = random_projection.SparseRandomProjection(n_components=2, random_state=42)
X_projected = rp.fit_transform(X)
plot_embedding(X_projected, "Random Projection of the digits")
#----------------------------------------------------------------------
# Projection on to the first 2 principal components
print("Computing PCA projection")

t0 = time()
X_pca = decomposition.TruncatedSVD(n_components=2).fit_transform(X)
plot_embedding(X_pca,
"Principal Components projection of the digits (time %.2fs)" %
(time() - t0))
#----------------------------------------------------------------------
# Projection on to the first 2 linear discriminant components
print("Computing Linear Discriminant Analysis projection")

X2 = X.copy()
X2.flat[::X.shape[1] + 1] += 0.01 # Make X invertible
t0 = time()
X_lda = discriminant_analysis.LinearDiscriminantAnalysis(n_components=2).fit_
˓→transform(X2, y)
plot_embedding(X_lda,
"Linear Discriminant projection of the digits (time %.2fs)" %
(time() - t0))
#----------------------------------------------------------------------
# Isomap projection of the digits dataset
print("Computing Isomap embedding")
t0 = time()
X_iso = manifold.Isomap(n_neighbors, n_components=2).fit_transform(X)
print("Done.")
plot_embedding(X_iso,
"Isomap projection of the digits (time %.2fs)" %
(time() - t0))
#----------------------------------------------------------------------
# Locally linear embedding of the digits dataset
print("Computing LLE embedding")
clf = manifold.LocallyLinearEmbedding(n_neighbors, n_components=2,

method='standard')
t0 = time()
X_lle = clf.fit_transform(X)
print("Done. Reconstruction error: %g" % clf.reconstruction_error_)
plot_embedding(X_lle,
"Locally Linear Embedding of the digits (time %.2fs)" %
(time() - t0))
#----------------------------------------------------------------------
# Modified Locally linear embedding of the digits dataset
print("Computing modified LLE embedding")
method='modified')
t0 = time()
X_mlle = clf.fit_transform(X)
plot_embedding(X_mlle,
"Modified Locally Linear Embedding of the digits (time %.2fs)" %
(time() - t0))
#----------------------------------------------------------------------
# HLLE embedding of the digits dataset
print("Computing Hessian LLE embedding")
method='hessian')
t0 = time()
X_hlle = clf.fit_transform(X)
plot_embedding(X_hlle,
"Hessian Locally Linear Embedding of the digits (time %.2fs)" %
(time() - t0))
#----------------------------------------------------------------------
# LTSA embedding of the digits dataset
print("Computing LTSA embedding")
method='ltsa')
t0 = time()
X_ltsa = clf.fit_transform(X)
plot_embedding(X_ltsa,
"Local Tangent Space Alignment of the digits (time %.2fs)" %
(time() - t0))
#----------------------------------------------------------------------
# MDS embedding of the digits dataset
print("Computing MDS embedding")
clf = manifold.MDS(n_components=2, n_init=1, max_iter=100)
t0 = time()
X_mds = clf.fit_transform(X)
print("Done. Stress: %f" % clf.stress_)
plot_embedding(X_mds,
"MDS embedding of the digits (time %.2fs)" %
(time() - t0))

#----------------------------------------------------------------------
# Random Trees embedding of the digits dataset
print("Computing Totally Random Trees embedding")
hasher = ensemble.RandomTreesEmbedding(n_estimators=200, random_state=0,
max_depth=5)
t0 = time()
X_transformed = hasher.fit_transform(X)
pca = decomposition.TruncatedSVD(n_components=2)
X_reduced = pca.fit_transform(X_transformed)
plot_embedding(X_reduced,
"Random forest embedding of the digits (time %.2fs)" %
(time() - t0))
#----------------------------------------------------------------------
# Spectral embedding of the digits dataset
print("Computing Spectral embedding")
embedder = manifold.SpectralEmbedding(n_components=2, random_state=0,
eigen_solver="arpack")
t0 = time()
X_se = embedder.fit_transform(X)
plot_embedding(X_se,
"Spectral embedding of the digits (time %.2fs)" %
(time() - t0))
#----------------------------------------------------------------------
# t-SNE embedding of the digits dataset
print("Computing t-SNE embedding")
tsne = manifold.TSNE(n_components=2, init='pca', random_state=0)
t0 = time()
X_tsne = tsne.fit_transform(X)
plot_embedding(X_tsne,
"t-SNE embedding of the digits (time %.2fs)" %
(time() - t0))
plt.show()
5.18 Gaussian Mixture Models
Examples concerning the sklearn.mixture module.
5.18.1 Density Estimation for a Gaussian mixture
Plot the density estimation of a mixture of two Gaussians. Data is generated from two Gaussians with different centers
and covariance matrices.
5.18. Gaussian Mixture Models 1193

import numpy as np
from matplotlib.colors import LogNorm
from sklearn import mixture
n_samples = 300
# generate random sample, two components

np.random.seed(0)
# generate spherical data centered on (20, 20)

shifted_gaussian = np.random.randn(n_samples, 2) + np.array([20, 20])
# generate zero centered stretched Gaussian data

C = np.array([[0., -0.7], [3.5, .7]])
stretched_gaussian = np.dot(np.random.randn(n_samples, 2), C)
# concatenate the two datasets into the final training set

X_train = np.vstack([shifted_gaussian, stretched_gaussian])
# fit a Gaussian Mixture Model with two components

clf = mixture.GaussianMixture(n_components=2, covariance_type='full')
clf.fit(X_train)
# display predicted scores by the model as a contour plot

x = np.linspace(-20., 30.)

y = np.linspace(-20., 40.)
X, Y = np.meshgrid(x, y)
XX = np.array([X.ravel(), Y.ravel()]).T
Z = -clf.score_samples(XX)
Z = Z.reshape(X.shape)
CS = plt.contour(X, Y, Z, norm=LogNorm(vmin=1.0, vmax=1000.0),

levels=np.logspace(0, 3, 10))
CB = plt.colorbar(CS, shrink=0.8, extend='both')
plt.scatter(X_train[:, 0], X_train[:, 1], .8)
plt.title('Negative log-likelihood predicted by a GMM')

plt.axis('tight')
plt.show()
5.18.2 Gaussian Mixture Model Ellipsoids
Plot the confidence ellipsoids of a mixture of two Gaussians obtained with Expectation Maximisation
(GaussianMixture class) and Variational Inference (BayesianGaussianMixture class models with a
Dirichlet process prior).
Both models have access to five components with which to fit the data. Note that the Expectation Maximisation
model will necessarily use all five components while the Variational Inference model will effectively only use as many
as are needed for a good fit. Here we can see that the Expectation Maximisation model splits some components
arbitrarily, because it is trying to fit too many components, while the Dirichlet Process model adapts it number of state
automatically.
This example doesn’t show it, as we’re in a low-dimensional space, but another advantage of the Dirichlet process
model is that it can fit full covariance matrices effectively even when there are less examples per cluster than there are
dimensions in the data, due to regularization properties of the inference algorithm.

import itertools
import numpy as np
color_iter = itertools.cycle(['navy', 'c', 'cornflowerblue', 'gold',

'darkorange'])
def plot_results(X, Y_, means, covariances, index, title):

splot = plt.subplot(2, 1, 1 + index)
for i, (mean, covar, color) in enumerate(zip(
means, covariances, color_iter)):
v, w = linalg.eigh(covar)
v = 2. * np.sqrt(2.) * np.sqrt(v)
# as the DP will not use every component it has access to
# unless it needs it, we shouldn't plot the redundant
# components.
if not np.any(Y_ == i):
continue
plt.scatter(X[Y_ == i, 0], X[Y_ == i, 1], .8, color=color)

# Plot an ellipse to show the Gaussian component

angle = 180. * angle / np.pi # convert to degrees
ell = mpl.patches.Ellipse(mean, v[0], v[1], 180. + angle, color=color)
ell.set_alpha(0.5)
plt.xlim(-9., 5.)
plt.ylim(-3., 6.)
plt.xticks(())
plt.yticks(())
plt.title(title)
# Number of samples per component

n_samples = 500
# Generate random sample, two components

np.random.seed(0)
C = np.array([[0., -0.1], [1.7, .4]])
X = np.r_[np.dot(np.random.randn(n_samples, 2), C),
.7 * np.random.randn(n_samples, 2) + np.array([-6, 3])]
# Fit a Gaussian mixture with EM using five components

gmm = mixture.GaussianMixture(n_components=5, covariance_type='full').fit(X)
plot_results(X, gmm.predict(X), gmm.means_, gmm.covariances_, 0,
'Gaussian Mixture')
# Fit a Dirichlet process Gaussian mixture using five components

dpgmm = mixture.BayesianGaussianMixture(n_components=5,
covariance_type='full').fit(X)
plot_results(X, dpgmm.predict(X), dpgmm.means_, dpgmm.covariances_, 1,
'Bayesian Gaussian Mixture with a Dirichlet process prior')
plt.show()
5.18.3 Gaussian Mixture Model Selection
This example shows that model selection can be performed with Gaussian Mixture Models using information-theoretic
criteria (BIC). Model selection concerns both the covariance type and the number of components in the model. In that
case, AIC also provides the right result (not shown to save time), but BIC is better suited if the problem is to identify
the right model. Unlike Bayesian procedures, such inferences are prior-free.
In that case, the model with 2 components and full covariance (which corresponds to the true generative model) is
selected.

import numpy as np
import itertools

print(__doc__)
# Number of samples per component

n_samples = 500
# Generate random sample, two components

np.random.seed(0)
C = np.array([[0., -0.1], [1.7, .4]])
X = np.r_[np.dot(np.random.randn(n_samples, 2), C),
.7 * np.random.randn(n_samples, 2) + np.array([-6, 3])]
lowest_bic = np.infty
bic = []
n_components_range = range(1, 7)
cv_types = ['spherical', 'tied', 'diag', 'full']
for cv_type in cv_types:
for n_components in n_components_range:

# Fit a Gaussian mixture with EM

gmm = mixture.GaussianMixture(n_components=n_components,
covariance_type=cv_type)
gmm.fit(X)
bic.append(gmm.bic(X))
if bic[-1] < lowest_bic:
lowest_bic = bic[-1]
best_gmm = gmm
bic = np.array(bic)
color_iter = itertools.cycle(['navy', 'turquoise', 'cornflowerblue',
'darkorange'])
clf = best_gmm
bars = []
# Plot the BIC scores

spl = plt.subplot(2, 1, 1)
for i, (cv_type, color) in enumerate(zip(cv_types, color_iter)):
xpos = np.array(n_components_range) + .2 * (i - 2)
bars.append(plt.bar(xpos, bic[i * len(n_components_range):
(i + 1) * len(n_components_range)],
width=.2, color=color))
plt.xticks(n_components_range)
plt.ylim([bic.min() * 1.01 - .01 * bic.max(), bic.max()])
plt.title('BIC score per model')
xpos = np.mod(bic.argmin(), len(n_components_range)) + .65 +\
.2 * np.floor(bic.argmin() / len(n_components_range))
plt.text(xpos, bic.min() * 0.97 + .03 * bic.max(), '*', fontsize=14)
spl.set_xlabel('Number of components')
spl.legend([b[0] for b in bars], cv_types)
# Plot the winner

splot = plt.subplot(2, 1, 2)
Y_ = clf.predict(X)
for i, (mean, cov, color) in enumerate(zip(clf.means_, clf.covariances_,
color_iter)):
v, w = linalg.eigh(cov)
if not np.any(Y_ == i):
continue
plt.scatter(X[Y_ == i, 0], X[Y_ == i, 1], .8, color=color)

angle = np.arctan2(w[0][1], w[0][0])
ell.set_alpha(.5)
plt.xticks(())
plt.yticks(())
plt.title('Selected GMM: full model, 2 components')
plt.subplots_adjust(hspace=.35, bottom=.02)
plt.show()

5.18.4 GMM covariances
Demonstration of several covariances types for Gaussian mixture models.

See Gaussian mixture models for more information on the estimator.
Although GMM are often used for clustering, we can compare the obtained clusters with the actual classes from the
dataset. We initialize the means of the Gaussians with the means of the classes from the training set to make this
comparison valid.
We plot predicted labels on both training and held out test data using a variety of GMM covariance types on the
iris dataset. We compare GMMs with spherical, diagonal, full, and tied covariance matrices in increasing order of
performance. Although one would expect full covariance to perform best in general, it is prone to overfitting on small
datasets and does not generalize well to held out test data.
On the plots, train data is shown as dots, while test data is shown as crosses. The iris dataset is four-dimensional. Only
the first two dimensions are shown here, and thus some points are separated in other dimensions.

# Author: Ron Weiss <ronweiss@gmail.com>, Gael Varoquaux

# Modified by Thierry Guillemot <thierry.guillemot.work@gmail.com>

import numpy as np

from sklearn.mixture import GaussianMixture
print(__doc__)

def make_ellipses(gmm, ax):

for n, color in enumerate(colors):
if gmm.covariance_type == 'full':
covariances = gmm.covariances_[n][:2, :2]
elif gmm.covariance_type == 'tied':
covariances = gmm.covariances_[:2, :2]
elif gmm.covariance_type == 'diag':
covariances = np.diag(gmm.covariances_[n][:2])
elif gmm.covariance_type == 'spherical':
covariances = np.eye(gmm.means_.shape[1]) * gmm.covariances_[n]
v, w = np.linalg.eigh(covariances)
u = w[0] / np.linalg.norm(w[0])
angle = np.arctan2(u[1], u[0])
angle = 180 * angle / np.pi # convert to degrees
ell = mpl.patches.Ellipse(gmm.means_[n, :2], v[0], v[1],
180 + angle, color=color)
ell.set_clip_box(ax.bbox)
ell.set_alpha(0.5)
ax.add_artist(ell)
ax.set_aspect('equal', 'datalim')
# Break up the dataset into non-overlapping training (75%) and testing

# (25%) sets.
skf = StratifiedKFold(n_splits=4)
# Only take the first fold.
train_index, test_index = next(iter(skf.split(iris.data, iris.target)))
X_train = iris.data[train_index]
y_train = iris.target[train_index]
X_test = iris.data[test_index]
y_test = iris.target[test_index]
n_classes = len(np.unique(y_train))
# Try GMMs using different types of covariances.

estimators = dict((cov_type, GaussianMixture(n_components=n_classes,
covariance_type=cov_type, max_iter=20, random_state=0))
for cov_type in ['spherical', 'diag', 'tied', 'full'])
n_estimators = len(estimators)
plt.figure(figsize=(3 * n_estimators // 2, 6))

plt.subplots_adjust(bottom=.01, top=0.95, hspace=.15, wspace=.05,
left=.01, right=.99)
for index, (name, estimator) in enumerate(estimators.items()):

# Since we have class labels for the training data, we can
# initialize the GMM parameters in a supervised manner.
estimator.means_init = np.array([X_train[y_train == i].mean(axis=0)
for i in range(n_classes)])
# Train the other parameters using the EM algorithm.

estimator.fit(X_train)

h = plt.subplot(2, n_estimators // 2, index + 1)

make_ellipses(estimator, h)

data = iris.data[iris.target == n]
plt.scatter(data[:, 0], data[:, 1], s=0.8, color=color,
label=iris.target_names[n])
# Plot the test data with crosses
data = X_test[y_test == n]
plt.scatter(data[:, 0], data[:, 1], marker='x', color=color)
y_train_pred = estimator.predict(X_train)
train_accuracy = np.mean(y_train_pred.ravel() == y_train.ravel()) * 100
plt.text(0.05, 0.9, 'Train accuracy: %.1f' % train_accuracy,
transform=h.transAxes)
y_test_pred = estimator.predict(X_test)
test_accuracy = np.mean(y_test_pred.ravel() == y_test.ravel()) * 100
plt.text(0.05, 0.8, 'Test accuracy: %.1f' % test_accuracy,
transform=h.transAxes)
plt.xticks(())
plt.yticks(())
plt.title(name)
plt.legend(scatterpoints=1, loc='lower right', prop=dict(size=12))
plt.show()
5.18.5 Gaussian Mixture Model Sine Curve
This example demonstrates the behavior of Gaussian mixture models fit on data that was not sampled from a mixture
of Gaussian random variables. The dataset is formed by 100 points loosely spaced following a noisy sine curve. There
is therefore no ground truth value for the number of Gaussian components.
The first model is a classical Gaussian Mixture Model with 10 components fit with the Expectation-Maximization
algorithm.
The second model is a Bayesian Gaussian Mixture Model with a Dirichlet process prior fit with variational inference.
The low value of the concentration prior makes the model favor a lower number of active components. This models
“decides” to focus its modeling power on the big picture of the structure of the dataset: groups of points with alternating
directions modeled by non-diagonal covariance matrices. Those alternating directions roughly capture the alternating
nature of the original sine signal.
The third model is also a Bayesian Gaussian mixture model with a Dirichlet process prior but this time the value of the
concentration prior is higher giving the model more liberty to model the fine-grained structure of the data. The result
is a mixture with a larger number of active components that is similar to the first model where we arbitrarily decided
to fix the number of components to 10.

Which model is the best is a matter of subjective judgement: do we want to favor models that only capture the big
picture to summarize and explain most of the structure of the data while ignoring the details or do we prefer models
that closely follow the high density regions of the signal?
The last two panels show how we can sample from the last two models. The resulting samples distributions do not
look exactly like the original data distribution. The difference primarily stems from the approximation error we made
by using a model that assumes that the data was generated by a finite number of Gaussian components instead of a
continuous noisy sine curve.
import itertools
import numpy as np

print(__doc__)
color_iter = itertools.cycle(['navy', 'c', 'cornflowerblue', 'gold',

'darkorange'])
def plot_results(X, Y, means, covariances, index, title):

splot = plt.subplot(5, 1, 1 + index)
for i, (mean, covar, color) in enumerate(zip(
means, covariances, color_iter)):
v, w = linalg.eigh(covar)
# components.
if not np.any(Y == i):
continue
plt.scatter(X[Y == i, 0], X[Y == i, 1], .8, color=color)

ell.set_alpha(0.5)
plt.xlim(-6., 4. * np.pi - 6.)

plt.ylim(-5., 5.)
plt.title(title)
plt.xticks(())
plt.yticks(())
def plot_samples(X, Y, n_components, index, title):

plt.subplot(5, 1, 4 + index)
for i, color in zip(range(n_components), color_iter):
# components.
if not np.any(Y == i):
continue
plt.scatter(X[Y == i, 0], X[Y == i, 1], .8, color=color)
plt.xlim(-6., 4. * np.pi - 6.)

plt.ylim(-5., 5.)
plt.title(title)
plt.xticks(())
plt.yticks(())
# Parameters
n_samples = 100

# Generate random sample following a sine curve

np.random.seed(0)
X = np.zeros((n_samples, 2))
step = 4. * np.pi / n_samples
x = i * step - 6.
X[i, 0] = x + np.random.normal(0, 0.1)
X[i, 1] = 3. * (np.sin(x) + np.random.normal(0, .2))
# Fit a Gaussian mixture with EM using ten components

gmm = mixture.GaussianMixture(n_components=10, covariance_type='full',
max_iter=100).fit(X)
plot_results(X, gmm.predict(X), gmm.means_, gmm.covariances_, 0,
'Expectation-maximization')
dpgmm = mixture.BayesianGaussianMixture(
n_components=10, covariance_type='full', weight_concentration_prior=1e-2,
weight_concentration_prior_type='dirichlet_process',
mean_precision_prior=1e-2, covariance_prior=1e0 * np.eye(2),
init_params="random", max_iter=100, random_state=2).fit(X)
"Bayesian Gaussian mixture models with a Dirichlet process prior "
r"for $\gamma_0=0.01$.")
X_s, y_s = dpgmm.sample(n_samples=2000)

plot_samples(X_s, y_s, dpgmm.n_components, 0,
"Gaussian mixture with a Dirichlet process prior "
r"for $\gamma_0=0.01$ sampled with $2000$ samples.")
dpgmm = mixture.BayesianGaussianMixture(
n_components=10, covariance_type='full', weight_concentration_prior=1e+2,
weight_concentration_prior_type='dirichlet_process',
mean_precision_prior=1e-2, covariance_prior=1e0 * np.eye(2),
init_params="kmeans", max_iter=100, random_state=2).fit(X)
"Bayesian Gaussian mixture models with a Dirichlet process prior "
r"for $\gamma_0=100$")
X_s, y_s = dpgmm.sample(n_samples=2000)

plot_samples(X_s, y_s, dpgmm.n_components, 1,
"Gaussian mixture with a Dirichlet process prior "
r"for $\gamma_0=100$ sampled with $2000$ samples.")
plt.show()

5.18.6 Concentration Prior Type Analysis of Variation Bayesian Gaussian Mixture
This example plots the ellipsoids obtained from a toy dataset (mixture of three Gaussians) fit-
ted by the BayesianGaussianMixture class models with a Dirichlet distribution prior
(weight_concentration_prior_type='dirichlet_distribution') and a Dirichlet process
prior (weight_concentration_prior_type='dirichlet_process'). On each figure, we plot the
results for three different values of the weight concentration prior.
The BayesianGaussianMixture class can adapt its number of mixture components automatically. The param-
eter weight_concentration_prior has a direct link with the resulting number of components with non-zero
weights. Specifying a low value for the concentration prior will make the model put most of the weight on few com-
ponents set the remaining components weights very close to zero. High values of the concentration prior will allow a
larger number of components to be active in the mixture.
The Dirichlet process prior allows to define an infinite number of components and automatically selects the correct
number of components: it activates a component only if it is necessary.
On the contrary the classical finite mixture model with a Dirichlet distribution prior will favor more uniformly weighted
components and therefore tends to divide natural clusters into unnecessary sub-components.

•
# Author: Thierry Guillemot <thierry.guillemot.work@gmail.com>
import numpy as np
import matplotlib.gridspec as gridspec
from sklearn.mixture import BayesianGaussianMixture
print(__doc__)
def plot_ellipses(ax, weights, means, covars):

for n in range(means.shape[0]):
eig_vals, eig_vecs = np.linalg.eigh(covars[n])
unit_eig_vec = eig_vecs[0] / np.linalg.norm(eig_vecs[0])
angle = np.arctan2(unit_eig_vec[1], unit_eig_vec[0])
# Ellipse needs degrees
angle = 180 * angle / np.pi
# eigenvector normalization
eig_vals = 2 * np.sqrt(2) * np.sqrt(eig_vals)
ell = mpl.patches.Ellipse(means[n], eig_vals[0], eig_vals[1],
180 + angle, edgecolor='black')
ell.set_clip_box(ax.bbox)
ell.set_alpha(weights[n])
ell.set_facecolor('#56B4E9')
ax.add_artist(ell)
def plot_results(ax1, ax2, estimator, X, y, title, plot_title=False):

ax1.set_title(title)
ax1.scatter(X[:, 0], X[:, 1], s=5, marker='o', color=colors[y], alpha=0.8)
ax1.set_xlim(-2., 2.)
ax1.set_ylim(-3., 3.)

ax1.set_xticks(())
ax1.set_yticks(())
plot_ellipses(ax1, estimator.weights_, estimator.means_,
estimator.covariances_)
ax2.get_xaxis().set_tick_params(direction='out')
ax2.yaxis.grid(True, alpha=0.7)
for k, w in enumerate(estimator.weights_):
ax2.bar(k, w, width=0.9, color='#56B4E9', zorder=3,
align='center', edgecolor='black')
ax2.text(k, w + 0.007, "%.1f%%" % (w * 100.),
horizontalalignment='center')
ax2.set_xlim(-.6, 2 * n_components - .4)
ax2.set_ylim(0., 1.1)
ax2.tick_params(axis='y', which='both', left='off',
right='off', labelleft='off')
ax2.tick_params(axis='x', which='both', top='off')
if plot_title:
ax1.set_ylabel('Estimated Mixtures')
ax2.set_ylabel('Weight of each component')
# Parameters of the dataset

random_state, n_components, n_features = 2, 3, 2
colors = np.array(['#0072B2', '#F0E442', '#D55E00'])
covars = np.array([[[.7, .0], [.0, .1]],

[[.5, .0], [.0, .1]],
[[.5, .0], [.0, .1]]])
samples = np.array([200, 500, 200])
means = np.array([[.0, -.70],
[.0, .0],
[.0, .70]])
# mean_precision_prior= 0.8 to minimize the influence of the prior

estimators = [
("Finite mixture with a Dirichlet distribution\nprior and "
r"$\gamma_0=$", BayesianGaussianMixture(
weight_concentration_prior_type="dirichlet_distribution",
n_components=2 * n_components, reg_covar=0, init_params='random',
max_iter=1500, mean_precision_prior=.8,
random_state=random_state), [0.001, 1, 1000]),
("Infinite mixture with a Dirichlet process\n prior and" r"$\gamma_0=$",
BayesianGaussianMixture(
weight_concentration_prior_type="dirichlet_process",
n_components=2 * n_components, reg_covar=0, init_params='random',
max_iter=1500, mean_precision_prior=.8,
random_state=random_state), [1, 1000, 100000])]
# Generate data
rng = np.random.RandomState(random_state)
X = np.vstack([
rng.multivariate_normal(means[j], covars[j], samples[j])
for j in range(n_components)])
y = np.concatenate([np.full(samples[j], j, dtype=int)
for j in range(n_components)])
# Plot results in two different figures

for (title, estimator, concentrations_prior) in estimators:

plt.figure(figsize=(4.7 * 3, 8))
gs = gridspec.GridSpec(3, len(concentrations_prior))
for k, concentration in enumerate(concentrations_prior):
estimator.weight_concentration_prior = concentration
estimator.fit(X)
plot_results(plt.subplot(gs[0:2, k]), plt.subplot(gs[2, k]), estimator,
X, y, r"%s$%.1e$" % (title, concentration),
plot_title=k == 0)
plt.show()
5.19 Model Selection
Examples related to the sklearn.model_selection module.
5.19.1 Plotting Cross-Validated Predictions
This example shows how to use cross_val_predict to visualize prediction errors.


from sklearn.model_selection import cross_val_predict
lr = linear_model.LinearRegression()
boston = datasets.load_boston()
y = boston.target
# cross_val_predict returns an array of the same size as `y` where each entry
# is a prediction obtained by cross validation:
predicted = cross_val_predict(lr, boston.data, y, cv=10)
ax.scatter(y, predicted, edgecolors=(0, 0, 0))
ax.plot([y.min(), y.max()], [y.min(), y.max()], 'k--', lw=4)
ax.set_xlabel('Measured')
ax.set_ylabel('Predicted')
plt.show()
5.19. Model Selection 1211

5.19.2 Plotting Validation Curves
In this plot you can see the training scores and validation scores of an SVM for different values of the kernel parameter
gamma. For very low values of gamma, you can see that both the training score and the validation score are low.
This is called underfitting. Medium values of gamma will result in high values for both scores, i.e. the classifier is
performing fairly well. If gamma is too high, the classifier will overfit, which means that the training score is good but
the validation score is poor.
print(__doc__)

import numpy as np

from sklearn.model_selection import validation_curve
param_range = np.logspace(-6, -1, 5)

train_scores, test_scores = validation_curve(
SVC(), X, y, param_name="gamma", param_range=param_range,
cv=5, scoring="accuracy", n_jobs=1)
train_scores_mean = np.mean(train_scores, axis=1)
train_scores_std = np.std(train_scores, axis=1)

test_scores_mean = np.mean(test_scores, axis=1)

test_scores_std = np.std(test_scores, axis=1)
plt.title("Validation Curve with SVM")

plt.xlabel(r"$\gamma$")
plt.ylabel("Score")
plt.ylim(0.0, 1.1)
lw = 2
plt.semilogx(param_range, train_scores_mean, label="Training score",
color="darkorange", lw=lw)
plt.fill_between(param_range, train_scores_mean - train_scores_std,
train_scores_mean + train_scores_std, alpha=0.2,
color="darkorange", lw=lw)
plt.semilogx(param_range, test_scores_mean, label="Cross-validation score",
color="navy", lw=lw)
plt.fill_between(param_range, test_scores_mean - test_scores_std,
test_scores_mean + test_scores_std, alpha=0.2,
color="navy", lw=lw)
plt.show()
5.19.3 Underfitting vs. Overfitting
This example demonstrates the problems of underfitting and overfitting and how we can use linear regression with
polynomial features to approximate nonlinear functions. The plot shows the function that we want to approximate,
which is a part of the cosine function. In addition, the samples from the real function and the approximations of
different models are displayed. The models have polynomial features of different degrees. We can see that a linear
function (polynomial with degree 1) is not sufficient to fit the training samples. This is called underfitting. A
polynomial of degree 4 approximates the true function almost perfectly. However, for higher degrees the model
will overfit the training data, i.e. it learns the noise of the training data. We evaluate quantitatively overfitting /
underfitting by using cross-validation. We calculate the mean squared error (MSE) on the validation set, the higher,
the less likely the model generalizes correctly from the training data.

print(__doc__)
import numpy as np
def true_fun(X):
return np.cos(1.5 * np.pi * X)
np.random.seed(0)
n_samples = 30
degrees = [1, 4, 15]
X = np.sort(np.random.rand(n_samples))
y = true_fun(X) + np.random.randn(n_samples) * 0.1
for i in range(len(degrees)):
ax = plt.subplot(1, len(degrees), i + 1)
plt.setp(ax, xticks=(), yticks=())
polynomial_features = PolynomialFeatures(degree=degrees[i],
include_bias=False)
linear_regression = LinearRegression()
pipeline = Pipeline([("polynomial_features", polynomial_features),
("linear_regression", linear_regression)])
pipeline.fit(X[:, np.newaxis], y)
# Evaluate the models using crossvalidation

scores = cross_val_score(pipeline, X[:, np.newaxis], y,
X_test = np.linspace(0, 1, 100)

plt.plot(X_test, pipeline.predict(X_test[:, np.newaxis]), label="Model")
plt.plot(X_test, true_fun(X_test), label="True function")
plt.scatter(X, y, edgecolor='b', s=20, label="Samples")
plt.xlabel("x")
plt.ylabel("y")
plt.xlim((0, 1))
plt.ylim((-2, 2))
plt.title("Degree {}\nMSE = {:.2e}(+/- {:.2e})".format(
degrees[i], -scores.mean(), scores.std()))
plt.show()

5.19.4 Parameter estimation using grid search with cross-validation
This examples shows how a classifier is optimized by cross-validation, which is done using the sklearn.
model_selection.GridSearchCV object on a development set that comprises only half of the available labeled
data.
The performance of the selected hyper-parameters and trained model is then measured on a dedicated evaluation set
that was not used during the model selection step.
More details on tools available for model selection can be found in the sections on Cross-validation: evaluating
estimator performance and Tuning the hyper-parameters of an estimator.
Out:
# Tuning hyper-parameters for precision
Best parameters set found on development set:
{'C': 10, 'gamma': 0.001, 'kernel': 'rbf'}
Grid scores on development set:
0.986 (+/-0.016) for {'C': 1, 'gamma': 0.001, 'kernel': 'rbf'}

0.975 (+/-0.014) for {'C': 1, 'kernel': 'linear'}
Detailed classification report:
The model is trained on the full development set.

The scores are computed on the full evaluation set.
0 1.00 1.00 1.00 89

1 0.97 1.00 0.98 90
2 0.99 0.98 0.98 92
3 1.00 0.99 0.99 93
4 1.00 1.00 1.00 76
5 0.99 0.98 0.99 108
6 0.99 1.00 0.99 89
7 0.99 1.00 0.99 78
8 1.00 0.98 0.99 92
9 0.99 0.99 0.99 92
micro avg 0.99 0.99 0.99 899

macro avg 0.99 0.99 0.99 899
weighted avg 0.99 0.99 0.99 899
# Tuning hyper-parameters for recall

Best parameters set found on development set:
{'C': 10, 'gamma': 0.001, 'kernel': 'rbf'}
Grid scores on development set:

Detailed classification report:
The model is trained on the full development set.

The scores are computed on the full evaluation set.
0 1.00 1.00 1.00 89

1 0.97 1.00 0.98 90
2 0.99 0.98 0.98 92
3 1.00 0.99 0.99 93
4 1.00 1.00 1.00 76
5 0.99 0.98 0.99 108
6 0.99 1.00 0.99 89
7 0.99 1.00 0.99 78
8 1.00 0.98 0.99 92
9 0.99 0.99 0.99 92
micro avg 0.99 0.99 0.99 899

macro avg 0.99 0.99 0.99 899
weighted avg 0.99 0.99 0.99 899

print(__doc__)
# Loading the Digits dataset

# To apply an classifier on this data, we need to flatten the image, to

n_samples = len(digits.images)
X = digits.images.reshape((n_samples, -1))
y = digits.target
# Split the dataset in two equal parts

# Set the parameters by cross-validation

tuned_parameters = [{'kernel': ['rbf'], 'gamma': [1e-3, 1e-4],
'C': [1, 10, 100, 1000]},
{'kernel': ['linear'], 'C': [1, 10, 100, 1000]}]
scores = ['precision', 'recall']
for score in scores:

print("# Tuning hyper-parameters for %s" % score)
print()
clf = GridSearchCV(SVC(), tuned_parameters, cv=5,

scoring='%s_macro' % score)
print("Best parameters set found on development set:")

print()
print(clf.best_params_)
print()
print("Grid scores on development set:")
print()
means = clf.cv_results_['mean_test_score']
stds = clf.cv_results_['std_test_score']
for mean, std, params in zip(means, stds, clf.cv_results_['params']):
print("%0.3f (+/-%0.03f) for %r"
% (mean, std * 2, params))
print()
print("Detailed classification report:")

print()
print("The model is trained on the full development set.")
print("The scores are computed on the full evaluation set.")
print()
y_true, y_pred = y_test, clf.predict(X_test)
print(classification_report(y_true, y_pred))
print()
# Note the problem is too easy: the hyperparameter plateau is too flat and the
# output model is the same for precision and recall with ties in quality.

5.19.5 Train error vs Test error
Illustration of how the performance of an estimator on unseen data (test data) is not the same as the performance on
training data. As the regularization increases the performance on train decreases while the performance on test is
optimal within a range of values of the regularization parameter. The example with an Elastic-Net regression model
and the performance is measured using the explained variance a.k.a. R^2.
Out:
Optimal regularization parameter : 0.000335292414924956
print(__doc__)

import numpy as np
# #############################################################################

n_samples_train, n_samples_test, n_features = 75, 150, 500

np.random.seed(0)
coef = np.random.randn(n_features)
coef[50:] = 0.0 # only the top 10 features are impacting the model
X = np.random.randn(n_samples_train + n_samples_test, n_features)
y = np.dot(X, coef)
# Split train and test data

X_train, X_test = X[:n_samples_train], X[n_samples_train:]
y_train, y_test = y[:n_samples_train], y[n_samples_train:]
# #############################################################################
# Compute train and test errors
enet = linear_model.ElasticNet(l1_ratio=0.7)
train_errors = list()
test_errors = list()
for alpha in alphas:
enet.set_params(alpha=alpha)
enet.fit(X_train, y_train)
train_errors.append(enet.score(X_train, y_train))
test_errors.append(enet.score(X_test, y_test))
i_alpha_optim = np.argmax(test_errors)
alpha_optim = alphas[i_alpha_optim]
print("Optimal regularization parameter : %s" % alpha_optim)
# Estimate the coef_ on full data with optimal regularization parameter

enet.set_params(alpha=alpha_optim)
coef_ = enet.fit(X, y).coef_
# #############################################################################
# Plot results functions

plt.semilogx(alphas, train_errors, label='Train')
plt.semilogx(alphas, test_errors, label='Test')
plt.vlines(alpha_optim, plt.ylim()[0], np.max(test_errors), color='k',
linewidth=3, label='Optimum on test')
plt.legend(loc='lower left')
plt.ylim([0, 1.2])
plt.xlabel('Regularization parameter')
plt.ylabel('Performance')
# Show estimated coef_ vs true coef

plt.plot(coef, label='True coef')
plt.plot(coef_, label='Estimated coef')
plt.legend()
plt.show()

5.19.6 Comparing randomized search and grid search for hyperparameter estima-
tion
Compare randomized search and grid search for optimizing hyperparameters of a random forest. All parameters that
influence the learning are searched simultaneously (except for the number of estimators, which poses a time / quality
tradeoff).
The randomized search and the grid search explore exactly the same space of parameters. The result in parameter
settings is quite similar, while the run time for randomized search is drastically lower.
The performance is slightly worse for the randomized search, though this is most likely a noise effect and would not
carry over to a held-out test set.
Note that in practice, one would not search over this many different parameters simultaneously using grid search, but
pick only the ones deemed most important.
Out:
RandomizedSearchCV took 5.68 seconds for 20 candidates parameter settings.
Model with rank: 1
Mean validation score: 0.939 (std: 0.024)
Parameters: {'bootstrap': False, 'criterion': 'entropy', 'max_depth': None, 'max_
˓→features': 7, 'min_samples_split': 3}
Model with rank: 2

Parameters: {'bootstrap': False, 'criterion': 'gini', 'max_depth': None, 'max_features
˓→': 6, 'min_samples_split': 6}
Model with rank: 3

Parameters: {'bootstrap': True, 'criterion': 'gini', 'max_depth': None, 'max_features
GridSearchCV took 17.43 seconds for 72 candidate parameter settings.

Model with rank: 1
Model with rank: 2

Parameters: {'bootstrap': False, 'criterion': 'gini', 'max_depth': None, 'max_features
Model with rank: 3

print(__doc__)
import numpy as np


from scipy.stats import randint as sp_randint

from sklearn.model_selection import RandomizedSearchCV
# get some data

# build a classifier
# Utility function to report best scores

def report(results, n_top=3):
for i in range(1, n_top + 1):
candidates = np.flatnonzero(results['rank_test_score'] == i)
for candidate in candidates:
print("Model with rank: {0}".format(i))
print("Mean validation score: {0:.3f} (std: {1:.3f})".format(
results['mean_test_score'][candidate],
results['std_test_score'][candidate]))
print("Parameters: {0}".format(results['params'][candidate]))
print("")
# specify parameters and distributions to sample from

param_dist = {"max_depth": [3, None],
"max_features": sp_randint(1, 11),
"min_samples_split": sp_randint(2, 11),
"bootstrap": [True, False],
"criterion": ["gini", "entropy"]}
# run randomized search

n_iter_search = 20
random_search = RandomizedSearchCV(clf, param_distributions=param_dist,
n_iter=n_iter_search, cv=5)
start = time()
random_search.fit(X, y)
print("RandomizedSearchCV took %.2f seconds for %d candidates"
" parameter settings." % ((time() - start), n_iter_search))
report(random_search.cv_results_)
# use a full grid over all parameters

param_grid = {"max_depth": [3, None],
"max_features": [1, 3, 10],
"min_samples_split": [2, 3, 10],
"bootstrap": [True, False],
"criterion": ["gini", "entropy"]}
# run grid search

grid_search = GridSearchCV(clf, param_grid=param_grid, cv=5)
start = time()
grid_search.fit(X, y)

print("GridSearchCV took %.2f seconds for %d candidate parameter settings."

% (time() - start, len(grid_search.cv_results_['params'])))
report(grid_search.cv_results_)
5.19.7 Receiver Operating Characteristic (ROC) with cross validation
Example of Receiver Operating Characteristic (ROC) metric to evaluate classifier output quality using cross-validation.
ROC curves typically feature true positive rate on the Y axis, and false positive rate on the X axis. This means that the
top left corner of the plot is the “ideal” point - a false positive rate of zero, and a true positive rate of one. This is not
very realistic, but it does mean that a larger area under the curve (AUC) is usually better.
The “steepness” of ROC curves is also important, since it is ideal to maximize the true positive rate while minimizing
the false positive rate.
This example shows the ROC response of different datasets, created from K-fold cross-validation. Taking all of these
curves, it is possible to calculate the mean area under curve, and see the variance of the curve when the training set
is split into different subsets. This roughly shows how the classifier output is affected by changes in the training data,
and how different the splits generated by K-fold cross-validation are from one another.
Note:
See also sklearn.metrics.roc_auc_score, sklearn.model_selection.cross_val_score,
Receiver Operating Characteristic (ROC),

print(__doc__)
import numpy as np
from scipy import interp
from sklearn import svm, datasets

from sklearn.metrics import roc_curve, auc
# #############################################################################
# Data IO and generation

X = iris.data
y = iris.target
X, y = X[y != 2], y[y != 2]
# Add noisy features

X = np.c_[X, random_state.randn(n_samples, 200 * n_features)]
# #############################################################################
# Classification and ROC analysis

# Run classifier with cross-validation and plot ROC curves

cv = StratifiedKFold(n_splits=6)
classifier = svm.SVC(kernel='linear', probability=True,
tprs = []
aucs = []
mean_fpr = np.linspace(0, 1, 100)
i = 0
for train, test in cv.split(X, y):
probas_ = classifier.fit(X[train], y[train]).predict_proba(X[test])
# Compute ROC curve and area the curve
fpr, tpr, thresholds = roc_curve(y[test], probas_[:, 1])
tprs.append(interp(mean_fpr, fpr, tpr))
tprs[-1][0] = 0.0
roc_auc = auc(fpr, tpr)
aucs.append(roc_auc)
plt.plot(fpr, tpr, lw=1, alpha=0.3,
label='ROC fold %d (AUC = %0.2f)' % (i, roc_auc))
i += 1
plt.plot([0, 1], [0, 1], linestyle='--', lw=2, color='r',
label='Chance', alpha=.8)
mean_tpr = np.mean(tprs, axis=0)

mean_tpr[-1] = 1.0
mean_auc = auc(mean_fpr, mean_tpr)
std_auc = np.std(aucs)
plt.plot(mean_fpr, mean_tpr, color='b',
label=r'Mean ROC (AUC = %0.2f $\pm$ %0.2f)' % (mean_auc, std_auc),
lw=2, alpha=.8)
std_tpr = np.std(tprs, axis=0)

tprs_upper = np.minimum(mean_tpr + std_tpr, 1)
tprs_lower = np.maximum(mean_tpr - std_tpr, 0)
plt.fill_between(mean_fpr, tprs_lower, tprs_upper, color='grey', alpha=.2,
label=r'$\pm$ 1 std. dev.')
plt.xlim([-0.05, 1.05])
plt.ylim([-0.05, 1.05])
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('Receiver operating characteristic example')
plt.show()

5.19.8 Nested versus non-nested cross-validation
This example compares non-nested and nested cross-validation strategies on a classifier of the iris data set. Nested
cross-validation (CV) is often used to train a model in which hyperparameters also need to be optimized. Nested CV
estimates the generalization error of the underlying model and its (hyper)parameter search. Choosing the parameters
that maximize non-nested CV biases the model to the dataset, yielding an overly-optimistic score.
Model selection without nested CV uses the same data to tune model parameters and evaluate model performance.
Information may thus “leak” into the model and overfit the data. The magnitude of this effect is primarily dependent
on the size of the dataset and the stability of the model. See Cawley and Talbot1 for an analysis of these issues.
To avoid this problem, nested CV effectively uses a series of train/validation/test set splits. In the inner loop
(here executed by GridSearchCV ), the score is approximately maximized by fitting a model to each training
set, and then directly maximized in selecting (hyper)parameters over the validation set. In the outer loop (here in
cross_val_score), generalization error is estimated by averaging test set scores over several dataset splits.
The example below uses a support vector classifier with a non-linear kernel to build a model with optimized hyperpa-
rameters by grid search. We compare the performance of non-nested and nested CV strategies by taking the difference
between their scores.
See Also:
• Cross-validation: evaluating estimator performance

• Tuning the hyper-parameters of an estimator
References:
1 Cawley, G.C.; Talbot, N.L.C. On over-fitting in model selection and subsequent selection bias in performance evaluation. J. Mach. Learn. Res
2010,11, 2079-2107.

Out:
Average difference of 0.007742 with std. dev. of 0.007688.

from sklearn.model_selection import GridSearchCV, cross_val_score, KFold
import numpy as np
print(__doc__)
# Number of random trials

NUM_TRIALS = 30
# Load the dataset

iris = load_iris()
X_iris = iris.data
y_iris = iris.target
# Set up possible values of parameters to optimize over

p_grid = {"C": [1, 10, 100],

"gamma": [.01, .1]}
# We will use a Support Vector Classifier with "rbf" kernel

svm = SVC(kernel="rbf")
# Arrays to store scores

non_nested_scores = np.zeros(NUM_TRIALS)
nested_scores = np.zeros(NUM_TRIALS)
# Loop for each trial

for i in range(NUM_TRIALS):
# Choose cross-validation techniques for the inner and outer loops,

# independently of the dataset.
# E.g "GroupKFold", "LeaveOneOut", "LeaveOneGroupOut", etc.
inner_cv = KFold(n_splits=4, shuffle=True, random_state=i)
outer_cv = KFold(n_splits=4, shuffle=True, random_state=i)
# Non_nested parameter search and scoring

clf = GridSearchCV(estimator=svm, param_grid=p_grid, cv=inner_cv)
clf.fit(X_iris, y_iris)
non_nested_scores[i] = clf.best_score_
# Nested CV with parameter optimization

nested_score = cross_val_score(clf, X=X_iris, y=y_iris, cv=outer_cv)
nested_scores[i] = nested_score.mean()
score_difference = non_nested_scores - nested_scores
print("Average difference of {0:6f} with std. dev. of {1:6f}."

.format(score_difference.mean(), score_difference.std()))
# Plot scores on each trial for nested and non-nested CV

plt.figure()
plt.subplot(211)
non_nested_scores_line, = plt.plot(non_nested_scores, color='r')
nested_line, = plt.plot(nested_scores, color='b')
plt.ylabel("score", fontsize="14")
plt.legend([non_nested_scores_line, nested_line],
["Non-Nested CV", "Nested CV"],
bbox_to_anchor=(0, .4, .5, 0))
plt.title("Non-Nested and Nested Cross Validation on Iris Dataset",
x=.5, y=1.1, fontsize="15")
# Plot bar chart of the difference.

plt.subplot(212)
difference_plot = plt.bar(range(NUM_TRIALS), score_difference)
plt.xlabel("Individual Trial #")
plt.legend([difference_plot],
["Non-Nested CV - Nested CV Score"],
bbox_to_anchor=(0, 1, .8, 0))
plt.ylabel("score difference", fontsize="14")
plt.show()

5.19.9 Demonstration of multi-metric evaluation on cross_val_score and Grid-

SearchCV
Multiple metric parameter search can be done by setting the scoring parameter to a list of metric scorer names or a
dict mapping the scorer names to the scorer callables.
The scores of all the scorers are available in the cv_results_ dict at keys ending in '_<scorer_name>'
('mean_test_precision', 'rank_test_precision', etc. . . )
The best_estimator_, best_index_, best_score_ and best_params_ correspond to the scorer (key)
that is set to the refit attribute.
# Author: Raghav RV <rvraghav93@gmail.com>
# License: BSD
import numpy as np
from sklearn.datasets import make_hastie_10_2

from sklearn.metrics import make_scorer
print(__doc__)
Running GridSearchCV using multiple evaluation metrics
X, y = make_hastie_10_2(n_samples=8000, random_state=42)
# The scorers can be either be one of the predefined metric strings or a scorer
# callable, like the one returned by make_scorer
scoring = {'AUC': 'roc_auc', 'Accuracy': make_scorer(accuracy_score)}
# Setting refit='AUC', refits an estimator on the whole dataset with the

# parameter setting that has the best cross-validated AUC score.
# That estimator is made available at ``gs.best_estimator_`` along with
# parameters like ``gs.best_score_``, ``gs.best_params_`` and
# ``gs.best_index_``
gs = GridSearchCV(DecisionTreeClassifier(random_state=42),
param_grid={'min_samples_split': range(2, 403, 10)},
scoring=scoring, cv=5, refit='AUC', return_train_score=True)
gs.fit(X, y)
results = gs.cv_results_
Plotting the result
plt.title("GridSearchCV evaluating using multiple scorers simultaneously",
fontsize=16)

plt.xlabel("min_samples_split")
plt.ylabel("Score")
ax = plt.gca()
ax.set_xlim(0, 402)
ax.set_ylim(0.73, 1)
# Get the regular numpy array from the MaskedArray

X_axis = np.array(results['param_min_samples_split'].data, dtype=float)
for scorer, color in zip(sorted(scoring), ['g', 'k']):

for sample, style in (('train', '--'), ('test', '-')):
sample_score_mean = results['mean_%s_%s' % (sample, scorer)]
sample_score_std = results['std_%s_%s' % (sample, scorer)]
ax.fill_between(X_axis, sample_score_mean - sample_score_std,
sample_score_mean + sample_score_std,
alpha=0.1 if sample == 'test' else 0, color=color)
ax.plot(X_axis, sample_score_mean, style, color=color,
alpha=1 if sample == 'test' else 0.7,
label="%s (%s)" % (scorer, sample))
best_index = np.nonzero(results['rank_test_%s' % scorer] == 1)[0][0]

best_score = results['mean_test_%s' % scorer][best_index]
# Plot a dotted vertical line at the best score for that scorer marked by x
ax.plot([X_axis[best_index], ] * 2, [0, best_score],
linestyle='-.', color=color, marker='x', markeredgewidth=3, ms=8)
# Annotate the best score for that scorer

ax.annotate("%0.2f" % best_score,
(X_axis[best_index], best_score + 0.005))
plt.grid('off')
plt.show()

5.19.10 Sample pipeline for text feature extraction and evaluation
The dataset used in this example is the 20 newsgroups dataset which will be automatically downloaded and then cached
and reused for the document classification example.
You can adjust the number of categories by giving their names to the dataset loader or setting them to None to get the
20 of them.
Here is a sample output of a run on a quad-core machine:

Loading 20 newsgroups dataset for categories:

['alt.atheism', 'talk.religion.misc']
1427 documents
2 categories
Performing grid search...

pipeline: ['vect', 'tfidf', 'clf']
parameters:
{'clf__alpha': (1.0000000000000001e-05, 9.9999999999999995e-07),
'clf__max_iter': (10, 50, 80),
'clf__penalty': ('l2', 'elasticnet'),
'tfidf__use_idf': (True, False),
'vect__max_n': (1, 2),
'vect__max_df': (0.5, 0.75, 1.0),
'vect__max_features': (None, 5000, 10000, 50000)}
done in 1737.030s
Best score: 0.940

Best parameters set:
clf__alpha: 9.9999999999999995e-07
clf__max_iter: 50
clf__penalty: 'elasticnet'
tfidf__use_idf: True
vect__max_n: 2
vect__max_df: 0.75
vect__max_features: 50000

# Peter Prettenhofer <peter.prettenhofer@gmail.com>
from pprint import pprint

import logging

from sklearn.feature_extraction.text import CountVectorizer
from sklearn.feature_extraction.text import TfidfTransformer
print(__doc__)

# #############################################################################
# Load some categories from the training set
categories = [
'alt.atheism',
'talk.religion.misc',
]

# Uncomment the following to do the analysis on all the categories

#categories = None
print("Loading 20 newsgroups dataset for categories:")

print(categories)
data = fetch_20newsgroups(subset='train', categories=categories)

print("%d documents" % len(data.filenames))
print("%d categories" % len(data.target_names))
print()
# #############################################################################
# Define a pipeline combining a text feature extractor with a simple
# classifier
pipeline = Pipeline([
('vect', CountVectorizer()),
('tfidf', TfidfTransformer()),
('clf', SGDClassifier()),
])
# uncommenting more parameters will give better exploring power but will
# increase processing time in a combinatorial way
parameters = {
'vect__max_df': (0.5, 0.75, 1.0),
# 'vect__max_features': (None, 5000, 10000, 50000),
'vect__ngram_range': ((1, 1), (1, 2)), # unigrams or bigrams
# 'tfidf__use_idf': (True, False),
# 'tfidf__norm': ('l1', 'l2'),
'clf__max_iter': (5,),
'clf__alpha': (0.00001, 0.000001),
'clf__penalty': ('l2', 'elasticnet'),
# 'clf__max_iter': (10, 50, 80),
}
if __name__ == "__main__":
# multiprocessing requires the fork to happen in a __main__ protected
# block
# find the best parameters for both the feature extraction and the
# classifier
grid_search = GridSearchCV(pipeline, parameters, cv=5,
n_jobs=-1, verbose=1)
print("Performing grid search...")

print("pipeline:", [name for name, _ in pipeline.steps])
print("parameters:")
pprint(parameters)
t0 = time()
grid_search.fit(data.data, data.target)
print()
print("Best score: %0.3f" % grid_search.best_score_)

print("Best parameters set:")
best_parameters = grid_search.best_estimator_.get_params()
for param_name in sorted(parameters.keys()):
print("\t%s: %r" % (param_name, best_parameters[param_name]))

5.19.11 Confusion matrix
Example of confusion matrix usage to evaluate the quality of the output of a classifier on the iris data set. The diagonal
elements represent the number of points for which the predicted label is equal to the true label, while off-diagonal
elements are those that are mislabeled by the classifier. The higher the diagonal values of the confusion matrix the
better, indicating many correct predictions.
The figures show the confusion matrix with and without normalization by class support size (number of elements in
each class). This kind of normalization can be interesting in case of class imbalance to have a more visual interpretation
of which class is being misclassified.
Here the results are not as good as they could be as our choice for the regularization parameter C was not the best. In
real life applications this parameter is usually chosen using Tuning the hyper-parameters of an estimator.

•
Out:
Confusion matrix, without normalization

[[13 0 0]
[ 0 10 6]
[ 0 0 9]]
Normalized confusion matrix
[[1. 0. 0. ]
[0. 0.62 0.38]
[0. 0. 1. ]]
print(__doc__)
import numpy as np

from sklearn.utils.multiclass import unique_labels

X = iris.data
y = iris.target
class_names = iris.target_names
# Split the data into a training set and a test set

# Run classifier, using a model that is too regularized (C too low) to see
# the impact on the results
classifier = svm.SVC(kernel='linear', C=0.01)
y_pred = classifier.fit(X_train, y_train).predict(X_test)
def plot_confusion_matrix(y_true, y_pred, classes,

normalize=False,
title=None,
cmap=plt.cm.Blues):
"""
This function prints and plots the confusion matrix.
Normalization can be applied by setting `normalize=True`.
"""
if not title:
if normalize:
title = 'Normalized confusion matrix'
else:
title = 'Confusion matrix, without normalization'
# Compute confusion matrix

cm = confusion_matrix(y_true, y_pred)
# Only use the labels that appear in the data
classes = classes[unique_labels(y_true, y_pred)]
if normalize:
cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]
print("Normalized confusion matrix")
else:
print('Confusion matrix, without normalization')
print(cm)
im = ax.imshow(cm, interpolation='nearest', cmap=cmap)
ax.figure.colorbar(im, ax=ax)
# We want to show all ticks...
ax.set(xticks=np.arange(cm.shape[1]),
yticks=np.arange(cm.shape[0]),
# ... and label them with the respective list entries
xticklabels=classes, yticklabels=classes,
title=title,
ylabel='True label',
xlabel='Predicted label')
# Rotate the tick labels and set their alignment.

plt.setp(ax.get_xticklabels(), rotation=45, ha="right",
rotation_mode="anchor")
# Loop over data dimensions and create text annotations.

fmt = '.2f' if normalize else 'd'

thresh = cm.max() / 2.
for i in range(cm.shape[0]):
for j in range(cm.shape[1]):
ax.text(j, i, format(cm[i, j], fmt),
ha="center", va="center",
color="white" if cm[i, j] > thresh else "black")
fig.tight_layout()
return ax
np.set_printoptions(precision=2)
# Plot non-normalized confusion matrix

plot_confusion_matrix(y_test, y_pred, classes=class_names,
title='Confusion matrix, without normalization')
# Plot normalized confusion matrix

plot_confusion_matrix(y_test, y_pred, classes=class_names, normalize=True,
title='Normalized confusion matrix')
plt.show()
5.19.12 Visualizing cross-validation behavior in scikit-learn
Choosing the right cross-validation object is a crucial part of fitting a model properly. There are many ways to split
data into training and test sets in order to avoid model overfitting, to standardize the number of groups in test sets, etc.
This example visualizes the behavior of several common scikit-learn objects for comparison.
from sklearn.model_selection import (TimeSeriesSplit, KFold, ShuffleSplit,

StratifiedKFold, GroupShuffleSplit,
GroupKFold, StratifiedShuffleSplit)
import numpy as np
from matplotlib.patches import Patch
np.random.seed(1338)
cmap_data = plt.cm.Paired
cmap_cv = plt.cm.coolwarm
n_splits = 4
Visualize our data
First, we must understand the structure of our data. It has 100 randomly generated input datapoints, 3 classes split
unevenly across datapoints, and 10 “groups” split evenly across datapoints.
As we’ll see, some cross-validation objects do specific things with labeled data, others behave differently with grouped
data, and others do not use this information.
To begin, we’ll visualize our data.

# Generate the class/group data

n_points = 100
X = np.random.randn(100, 10)
percentiles_classes = [.1, .3, .6]

y = np.hstack([[ii] * int(100 * perc)
for ii, perc in enumerate(percentiles_classes)])
# Evenly spaced groups repeated once

groups = np.hstack([[ii] * 10 for ii in range(10)])
def visualize_groups(classes, groups, name):

# Visualize dataset groups
ax.scatter(range(len(groups)), [.5] * len(groups), c=groups, marker='_',
lw=50, cmap=cmap_data)
ax.scatter(range(len(groups)), [3.5] * len(groups), c=classes, marker='_',
lw=50, cmap=cmap_data)
ax.set(ylim=[-1, 5], yticks=[.5, 3.5],
yticklabels=['Data\ngroup', 'Data\nclass'], xlabel="Sample index")
visualize_groups(y, groups, 'no groups')

Define a function to visualize cross-validation behavior
We’ll define a function that lets us visualize the behavior of each cross-validation object. We’ll perform 4 splits of the
data. On each split, we’ll visualize the indices chosen for the training set (in blue) and the test set (in red).
def plot_cv_indices(cv, X, y, group, ax, n_splits, lw=10):

"""Create a sample plot for indices of a cross-validation object."""
# Generate the training/testing visualizations for each CV split

for ii, (tr, tt) in enumerate(cv.split(X=X, y=y, groups=group)):
# Fill in indices with the training/test groups
indices = np.array([np.nan] * len(X))
indices[tt] = 1
indices[tr] = 0
# Visualize the results

ax.scatter(range(len(indices)), [ii + .5] * len(indices),
c=indices, marker='_', lw=lw, cmap=cmap_cv,
vmin=-.2, vmax=1.2)
# Plot the data classes and groups at the end

ax.scatter(range(len(X)), [ii + 1.5] * len(X),
c=y, marker='_', lw=lw, cmap=cmap_data)
ax.scatter(range(len(X)), [ii + 2.5] * len(X),

c=group, marker='_', lw=lw, cmap=cmap_data)
# Formatting
yticklabels = list(range(n_splits)) + ['class', 'group']
ax.set(yticks=np.arange(n_splits+2) + .5, yticklabels=yticklabels,
xlabel='Sample index', ylabel="CV iteration",
ylim=[n_splits+2.2, -.2], xlim=[0, 100])
ax.set_title('{}'.format(type(cv).__name__), fontsize=15)
return ax
Let’s see how it looks for the KFold cross-validation object:
cv = KFold(n_splits)
plot_cv_indices(cv, X, y, groups, ax, n_splits)

As you can see, by default the KFold cross-validation iterator does not take either datapoint class or group into
consideration. We can change this by using the StratifiedKFold like so.
cv = StratifiedKFold(n_splits)
plot_cv_indices(cv, X, y, groups, ax, n_splits)

In this case, the cross-validation retained the same ratio of classes across each CV split. Next we’ll visualize this
behavior for a number of CV iterators.
Visualize cross-validation indices for many CV objects
Let’s visually compare the cross validation behavior for many scikit-learn cross-validation objects. Below we will
loop through several common cross-validation objects, visualizing the behavior of each.
Note how some use the group/class information while others do not.
cvs = [KFold, GroupKFold, ShuffleSplit, StratifiedKFold,

GroupShuffleSplit, StratifiedShuffleSplit, TimeSeriesSplit]
for cv in cvs:
this_cv = cv(n_splits=n_splits)
plot_cv_indices(this_cv, X, y, groups, ax, n_splits)
ax.legend([Patch(color=cmap_cv(.8)), Patch(color=cmap_cv(.02))],
['Testing set', 'Training set'], loc=(1.02, .8))
# Make the legend fit
plt.tight_layout()
fig.subplots_adjust(right=.7)
plt.show()




•
5.19.13 Receiver Operating Characteristic (ROC)
Example of Receiver Operating Characteristic (ROC) metric to evaluate classifier output quality.
ROC curves typically feature true positive rate on the Y axis, and false positive rate on the X axis. This means that the
top left corner of the plot is the “ideal” point - a false positive rate of zero, and a true positive rate of one. This is not
very realistic, but it does mean that a larger area under the curve (AUC) is usually better.
The “steepness” of ROC curves is also important, since it is ideal to maximize the true positive rate while minimizing
the false positive rate.
Multiclass settings
ROC curves are typically used in binary classification to study the output of a classifier. In order to extend ROC curve
and ROC area to multi-class or multi-label classification, it is necessary to binarize the output. One ROC curve can
be drawn per label, but one can also draw a ROC curve by considering each element of the label indicator matrix as a
binary prediction (micro-averaging).
Another evaluation measure for multi-class classification is macro-averaging, which gives equal weight to the classi-
fication of each label.
Note:
See also sklearn.metrics.roc_auc_score, Receiver Operating Characteristic (ROC) with cross valida-
tion.
print(__doc__)
import numpy as np



from sklearn.metrics import roc_curve, auc
from sklearn.preprocessing import label_binarize
from scipy import interp

X = iris.data
y = iris.target
# Binarize the output

y = label_binarize(y, classes=[0, 1, 2])
n_classes = y.shape[1]
# Add noisy features to make the problem harder

# shuffle and split training and test sets

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=.5,
random_state=0)
# Learn to predict each class against the other

classifier = OneVsRestClassifier(svm.SVC(kernel='linear', probability=True,
random_state=random_state))
y_score = classifier.fit(X_train, y_train).decision_function(X_test)
# Compute ROC curve and ROC area for each class

fpr = dict()
tpr = dict()
roc_auc = dict()
for i in range(n_classes):
fpr[i], tpr[i], _ = roc_curve(y_test[:, i], y_score[:, i])
roc_auc[i] = auc(fpr[i], tpr[i])
# Compute micro-average ROC curve and ROC area

fpr["micro"], tpr["micro"], _ = roc_curve(y_test.ravel(), y_score.ravel())
roc_auc["micro"] = auc(fpr["micro"], tpr["micro"])
Plot of a ROC curve for a specific class
plt.figure()
lw = 2
plt.plot(fpr[2], tpr[2], color='darkorange',
lw=lw, label='ROC curve (area = %0.2f)' % roc_auc[2])
plt.plot([0, 1], [0, 1], color='navy', lw=lw, linestyle='--')
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.title('Receiver operating characteristic example')

plt.show()
Plot ROC curves for the multiclass problem

# Compute macro-average ROC curve and ROC area
# First aggregate all false positive rates

all_fpr = np.unique(np.concatenate([fpr[i] for i in range(n_classes)]))
# Then interpolate all ROC curves at this points

mean_tpr = np.zeros_like(all_fpr)
mean_tpr += interp(all_fpr, fpr[i], tpr[i])
# Finally average it and compute AUC

mean_tpr /= n_classes
fpr["macro"] = all_fpr
tpr["macro"] = mean_tpr
roc_auc["macro"] = auc(fpr["macro"], tpr["macro"])
# Plot all ROC curves

plt.figure()
plt.plot(fpr["micro"], tpr["micro"],
label='micro-average ROC curve (area = {0:0.2f})'
''.format(roc_auc["micro"]),

color='deeppink', linestyle=':', linewidth=4)
plt.plot(fpr["macro"], tpr["macro"],
label='macro-average ROC curve (area = {0:0.2f})'
''.format(roc_auc["macro"]),
color='navy', linestyle=':', linewidth=4)
colors = cycle(['aqua', 'darkorange', 'cornflowerblue'])

for i, color in zip(range(n_classes), colors):
plt.plot(fpr[i], tpr[i], color=color, lw=lw,
label='ROC curve of class {0} (area = {1:0.2f})'
''.format(i, roc_auc[i]))
plt.plot([0, 1], [0, 1], 'k--', lw=lw)

plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.title('Some extension of Receiver operating characteristic to multi-class')
plt.show()

5.19.14 Plotting Learning Curves
On the left side the learning curve of a naive Bayes classifier is shown for the digits dataset. Note that the training
score and the cross-validation score are both not very good at the end. However, the shape of the curve can be found
in more complex datasets very often: the training score is very high at the beginning and decreases and the cross-
validation score is very low at the beginning and increases. On the right side we see the learning curve of an SVM
with RBF kernel. We can see clearly that the training score is still around the maximum and the validation score could
be increased with more training samples.

print(__doc__)
import numpy as np
from sklearn.model_selection import learning_curve
from sklearn.model_selection import ShuffleSplit
def plot_learning_curve(estimator, title, X, y, ylim=None, cv=None,

n_jobs=None, train_sizes=np.linspace(.1, 1.0, 5)):
"""
Generate a simple plot of the test and training learning curve.
Parameters
----------
estimator : object type that implements the "fit" and "predict" methods
An object of that type which is cloned for each validation.
title : string
Title for the chart.
X : array-like, shape (n_samples, n_features)

Training vector, where n_samples is the number of samples and
n_features is the number of features.

y : array-like, shape (n_samples) or (n_samples, n_features), optional

Target relative to X for classification or regression;
None for unsupervised learning.
ylim : tuple, shape (ymin, ymax), optional

Defines minimum and maximum yvalues plotted.
cv : int, cross-validation generator or an iterable, optional

Determines the cross-validation splitting strategy.
Possible inputs for cv are:
- None, to use the default 3-fold cross-validation,
- integer, to specify the number of folds.
- :term:`CV splitter`,
- An iterable yielding (train, test) splits as arrays of indices.
For integer/None inputs, if ``y`` is binary or multiclass,

:class:`StratifiedKFold` used. If the estimator is not a classifier
or if ``y`` is neither binary nor multiclass, :class:`KFold` is used.
Refer :ref:Ùser Guide <cross_validation>` for the various

cross-validators that can be used here.
n_jobs : int or None, optional (default=None)

Number of jobs to run in parallel.
``None`` means 1 unless in a :obj:`joblib.parallel_backend` context.
``-1`` means using all processors. See :term:`Glossary <n_jobs>`
for more details.
train_sizes : array-like, shape (n_ticks,), dtype float or int

Relative or absolute numbers of training examples that will be used to
generate the learning curve. If the dtype is float, it is regarded as a
fraction of the maximum size of the training set (that is determined
by the selected validation method), i.e. it has to be within (0, 1].
Otherwise it is interpreted as absolute sizes of the training sets.
Note that for classification the number of samples usually have to
be big enough to contain at least one sample from each class.
(default: np.linspace(0.1, 1.0, 5))
"""
plt.figure()
plt.title(title)
if ylim is not None:
plt.ylim(*ylim)
plt.xlabel("Training examples")
plt.ylabel("Score")
train_sizes, train_scores, test_scores = learning_curve(
estimator, X, y, cv=cv, n_jobs=n_jobs, train_sizes=train_sizes)
train_scores_mean = np.mean(train_scores, axis=1)
train_scores_std = np.std(train_scores, axis=1)
test_scores_mean = np.mean(test_scores, axis=1)
test_scores_std = np.std(test_scores, axis=1)
plt.grid()
plt.fill_between(train_sizes, train_scores_mean - train_scores_std,

train_scores_mean + train_scores_std, alpha=0.1,
color="r")
plt.fill_between(train_sizes, test_scores_mean - test_scores_std,
test_scores_mean + test_scores_std, alpha=0.1, color="g")

plt.plot(train_sizes, train_scores_mean, 'o-', color="r",

label="Training score")
plt.plot(train_sizes, test_scores_mean, 'o-', color="g",
label="Cross-validation score")
return plt
title = "Learning Curves (Naive Bayes)"

# Cross validation with 100 iterations to get smoother mean test and train
# score curves, each time with 20% data randomly selected as a validation set.
cv = ShuffleSplit(n_splits=100, test_size=0.2, random_state=0)
estimator = GaussianNB()
plot_learning_curve(estimator, title, X, y, ylim=(0.7, 1.01), cv=cv, n_jobs=4)
title = r"Learning Curves (SVM, RBF kernel, $\gamma=0.001$)"

# SVC is more expensive so we do a lower number of CV iterations:
cv = ShuffleSplit(n_splits=10, test_size=0.2, random_state=0)
estimator = SVC(gamma=0.001)
plot_learning_curve(estimator, title, X, y, (0.7, 1.01), cv=cv, n_jobs=4)
plt.show()
5.19.15 Precision-Recall
Example of Precision-Recall metric to evaluate classifier output quality.

Precision-Recall is a useful measure of success of prediction when the classes are very imbalanced. In information
retrieval, precision is a measure of result relevancy, while recall is a measure of how many truly relevant results are
returned.
The precision-recall curve shows the tradeoff between precision and recall for different threshold. A high area under
the curve represents both high recall and high precision, where high precision relates to a low false positive rate, and
high recall relates to a low false negative rate. High scores for both show that the classifier is returning accurate results
(high precision), as well as returning a majority of all positive results (high recall).
A system with high recall but low precision returns many results, but most of its predicted labels are incorrect when
compared to the training labels. A system with high precision but low recall is just the opposite, returning very few
results, but most of its predicted labels are correct when compared to the training labels. An ideal system with high
precision and high recall will return many results, with all results labeled correctly.
Precision (𝑃 ) is defined as the number of true positives (𝑇𝑝 ) over the number of true positives plus the number of false
positives (𝐹𝑝 ).
𝑇𝑝
𝑃 = 𝑇𝑝 +𝐹𝑝

Recall (𝑅) is defined as the number of true positives (𝑇𝑝 ) over the number of true positives plus the number of false
negatives (𝐹𝑛 ).
𝑇𝑝
𝑅= 𝑇𝑝 +𝐹𝑛
These quantities are also related to the (𝐹1 ) score, which is defined as the harmonic mean of precision and recall.
×𝑅
𝐹1 = 2𝑃
𝑃 +𝑅
𝑇
𝑝
Note that the precision may not decrease with recall. The definition of precision ( 𝑇𝑝 +𝐹 𝑝
) shows that lowering the
threshold of a classifier may increase the denominator, by increasing the number of results returned. If the threshold
was previously set too high, the new results may all be true positives, which will increase precision. If the previous
threshold was about right or too low, further lowering the threshold will introduce false positives, decreasing precision.
𝑇
𝑝
Recall is defined as 𝑇𝑝 +𝐹 𝑛
, where 𝑇𝑝 + 𝐹𝑛 does not depend on the classifier threshold. This means that lowering
the classifier threshold may increase recall, by increasing the number of true positive results. It is also possible that
lowering the threshold may leave recall unchanged, while the precision fluctuates.
The relationship between recall and precision can be observed in the stairstep area of the plot - at the edges of these
steps a small change in the threshold considerably reduces precision, with only a minor gain in recall.
Average precision (AP) summarizes such a plot as the weighted mean of precisions achieved at each threshold, with
the increase in recall from the previous threshold used as the weight:
∑︀
AP = 𝑛 (𝑅𝑛 − 𝑅𝑛−1 )𝑃𝑛
where 𝑃𝑛 and 𝑅𝑛 are the precision and recall at the nth threshold. A pair (𝑅𝑘 , 𝑃𝑘 ) is referred to as an operating point.
AP and the trapezoidal area under the operating points (sklearn.metrics.auc) are common ways to summarize
a precision-recall curve that lead to different results. Read more in the User Guide.
Precision-recall curves are typically used in binary classification to study the output of a classifier. In order to extend
the precision-recall curve and average precision to multi-class or multi-label classification, it is necessary to binarize
the output. One curve can be drawn per label, but one can also draw a precision-recall curve by considering each
element of the label indicator matrix as a binary prediction (micro-averaging).
Note:
See also sklearn.metrics.average_precision_score, sklearn.metrics.recall_score,
sklearn.metrics.precision_score, sklearn.metrics.f1_score
In binary classification settings
Create simple data
Try to differentiate the two first classes of the iris data

import numpy as np
X = iris.data
y = iris.target

# Add noisy features

# Limit to the two first classes, and split into training and test
X_train, X_test, y_train, y_test = train_test_split(X[y < 2], y[y < 2],
test_size=.5,
# Create a simple classifier

classifier = svm.LinearSVC(random_state=random_state)
classifier.fit(X_train, y_train)
y_score = classifier.decision_function(X_test)
Compute the average precision score
from sklearn.metrics import average_precision_score

average_precision = average_precision_score(y_test, y_score)
print('Average precision-recall score: {0:0.2f}'.format(

average_precision))
Out:
Average precision-recall score: 0.88
Plot the Precision-Recall curve
from sklearn.metrics import precision_recall_curve

from sklearn.utils.fixes import signature
precision, recall, _ = precision_recall_curve(y_test, y_score)
# In matplotlib < 1.5, plt.fill_between does not have a 'step' argument

step_kwargs = ({'step': 'post'}
if 'step' in signature(plt.fill_between).parameters
else {})
plt.step(recall, precision, color='b', alpha=0.2,
where='post')
plt.fill_between(recall, precision, alpha=0.2, color='b', **step_kwargs)
plt.xlabel('Recall')
plt.ylabel('Precision')
plt.ylim([0.0, 1.05])
plt.xlim([0.0, 1.0])
plt.title('2-class Precision-Recall curve: AP={0:0.2f}'.format(
average_precision))

In multi-label settings
Create multi-label data, fit, and predict
We create a multi-label dataset, to illustrate the precision-recall in multi-label settings
from sklearn.preprocessing import label_binarize
# Use label_binarize to be multi-label like settings

Y = label_binarize(y, classes=[0, 1, 2])
n_classes = Y.shape[1]
# Split into training and test

X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=.5,
# We use OneVsRestClassifier for multi-label prediction

# Run classifier
classifier = OneVsRestClassifier(svm.LinearSVC(random_state=random_state))
classifier.fit(X_train, Y_train)
y_score = classifier.decision_function(X_test)

The average precision score in multi-label settings
from sklearn.metrics import precision_recall_curve

from sklearn.metrics import average_precision_score
# For each class

precision = dict()
recall = dict()
average_precision = dict()
precision[i], recall[i], _ = precision_recall_curve(Y_test[:, i],
y_score[:, i])
average_precision[i] = average_precision_score(Y_test[:, i], y_score[:, i])
# A "micro-average": quantifying score on all classes jointly

precision["micro"], recall["micro"], _ = precision_recall_curve(Y_test.ravel(),
y_score.ravel())
average_precision["micro"] = average_precision_score(Y_test, y_score,
average="micro")
print('Average precision score, micro-averaged over all classes: {0:0.2f}'
.format(average_precision["micro"]))
Out:
Average precision score, micro-averaged over all classes: 0.43
Plot the micro-averaged Precision-Recall curve
plt.figure()
plt.step(recall['micro'], precision['micro'], color='b', alpha=0.2,
where='post')
plt.fill_between(recall["micro"], precision["micro"], alpha=0.2, color='b',
**step_kwargs)
plt.ylim([0.0, 1.05])
plt.xlim([0.0, 1.0])
plt.title(
'Average precision score, micro-averaged over all classes: AP={0:0.2f}'
.format(average_precision["micro"]))

Plot Precision-Recall curve for each class and iso-f1 curves

# setup plot details
colors = cycle(['navy', 'turquoise', 'darkorange', 'cornflowerblue', 'teal'])
f_scores = np.linspace(0.2, 0.8, num=4)
lines = []
labels = []
for f_score in f_scores:
x = np.linspace(0.01, 1)
y = f_score * x / (2 * x - f_score)
l, = plt.plot(x[y >= 0], y[y >= 0], color='gray', alpha=0.2)
plt.annotate('f1={0:0.1f}'.format(f_score), xy=(0.9, y[45] + 0.02))
lines.append(l)
labels.append('iso-f1 curves')
l, = plt.plot(recall["micro"], precision["micro"], color='gold', lw=2)
lines.append(l)
labels.append('micro-average Precision-recall (area = {0:0.2f})'
''.format(average_precision["micro"]))
for i, color in zip(range(n_classes), colors):

l, = plt.plot(recall[i], precision[i], color=color, lw=2)

lines.append(l)
labels.append('Precision-recall for class {0} (area = {1:0.2f})'
''.format(i, average_precision[i]))
fig = plt.gcf()
fig.subplots_adjust(bottom=0.25)
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.title('Extension of Precision-Recall curve to multi-class')
plt.legend(lines, labels, loc=(0, -.38), prop=dict(size=14))
plt.show()

5.20 Multioutput methods
Examples concerning the sklearn.multioutput module.

5.20.1 Classifier Chain
Example of using classifier chain on a multilabel dataset.

For this example we will use the yeast dataset which contains 2417 datapoints each with 103 features and 14 possible
labels. Each data point has at least one label. As a baseline we first train a logistic regression classifier for each of the
14 labels. To evaluate the performance of these classifiers we predict on a held-out test set and calculate the jaccard
similarity score.
Next we create 10 classifier chains. Each classifier chain contains a logistic regression model for each of the 14 labels.
The models in each chain are ordered randomly. In addition to the 103 features in the dataset, each model gets the
predictions of the preceding models in the chain as features (note that by default at training time each model gets the
true labels as features). These additional features allow each chain to exploit correlations among the classes. The
Jaccard similarity score for each chain tends to be greater than that of the set independent logistic models.
Because the models in each chain are arranged randomly there is significant variation in performance among the
chains. Presumably there is an optimal ordering of the classes in a chain that will yield the best performance. However
we do not know that ordering a priori. Instead we can construct an voting ensemble of classifier chains by averaging
the binary predictions of the chains and apply a threshold of 0.5. The Jaccard similarity score of the ensemble is
greater than that of the independent models and tends to exceed the score of each chain in the ensemble (although this
is not guaranteed with randomly ordered chains).
# Author: Adam Kleczewski

import numpy as np
5.20. Multioutput methods 1259

from sklearn.multioutput import ClassifierChain

from sklearn.metrics import jaccard_similarity_score
print(__doc__)
# Load a multi-label dataset from https://www.openml.org/d/40597

X, Y = fetch_openml('yeast', version=4, return_X_y=True)
Y = Y == 'TRUE'
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=.2,
random_state=0)
# Fit an independent logistic regression model for each class using the
# OneVsRestClassifier wrapper.
base_lr = LogisticRegression(solver='lbfgs')
ovr = OneVsRestClassifier(base_lr)
ovr.fit(X_train, Y_train)
Y_pred_ovr = ovr.predict(X_test)
ovr_jaccard_score = jaccard_similarity_score(Y_test, Y_pred_ovr)
# Fit an ensemble of logistic regression classifier chains and take the

# take the average prediction of all the chains.
chains = [ClassifierChain(base_lr, order='random', random_state=i)
for i in range(10)]
for chain in chains:
chain.fit(X_train, Y_train)
Y_pred_chains = np.array([chain.predict(X_test) for chain in

chains])
chain_jaccard_scores = [jaccard_similarity_score(Y_test, Y_pred_chain >= .5)
for Y_pred_chain in Y_pred_chains]
Y_pred_ensemble = Y_pred_chains.mean(axis=0)
ensemble_jaccard_score = jaccard_similarity_score(Y_test,
Y_pred_ensemble >= .5)
model_scores = [ovr_jaccard_score] + chain_jaccard_scores

model_scores.append(ensemble_jaccard_score)
model_names = ('Independent',
'Chain 1',
'Chain 2',
'Chain 3',
'Chain 4',
'Chain 5',
'Chain 6',
'Chain 7',
'Chain 8',
'Chain 9',
'Chain 10',
'Ensemble')
x_pos = np.arange(len(model_names))
# Plot the Jaccard similarity scores for the independent model, each of the
# chains, and the ensemble (note that the vertical axis on this plot does

# not begin at 0).

ax.grid(True)
ax.set_title('Classifier Chain Ensemble Performance Comparison')
ax.set_xticks(x_pos)
ax.set_xticklabels(model_names, rotation='vertical')
ax.set_ylabel('Jaccard Similarity Score')
ax.set_ylim([min(model_scores) * .9, max(model_scores) * 1.1])
colors = ['r'] + ['b'] * len(chain_jaccard_scores) + ['g']
ax.bar(x_pos, model_scores, alpha=0.5, color=colors)
plt.tight_layout()
plt.show()
5.21 Nearest Neighbors
Examples concerning the sklearn.neighbors module.
5.21.1 Nearest Neighbors regression
Demonstrate the resolution of a regression problem using a k-Nearest Neighbor and the interpolation of the target
using both barycenter and constant weights.
5.21. Nearest Neighbors 1261

print(__doc__)

# Fabian Pedregosa <fabian.pedregosa@inria.fr>
#
# License: BSD 3 clause (C) INRIA
# #############################################################################
import numpy as np
from sklearn import neighbors
np.random.seed(0)
X = np.sort(5 * np.random.rand(40, 1), axis=0)
T = np.linspace(0, 5, 500)[:, np.newaxis]

y[::5] += 1 * (0.5 - np.random.rand(8))
# #############################################################################
n_neighbors = 5

for i, weights in enumerate(['uniform', 'distance']):

knn = neighbors.KNeighborsRegressor(n_neighbors, weights=weights)
y_ = knn.fit(X, y).predict(T)
plt.scatter(X, y, c='k', label='data')
plt.plot(T, y_, c='g', label='prediction')
plt.axis('tight')
plt.legend()
plt.title("KNeighborsRegressor (k = %i, weights = '%s')" % (n_neighbors,
weights))
plt.tight_layout()
plt.show()
5.21.2 Outlier detection with Local Outlier Factor (LOF)
The Local Outlier Factor (LOF) algorithm is an unsupervised anomaly detection method which computes the local
density deviation of a given data point with respect to its neighbors. It considers as outliers the samples that have a
substantially lower density than their neighbors. This example shows how to use LOF for outlier detection which is
the default use case of this estimator in scikit-learn. Note that when LOF is used for outlier detection it has no predict,
decision_function and score_samples methods. See User Guide: for details on the difference between outlier detection
and novelty detection and how to use LOF for novelty detection.
The number of neighbors considered (parameter n_neighbors) is typically set 1) greater than the minimum number of
samples a cluster has to contain, so that other samples can be local outliers relative to this cluster, and 2) smaller than
the maximum number of close by samples that can potentially be local outliers. In practice, such informations are
generally not available, and taking n_neighbors=20 appears to work well in general.

import numpy as np
print(__doc__)
np.random.seed(42)

X_inliers = 0.3 * np.random.randn(100, 2)
X_inliers = np.r_[X_inliers + 2, X_inliers - 2]
# Generate some outliers

X_outliers = np.random.uniform(low=-4, high=4, size=(20, 2))
X = np.r_[X_inliers, X_outliers]
n_outliers = len(X_outliers)
ground_truth = np.ones(len(X), dtype=int)
ground_truth[-n_outliers:] = -1
# fit the model for outlier detection (default)

clf = LocalOutlierFactor(n_neighbors=20, contamination=0.1)
# use fit_predict to compute the predicted labels of the training samples
# (when LOF is used for outlier detection, the estimator has no predict,
# decision_function and score_samples methods).
y_pred = clf.fit_predict(X)

n_errors = (y_pred != ground_truth).sum()

X_scores = clf.negative_outlier_factor_
plt.title("Local Outlier Factor (LOF)")

plt.scatter(X[:, 0], X[:, 1], color='k', s=3., label='Data points')
# plot circles with radius proportional to the outlier scores
radius = (X_scores.max() - X_scores) / (X_scores.max() - X_scores.min())
plt.scatter(X[:, 0], X[:, 1], s=1000 * radius, edgecolors='r',
facecolors='none', label='Outlier scores')
plt.axis('tight')
plt.xlim((-5, 5))
plt.ylim((-5, 5))
plt.xlabel("prediction errors: %d" % (n_errors))
legend = plt.legend(loc='upper left')
legend.legendHandles[0]._sizes = [10]
legend.legendHandles[1]._sizes = [20]
plt.show()
5.21.3 Nearest Neighbors Classification
Sample usage of Nearest Neighbors classification. It will plot the decision boundaries for each class.


print(__doc__)
import numpy as np
from sklearn import neighbors, datasets
n_neighbors = 15

# we only take the first two features. We could avoid this ugly
# slicing by using a two-dim dataset
y = iris.target
# Create color maps

cmap_light = ListedColormap(['#FFAAAA', '#AAFFAA', '#AAAAFF'])
cmap_bold = ListedColormap(['#FF0000', '#00FF00', '#0000FF'])
for weights in ['uniform', 'distance']:

# we create an instance of Neighbours Classifier and fit the data.
clf = neighbors.KNeighborsClassifier(n_neighbors, weights=weights)
clf.fit(X, y)


plt.figure()
plt.pcolormesh(xx, yy, Z, cmap=cmap_light)

plt.scatter(X[:, 0], X[:, 1], c=y, cmap=cmap_bold,
plt.title("3-Class classification (k = %i, weights = '%s')"
% (n_neighbors, weights))
plt.show()
5.21.4 Nearest Centroid Classification
Sample usage of Nearest Centroid classification. It will plot the decision boundaries for each class.


•
Out:
None 0.8133333333333334
0.2 0.82
print(__doc__)
import numpy as np
from sklearn.neighbors import NearestCentroid
n_neighbors = 15

# we only take the first two features. We could avoid this ugly
# slicing by using a two-dim dataset
y = iris.target

# Create color maps

cmap_light = ListedColormap(['#FFAAAA', '#AAFFAA', '#AAAAFF'])
cmap_bold = ListedColormap(['#FF0000', '#00FF00', '#0000FF'])
for shrinkage in [None, .2]:

# we create an instance of Neighbours Classifier and fit the data.
clf = NearestCentroid(shrink_threshold=shrinkage)
clf.fit(X, y)
y_pred = clf.predict(X)
print(shrinkage, np.mean(y == y_pred))

plt.figure()
plt.pcolormesh(xx, yy, Z, cmap=cmap_light)

plt.scatter(X[:, 0], X[:, 1], c=y, cmap=cmap_bold,
plt.title("3-Class classification (shrink_threshold=%r)"
% shrinkage)
plt.axis('tight')
plt.show()
5.21.5 Kernel Density Estimation
This example shows how kernel density estimation (KDE), a powerful non-parametric density estimation technique,
can be used to learn a generative model for a dataset. With this generative model in place, new samples can be drawn.
These new samples reflect the underlying model of the data.

Out:
best bandwidth: 3.79269019073225
import numpy as np

from sklearn.neighbors import KernelDensity
# load the data

# project the 64-dimensional data to a lower dimension

pca = PCA(n_components=15, whiten=False)
data = pca.fit_transform(digits.data)
# use grid search cross-validation to optimize the bandwidth

params = {'bandwidth': np.logspace(-1, 1, 20)}

grid = GridSearchCV(KernelDensity(), params, cv=5)

grid.fit(data)
print("best bandwidth: {0}".format(grid.best_estimator_.bandwidth))
# use the best estimator to compute the kernel density estimate

kde = grid.best_estimator_
# sample 44 new points from the data

new_data = kde.sample(44, random_state=0)
new_data = pca.inverse_transform(new_data)
# turn data into a 4x11 grid

new_data = new_data.reshape((4, 11, -1))
real_data = digits.data[:44].reshape((4, 11, -1))
# plot real digits and resampled digits

fig, ax = plt.subplots(9, 11, subplot_kw=dict(xticks=[], yticks=[]))
for j in range(11):
ax[4, j].set_visible(False)
for i in range(4):
im = ax[i, j].imshow(real_data[i, j].reshape((8, 8)),
cmap=plt.cm.binary, interpolation='nearest')
im.set_clim(0, 16)
im = ax[i + 5, j].imshow(new_data[i, j].reshape((8, 8)),
cmap=plt.cm.binary, interpolation='nearest')
im.set_clim(0, 16)
ax[0, 5].set_title('Selection from the input data')

ax[5, 5].set_title('"New" digits drawn from the kernel density model')
plt.show()
5.21.6 Novelty detection with Local Outlier Factor (LOF)
The Local Outlier Factor (LOF) algorithm is an unsupervised anomaly detection method which computes the local
density deviation of a given data point with respect to its neighbors. It considers as outliers the samples that have a
substantially lower density than their neighbors. This example shows how to use LOF for novelty detection. Note
that when LOF is used for novelty detection you MUST not use predict, decision_function and score_samples on the
training set as this would lead to wrong results. You must only use these methods on new unseen data (which are not
in the training set). See User Guide: for details on the difference between outlier detection and novelty detection and
how to use LOF for outlier detection.
The number of neighbors considered, (parameter n_neighbors) is typically set 1) greater than the minimum number
of samples a cluster has to contain, so that other samples can be local outliers relative to this cluster, and 2) smaller
than the maximum number of close by samples that can potentially be local outliers. In practice, such informations
are generally not available, and taking n_neighbors=20 appears to work well in general.

import numpy as np
import matplotlib
print(__doc__)
np.random.seed(42)

# Generate normal (not abnormal) training observations
X = 0.3 * np.random.randn(100, 2)
X_train = np.r_[X + 2, X - 2]
# Generate new normal (not abnormal) observations
X_test = np.r_[X + 2, X - 2]
# fit the model for novelty detection (novelty=True)

clf = LocalOutlierFactor(n_neighbors=20, novelty=True, contamination=0.1)
clf.fit(X_train)
# DO NOT use predict, decision_function and score_samples on X_train as this
# would give wrong results but only on new unseen data (not used in X_train),
# e.g. X_test, X_outliers or the meshgrid

n_error_test = y_pred_test[y_pred_test == -1].size
n_error_outliers = y_pred_outliers[y_pred_outliers == 1].size
# plot the learned frontier, the points, and the nearest vectors to the plane
plt.title("Novelty Detection with LOF")

plt.contourf(xx, yy, Z, levels=np.linspace(Z.min(), 0, 7), cmap=plt.cm.PuBu)
a = plt.contour(xx, yy, Z, levels=[0], linewidths=2, colors='darkred')
plt.contourf(xx, yy, Z, levels=[0, Z.max()], colors='palevioletred')
s = 40
b1 = plt.scatter(X_train[:, 0], X_train[:, 1], c='white', s=s, edgecolors='k')
b2 = plt.scatter(X_test[:, 0], X_test[:, 1], c='blueviolet', s=s,
edgecolors='k')
c = plt.scatter(X_outliers[:, 0], X_outliers[:, 1], c='gold', s=s,
edgecolors='k')
plt.axis('tight')
plt.xlim((-5, 5))
plt.ylim((-5, 5))
plt.legend([a.collections[0], b1, b2, c],
["learned frontier", "training observations",
loc="upper left",
plt.xlabel(
"errors novel regular: %d/40 ; errors novel abnormal: %d/40"
% (n_error_test, n_error_outliers))
plt.show()
5.21.7 Kernel Density Estimate of Species Distributions
This shows an example of a neighbors-based query (in particular a kernel density estimate) on geospatial data, using
a Ball Tree built upon the Haversine distance metric – i.e. distances over points in latitude/longitude. The dataset
is provided by Phillips et. al. (2006). If available, the example uses basemap to plot the coast lines and national
boundaries of South America.
This example does not perform any learning over the data (see Species distribution modeling for an example of classi-
fication based on the attributes in this dataset). It simply shows the kernel density estimate of observed data points in
geospatial coordinates.

References
• “Maximum entropy modeling of species geographic distributions” S. J. Phillips, R. P. Anderson, R. E. Schapire

- Ecological Modelling, 190:231-259, 2006.
Out:
- computing KDE in spherical coordinates

- computing KDE in spherical coordinates
# Author: Jake Vanderplas <jakevdp@cs.washington.edu>

#
import numpy as np
from sklearn.datasets import fetch_species_distributions
from sklearn.datasets.species_distributions import construct_grids

# if basemap is available, we'll use it.

# otherwise, we'll improvise later...
try:
from mpl_toolkits.basemap import Basemap
basemap = True
except ImportError:
basemap = False
# Get matrices/arrays of species IDs and locations

data = fetch_species_distributions()
species_names = ['Bradypus Variegatus', 'Microryzomys Minutus']
Xtrain = np.vstack([data['train']['dd lat'],

data['train']['dd long']]).T
ytrain = np.array([d.decode('ascii').startswith('micro')
for d in data['train']['species']], dtype='int')
Xtrain *= np.pi / 180. # Convert lat/long to radians
# Set up the data grid for the contour plot

xgrid, ygrid = construct_grids(data)
X, Y = np.meshgrid(xgrid[::5], ygrid[::5][::-1])
land_reference = data.coverages[6][::5, ::5]
land_mask = (land_reference > -9999).ravel()
xy = np.vstack([Y.ravel(), X.ravel()]).T
xy = xy[land_mask]
xy *= np.pi / 180.
# Plot map of South America with distributions of each species

fig = plt.figure()
fig.subplots_adjust(left=0.05, right=0.95, wspace=0.05)
for i in range(2):
# construct a kernel density estimate of the distribution

print(" - computing KDE in spherical coordinates")
kde = KernelDensity(bandwidth=0.04, metric='haversine',
kernel='gaussian', algorithm='ball_tree')
kde.fit(Xtrain[ytrain == i])
# evaluate only on the land: -9999 indicates ocean

Z = np.full(land_mask.shape[0], -9999, dtype='int')
Z[land_mask] = np.exp(kde.score_samples(xy))
Z = Z.reshape(X.shape)
# plot contours of the density

levels = np.linspace(0, Z.max(), 25)
plt.contourf(X, Y, Z, levels=levels, cmap=plt.cm.Reds)
if basemap:
print(" - plot coastlines using basemap")
m = Basemap(projection='cyl', llcrnrlat=Y.min(),
urcrnrlat=Y.max(), llcrnrlon=X.min(),
urcrnrlon=X.max(), resolution='c')
m.drawcoastlines()
m.drawcountries()

else:
print(" - plot coastlines from coverage")
plt.contour(X, Y, land_reference,
levels=[-9998], colors="k",
linestyles="solid")
plt.xticks([])
plt.yticks([])
plt.title(species_names[i])
plt.show()
5.21.8 Simple 1D Kernel Density Estimation
This example uses the sklearn.neighbors.KernelDensity class to demonstrate the principles of Kernel
Density Estimation in one dimension.
The first plot shows one of the problems with using histograms to visualize the density of points in 1D. Intuitively, a
histogram can be thought of as a scheme in which a unit “block” is stacked above each point on a regular grid. As
the top two panels show, however, the choice of gridding for these blocks can lead to wildly divergent ideas about
the underlying shape of the density distribution. If we instead center each block on the point it represents, we get the
estimate shown in the bottom left panel. This is a kernel density estimation with a “top hat” kernel. This idea can be
generalized to other kernel shapes: the bottom-right panel of the first figure shows a Gaussian kernel density estimate
over the same distribution.
Scikit-learn implements efficient kernel density estimation using either a Ball Tree or KD Tree structure, through the
sklearn.neighbors.KernelDensity estimator. The available kernels are shown in the second figure of this
example.
The third figure compares kernel density estimates for a distribution of 100 samples in 1 dimension. Though this
example uses 1D distributions, kernel density estimation is easily and efficiently extensible to higher dimensions as
well.



# Author: Jake Vanderplas <jakevdp@cs.washington.edu>

#
import numpy as np
import matplotlib
from scipy.stats import norm

else:
#----------------------------------------------------------------------
# Plot the progression of histograms to kernels
np.random.seed(1)
N = 20
X = np.concatenate((np.random.normal(0, 1, int(0.3 * N)),
np.random.normal(5, 1, int(0.7 * N))))[:, np.newaxis]
X_plot = np.linspace(-5, 10, 1000)[:, np.newaxis]
bins = np.linspace(-5, 10, 10)
fig, ax = plt.subplots(2, 2, sharex=True, sharey=True)

fig.subplots_adjust(hspace=0.05, wspace=0.05)

# histogram 1
ax[0, 0].hist(X[:, 0], bins=bins, fc='#AAAAFF', **density_param)
ax[0, 0].text(-3.5, 0.31, "Histogram")
# histogram 2
ax[0, 1].hist(X[:, 0], bins=bins + 0.75, fc='#AAAAFF', **density_param)
ax[0, 1].text(-3.5, 0.31, "Histogram, bins shifted")
# tophat KDE
kde = KernelDensity(kernel='tophat', bandwidth=0.75).fit(X)
log_dens = kde.score_samples(X_plot)
ax[1, 0].fill(X_plot[:, 0], np.exp(log_dens), fc='#AAAAFF')
ax[1, 0].text(-3.5, 0.31, "Tophat Kernel Density")
# Gaussian KDE
kde = KernelDensity(kernel='gaussian', bandwidth=0.75).fit(X)
ax[1, 1].fill(X_plot[:, 0], np.exp(log_dens), fc='#AAAAFF')
ax[1, 1].text(-3.5, 0.31, "Gaussian Kernel Density")
for axi in ax.ravel():

axi.plot(X[:, 0], np.full(X.shape[0], -0.01), '+k')
axi.set_xlim(-4, 9)
axi.set_ylim(-0.02, 0.34)
for axi in ax[:, 0]:

axi.set_ylabel('Normalized Density')
for axi in ax[1, :]:

axi.set_xlabel('x')
#----------------------------------------------------------------------
# Plot all available kernels
X_plot = np.linspace(-6, 6, 1000)[:, None]
X_src = np.zeros((1, 1))
fig, ax = plt.subplots(2, 3, sharex=True, sharey=True)

fig.subplots_adjust(left=0.05, right=0.95, hspace=0.05, wspace=0.05)
def format_func(x, loc):

if x == 0:
return '0'
elif x == 1:
return 'h'
elif x == -1:
return '-h'
else:
return '%ih' % x
for i, kernel in enumerate(['gaussian', 'tophat', 'epanechnikov',

'exponential', 'linear', 'cosine']):
axi = ax.ravel()[i]
log_dens = KernelDensity(kernel=kernel).fit(X_src).score_samples(X_plot)
axi.fill(X_plot[:, 0], np.exp(log_dens), '-k', fc='#AAAAFF')
axi.text(-2.6, 0.95, kernel)
axi.xaxis.set_major_formatter(plt.FuncFormatter(format_func))

axi.xaxis.set_major_locator(plt.MultipleLocator(1))
axi.yaxis.set_major_locator(plt.NullLocator())
axi.set_ylim(0, 1.05)
axi.set_xlim(-2.9, 2.9)
ax[0, 1].set_title('Available Kernels')
#----------------------------------------------------------------------
# Plot a 1D density example
N = 100
np.random.seed(1)
X = np.concatenate((np.random.normal(0, 1, int(0.3 * N)),
np.random.normal(5, 1, int(0.7 * N))))[:, np.newaxis]
X_plot = np.linspace(-5, 10, 1000)[:, np.newaxis]
true_dens = (0.3 * norm(0, 1).pdf(X_plot[:, 0])

+ 0.7 * norm(5, 1).pdf(X_plot[:, 0]))
ax.fill(X_plot[:, 0], true_dens, fc='black', alpha=0.2,
label='input distribution')
for kernel in ['gaussian', 'tophat', 'epanechnikov']:

kde = KernelDensity(kernel=kernel, bandwidth=0.5).fit(X)
ax.plot(X_plot[:, 0], np.exp(log_dens), '-',
label="kernel = '{0}'".format(kernel))
ax.text(6, 0.38, "N={0} points".format(N))
ax.legend(loc='upper left')
ax.plot(X[:, 0], -0.005 - 0.01 * np.random.random(X.shape[0]), '+k')
ax.set_xlim(-4, 9)
ax.set_ylim(-0.02, 0.4)
plt.show()
5.22 Neural Networks
Examples concerning the sklearn.neural_network module.
5.22.1 Visualization of MLP weights on MNIST
Sometimes looking at the learned coefficients of a neural network can provide insight into the learning behavior. For
example if weights look unstructured, maybe some were not used at all, or if very large coefficients exist, maybe
regularization was too low or the learning rate too high.
5.22. Neural Networks 1283

This example shows how to plot some of the first layer weights in a MLPClassifier trained on the MNIST dataset.
The input data consists of 28x28 pixel handwritten digits, leading to 784 features in the dataset. Therefore the first
layer weight matrix have the shape (784, hidden_layer_sizes[0]). We can therefore visualize a single column of the
weight matrix as a 28x28 pixel image.
To make the example run faster, we use very few hidden units, and train only for a very short time. Training longer
would result in weights with a much smoother spatial appearance.
Out:
Iteration 1, loss = 0.32009978

Training set score: 0.986800
Test set score: 0.970000


print(__doc__)
# Load data from https://www.openml.org/d/554

X, y = fetch_openml('mnist_784', version=1, return_X_y=True)
X = X / 255.
# rescale the data, use the traditional train/test split

X_train, X_test = X[:60000], X[60000:]
y_train, y_test = y[:60000], y[60000:]
# mlp = MLPClassifier(hidden_layer_sizes=(100, 100), max_iter=400, alpha=1e-4,

# solver='sgd', verbose=10, tol=1e-4, random_state=1)
mlp = MLPClassifier(hidden_layer_sizes=(50,), max_iter=10, alpha=1e-4,
solver='sgd', verbose=10, tol=1e-4, random_state=1,
learning_rate_init=.1)
mlp.fit(X_train, y_train)
print("Training set score: %f" % mlp.score(X_train, y_train))
print("Test set score: %f" % mlp.score(X_test, y_test))
fig, axes = plt.subplots(4, 4)

# use global min / max to ensure all weights are shown on the same scale
vmin, vmax = mlp.coefs_[0].min(), mlp.coefs_[0].max()
for coef, ax in zip(mlp.coefs_[0].T, axes.ravel()):
ax.matshow(coef.reshape(28, 28), cmap=plt.cm.gray, vmin=.5 * vmin,
vmax=.5 * vmax)
ax.set_xticks(())
ax.set_yticks(())
plt.show()
5.22.2 Compare Stochastic learning strategies for MLPClassifier
This example visualizes some training loss curves for different stochastic learning strategies, including SGD and
Adam. Because of time-constraints, we use several small datasets, for which L-BFGS might be more suitable. The
general trend shown in these examples seems to carry over to larger datasets, however.
Note that those results can be highly dependent on the value of learning_rate_init.

Out:
learning on dataset iris

training: constant learning-rate
Training set loss: 0.096950
training: constant with momentum
training: constant with Nesterov's momentum
training: inv-scaling learning-rate
training: inv-scaling with momentum
training: inv-scaling with Nesterov's momentum
training: adam
learning on dataset digits



training: adam
learning on dataset circles

training: adam
learning on dataset moons



training: adam
print(__doc__)
from sklearn.preprocessing import MinMaxScaler
# different learning rate schedules and momentum parameters

params = [{'solver': 'sgd', 'learning_rate': 'constant', 'momentum': 0,
'learning_rate_init': 0.2},
{'solver': 'sgd', 'learning_rate': 'constant', 'momentum': .9,
'nesterovs_momentum': False, 'learning_rate_init': 0.2},
{'solver': 'sgd', 'learning_rate': 'constant', 'momentum': .9,
'nesterovs_momentum': True, 'learning_rate_init': 0.2},
{'solver': 'sgd', 'learning_rate': 'invscaling', 'momentum': 0,
'learning_rate_init': 0.2},
{'solver': 'sgd', 'learning_rate': 'invscaling', 'momentum': .9,
'nesterovs_momentum': True, 'learning_rate_init': 0.2},
{'solver': 'sgd', 'learning_rate': 'invscaling', 'momentum': .9,
'nesterovs_momentum': False, 'learning_rate_init': 0.2},
{'solver': 'adam', 'learning_rate_init': 0.01}]
labels = ["constant learning-rate", "constant with momentum",

"constant with Nesterov's momentum",
"inv-scaling learning-rate", "inv-scaling with momentum",
"inv-scaling with Nesterov's momentum", "adam"]
plot_args = [{'c': 'red', 'linestyle': '-'},

{'c': 'green', 'linestyle': '-'},
{'c': 'blue', 'linestyle': '-'},
{'c': 'red', 'linestyle': '--'},
{'c': 'green', 'linestyle': '--'},
{'c': 'blue', 'linestyle': '--'},
{'c': 'black', 'linestyle': '-'}]
def plot_on_dataset(X, y, ax, name):

# for each dataset, plot learning for each learning strategy
print("\nlearning on dataset %s" % name)
ax.set_title(name)
X = MinMaxScaler().fit_transform(X)
mlps = []
if name == "digits":
# digits is larger but converges fairly quickly
max_iter = 15
else:
max_iter = 400

for label, param in zip(labels, params):

print("training: %s" % label)
mlp = MLPClassifier(verbose=0, random_state=0,
max_iter=max_iter, **param)
mlp.fit(X, y)
mlps.append(mlp)
print("Training set score: %f" % mlp.score(X, y))
print("Training set loss: %f" % mlp.loss_)
for mlp, label, args in zip(mlps, labels, plot_args):
ax.plot(mlp.loss_curve_, label=label, **args)
fig, axes = plt.subplots(2, 2, figsize=(15, 10))

# load / generate some toy datasets
data_sets = [(iris.data, iris.target),
(digits.data, digits.target),
datasets.make_circles(noise=0.2, factor=0.5, random_state=1),
datasets.make_moons(noise=0.3, random_state=0)]
for ax, data, name in zip(axes.ravel(), data_sets, ['iris', 'digits',

'circles', 'moons']):
plot_on_dataset(*data, ax=ax, name=name)
fig.legend(ax.get_lines(), labels, ncol=3, loc="upper center")

plt.show()
5.22.3 Varying regularization in Multi-layer Perceptron
A comparison of different values for regularization parameter ‘alpha’ on synthetic datasets. The plot shows that
different alphas yield different decision functions.
Alpha is a parameter for regularization term, aka penalty term, that combats overfitting by constraining the size of the
weights. Increasing alpha may fix high variance (a sign of overfitting) by encouraging smaller weights, resulting in
a decision boundary plot that appears with lesser curvatures. Similarly, decreasing alpha may fix high bias (a sign of
underfitting) by encouraging larger weights, potentially resulting in a more complicated decision boundary.

print(__doc__)
# Author: Issam H. Laradji

import numpy as np
names = []
for i in alphas:
names.append('alpha ' + str(i))
classifiers = []
for i in alphas:
classifiers.append(MLPClassifier(alpha=i, random_state=1))
X, y = make_classification(n_features=2, n_redundant=0, n_informative=2,

random_state=0, n_clusters_per_class=1)
X += 2 * rng.uniform(size=X.shape)
linearly_separable = (X, y)
datasets = [make_moons(noise=0.3, random_state=0),

make_circles(noise=0.2, factor=0.5, random_state=1),
linearly_separable]


i = 1
for X, y in datasets:
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=.4)

# just plot the dataset first

cm = plt.cm.RdBu
cm_bright = ListedColormap(['#FF0000', '#0000FF'])
ax.scatter(X_train[:, 0], X_train[:, 1], c=y_train, cmap=cm_bright)
# and testing points
ax.scatter(X_test[:, 0], X_test[:, 1], c=y_test, cmap=cm_bright, alpha=0.6)
ax.set_xticks(())
ax.set_yticks(())
i += 1

for name, clf in zip(names, classifiers):
else:


edgecolors='black', s=25)
alpha=0.6, edgecolors='black', s=25)
ax.set_xticks(())
ax.set_yticks(())
ax.set_title(name)
ax.text(xx.max() - .3, yy.min() + .3, ('%.2f' % score).lstrip('0'),
size=15, horizontalalignment='right')

i += 1
figure.subplots_adjust(left=.02, right=.98)
plt.show()
5.22.4 Restricted Boltzmann Machine features for digit classification
For greyscale image data where pixel values can be interpreted as degrees of blackness on a white background, like
handwritten digit recognition, the Bernoulli Restricted Boltzmann machine model (BernoulliRBM ) can perform
effective non-linear feature extraction.
In order to learn good latent representations from a small dataset, we artificially generate more labeled data by per-
turbing the training data with linear shifts of 1 pixel in each direction.
This example shows how to build a classification pipeline with a BernoulliRBM feature extractor and a
LogisticRegression classifier. The hyperparameters of the entire model (learning rate, hidden layer size, regu-
larization) were optimized by grid search, but the search is not reproduced here because of runtime constraints.
Logistic regression on raw pixel values is presented for comparison. The example shows that the features extracted by
the BernoulliRBM help improve the classification accuracy.
Out:
[BernoulliRBM] Iteration 1, pseudo-likelihood = -25.39, time = 0.16s


Logistic regression using RBM features:
0 0.99 0.98 0.99 174

1 0.94 0.94 0.94 184
2 0.90 0.98 0.94 166
3 0.94 0.90 0.92 194
4 0.96 0.94 0.95 186
5 0.91 0.92 0.91 181
6 0.98 0.97 0.97 207
7 0.93 0.99 0.96 154
8 0.91 0.85 0.88 182
9 0.91 0.92 0.91 169
micro avg 0.94 0.94 0.94 1797

macro avg 0.94 0.94 0.94 1797
weighted avg 0.94 0.94 0.94 1797
Logistic regression using raw pixel features:

0 0.90 0.93 0.91 174

1 0.60 0.59 0.59 184
2 0.75 0.85 0.80 166
3 0.77 0.78 0.78 194
4 0.81 0.84 0.82 186
5 0.77 0.75 0.76 181
6 0.90 0.87 0.89 207
7 0.86 0.88 0.87 154
8 0.67 0.58 0.62 182
9 0.75 0.76 0.76 169
micro avg 0.78 0.78 0.78 1797

macro avg 0.78 0.78 0.78 1797
weighted avg 0.78 0.78 0.78 1797

print(__doc__)
# Authors: Yann N. Dauphin, Vlad Niculae, Gabriel Synnaeve

# License: BSD
import numpy as np
from scipy.ndimage import convolve

from sklearn import linear_model, datasets, metrics
from sklearn.neural_network import BernoulliRBM
from sklearn.base import clone
# #############################################################################
# Setting up
def nudge_dataset(X, Y):

"""
This produces a dataset 5 times bigger than the original one,
by moving the 8x8 images in X around by 1px to left, right, down, up
"""
direction_vectors = [
[[0, 1, 0],
[0, 0, 0],
[0, 0, 0]],
[[0, 0, 0],
[1, 0, 0],
[0, 0, 0]],
[[0, 0, 0],
[0, 0, 1],
[0, 0, 0]],
[[0, 0, 0],
[0, 0, 0],
[0, 1, 0]]]
def shift(x, w):

return convolve(x.reshape((8, 8)), mode='constant', weights=w).ravel()
X = np.concatenate([X] +
[np.apply_along_axis(shift, 1, X, vector)
for vector in direction_vectors])
Y = np.concatenate([Y for _ in range(5)], axis=0)
return X, Y
# Load Data
X = np.asarray(digits.data, 'float32')
X, Y = nudge_dataset(X, digits.target)
X = (X - np.min(X, 0)) / (np.max(X, 0) + 0.0001) # 0-1 scaling

X_train, X_test, Y_train, Y_test = train_test_split(

X, Y, test_size=0.2, random_state=0)
# Models we will use

logistic = linear_model.LogisticRegression(solver='lbfgs', max_iter=10000,
multi_class='multinomial')
rbm = BernoulliRBM(random_state=0, verbose=True)
rbm_features_classifier = Pipeline(
steps=[('rbm', rbm), ('logistic', logistic)])
# #############################################################################
# Training
# Hyper-parameters. These were set by cross-validation,

# using a GridSearchCV. Here we are not performing cross-validation to
# save time.
rbm.learning_rate = 0.06
rbm.n_iter = 20
# More components tend to give better prediction performance, but larger
# fitting time
rbm.n_components = 100
logistic.C = 6000
# Training RBM-Logistic Pipeline

rbm_features_classifier.fit(X_train, Y_train)
# Training the Logistic regression classifier directly on the pixel

raw_pixel_classifier = clone(logistic)
raw_pixel_classifier.C = 100.
raw_pixel_classifier.fit(X_train, Y_train)
# #############################################################################
# Evaluation
Y_pred = rbm_features_classifier.predict(X_test)
print("Logistic regression using RBM features:\n%s\n" % (
metrics.classification_report(Y_test, Y_pred)))
Y_pred = raw_pixel_classifier.predict(X_test)
print("Logistic regression using raw pixel features:\n%s\n" % (
metrics.classification_report(Y_test, Y_pred)))
# #############################################################################
# Plotting
for i, comp in enumerate(rbm.components_):
plt.imshow(comp.reshape((8, 8)), cmap=plt.cm.gray_r,
plt.xticks(())
plt.yticks(())
plt.suptitle('100 components extracted by RBM', fontsize=16)
plt.show()

5.23 Preprocessing
Examples concerning the sklearn.preprocessing module.
5.23.1 Using FunctionTransformer to select columns
Shows how to use a function transformer in a pipeline. If you know your dataset’s first principle component is irrelevant
for a classification task, you can use the FunctionTransformer to select all but the first column of the PCA transformed
data.


import numpy as np

from sklearn.preprocessing import FunctionTransformer
def _generate_vector(shift=0.5, noise=15):

return np.arange(1000) + (np.random.rand(1000) - shift) * noise
def generate_dataset():
"""
This dataset is two lines with a slope ~ 1, where one has
a y offset of ~100
"""
return np.vstack((
np.vstack((
_generate_vector(),
_generate_vector() + 100,
)).T,
np.vstack((
_generate_vector(),
_generate_vector(),
)).T,
5.23. Preprocessing 1297

)), np.hstack((np.zeros(1000), np.ones(1000)))
def all_but_first_column(X):
return X[:, 1:]
def drop_first_component(X, y):

"""
Create a pipeline with PCA and the column selector and use it to
transform the dataset.
"""
pipeline = make_pipeline(
PCA(), FunctionTransformer(all_but_first_column),
)
X_train, X_test, y_train, y_test = train_test_split(X, y)
pipeline.fit(X_train, y_train)
return pipeline.transform(X_test), y_test
if __name__ == '__main__':
X, y = generate_dataset()
lw = 0
plt.figure()
plt.scatter(X[:, 0], X[:, 1], c=y, lw=lw)
plt.figure()
X_transformed, y_transformed = drop_first_component(*generate_dataset())
plt.scatter(
X_transformed[:, 0],
np.zeros(len(X_transformed)),
c=y_transformed,
lw=lw,
s=60
)
plt.show()
5.23.2 Using KBinsDiscretizer to discretize continuous features
The example compares prediction result of linear regression (linear model) and decision tree (tree based model) with
and without discretization of real-valued features.
As is shown in the result before discretization, linear model is fast to build and relatively straightforward to interpret,
but can only model linear relationships, while decision tree can build a much more complex model of the data. One
way to make linear model more powerful on continuous data is to use discretization (also known as binning). In the
example, we discretize the feature and one-hot encode the transformed data. Note that if the bins are not reasonably
wide, there would appear to be a substantially increased risk of overfitting, so the discretizer parameters should usually
be tuned under cross validation.
After discretization, linear regression and decision tree make exactly the same prediction. As features are constant
within each bin, any model must predict the same value for all points within a bin. Compared with the result before

discretization, linear model become much more flexible while decision tree gets much less flexible. Note that bin-
ning features generally has no beneficial effect for tree-based models, as these models can learn to split up the data
anywhere.
# Author: Andreas Müller

# Hanmin Qin <qinhanmin2005@sina.com>
import numpy as np

from sklearn.preprocessing import KBinsDiscretizer
print(__doc__)
# construct the dataset

rnd = np.random.RandomState(42)
X = rnd.uniform(-3, 3, size=100)
y = np.sin(X) + rnd.normal(size=len(X)) / 3
X = X.reshape(-1, 1)
# transform the dataset with KBinsDiscretizer

enc = KBinsDiscretizer(n_bins=10, encode='onehot')
X_binned = enc.fit_transform(X)
# predict with original dataset

fig, (ax1, ax2) = plt.subplots(ncols=2, sharey=True, figsize=(10, 4))
line = np.linspace(-3, 3, 1000, endpoint=False).reshape(-1, 1)
reg = LinearRegression().fit(X, y)
ax1.plot(line, reg.predict(line), linewidth=2, color='green',
label="linear regression")
reg = DecisionTreeRegressor(min_samples_split=3, random_state=0).fit(X, y)
ax1.plot(line, reg.predict(line), linewidth=2, color='red',
label="decision tree")
ax1.plot(X[:, 0], y, 'o', c='k')
ax1.legend(loc="best")
ax1.set_ylabel("Regression output")
ax1.set_xlabel("Input feature")
ax1.set_title("Result before discretization")

# predict with transformed dataset

line_binned = enc.transform(line)
reg = LinearRegression().fit(X_binned, y)
ax2.plot(line, reg.predict(line_binned), linewidth=2, color='green',
linestyle='-', label='linear regression')
reg = DecisionTreeRegressor(min_samples_split=3,
random_state=0).fit(X_binned, y)
ax2.plot(line, reg.predict(line_binned), linewidth=2, color='red',
linestyle=':', label='decision tree')
ax2.plot(X[:, 0], y, 'o', c='k')
ax2.vlines(enc.bin_edges_[0], *plt.gca().get_ylim(), linewidth=1, alpha=.2)
ax2.legend(loc="best")
ax2.set_xlabel("Input feature")
ax2.set_title("Result after discretization")
plt.tight_layout()
plt.show()
5.23.3 Demonstrating the different strategies of KBinsDiscretizer
This example presents the different strategies implemented in KBinsDiscretizer:

• ‘uniform’: The discretization is uniform in each feature, which means that the bin widths are constant in each
dimension.
• quantile’: The discretization is done on the quantiled values, which means that each bin has approximately the
same number of samples.
• ‘kmeans’: The discretization is based on the centroids of a KMeans clustering procedure.
The plot shows the regions where the discretized encoding is constant.

# Author: Tom Dupré la Tour

import numpy as np

print(__doc__)
strategies = ['uniform', 'quantile', 'kmeans']
n_samples = 200
centers_0 = np.array([[0, 0], [0, 5], [2, 4], [8, 8]])
centers_1 = np.array([[0, 0], [3, 1]])
# construct the datasets

random_state = 42
X_list = [
np.random.RandomState(random_state).uniform(-3, 3, size=(n_samples, 2)),
make_blobs(n_samples=[n_samples // 10, n_samples * 4 // 10,
n_samples // 10, n_samples * 4 // 10],
cluster_std=0.5, centers=centers_0,
random_state=random_state)[0],
make_blobs(n_samples=[n_samples // 5, n_samples * 4 // 5],
cluster_std=0.5, centers=centers_1,
random_state=random_state)[0],
]


i = 1
for ds_cnt, X in enumerate(X_list):
ax = plt.subplot(len(X_list), len(strategies) + 1, i)
ax.scatter(X[:, 0], X[:, 1], edgecolors='k')
if ds_cnt == 0:
ax.set_title("Input data", size=14)
xx, yy = np.meshgrid(
np.linspace(X[:, 0].min(), X[:, 0].max(), 300),
np.linspace(X[:, 1].min(), X[:, 1].max(), 300))
grid = np.c_[xx.ravel(), yy.ravel()]
ax.set_xticks(())
ax.set_yticks(())
i += 1
# transform the dataset with KBinsDiscretizer
for strategy in strategies:
enc = KBinsDiscretizer(n_bins=4, encode='ordinal', strategy=strategy)
enc.fit(X)
grid_encoded = enc.transform(grid)
ax = plt.subplot(len(X_list), len(strategies) + 1, i)
# horizontal stripes
horizontal = grid_encoded[:, 0].reshape(xx.shape)
ax.contourf(xx, yy, horizontal, alpha=.5)
# vertical stripes
vertical = grid_encoded[:, 1].reshape(xx.shape)
ax.contourf(xx, yy, vertical, alpha=.5)
ax.scatter(X[:, 0], X[:, 1], edgecolors='k')

ax.set_xticks(())
ax.set_yticks(())
if ds_cnt == 0:
ax.set_title("strategy='%s'" % (strategy, ), size=14)
i += 1
plt.tight_layout()
plt.show()

5.23.4 Map data to a normal distribution
This example demonstrates the use of the Box-Cox and Yeo-Johnson transforms through preprocessing.
PowerTransformer to map data from various distributions to a normal distribution.
The power transform is useful as a transformation in modeling problems where homoscedasticity and normality are de-
sired. Below are examples of Box-Cox and Yeo-Johnwon applied to six different probability distributions: Lognormal,
Chi-squared, Weibull, Gaussian, Uniform, and Bimodal.
Note that the transformations successfully map the data to a normal distribution when applied to certain datasets, but
are ineffective with others. This highlights the importance of visualizing the data before and after transformation.
Also note that even though Box-Cox seems to perform better than Yeo-Johnson for lognormal and chi-squared distri-
butions, keep in mind that Box-Cox does not support inputs with negative values.
For comparison, we also add the output from preprocessing.QuantileTransformer. It can force any ar-
bitrary distribution into a gaussian, provided that there are enough training samples (thousands). Because it is a
non-parametric method, it is harder to interpret than the parametric ones (Box-Cox and Yeo-Johnson).
On “small” datasets (less than a few hundred points), the quantile transformer is prone to overfitting. The use of the
power transform is then recommended.

# Author: Eric Chang <ericchang2017@u.northwestern.edu>

# Nicolas Hug <contact@nicolas-hug.com>
import numpy as np

from sklearn.preprocessing import PowerTransformer

from sklearn.preprocessing import QuantileTransformer
print(__doc__)
N_SAMPLES = 1000
FONT_SIZE = 6
BINS = 30
bc = PowerTransformer(method='box-cox')
yj = PowerTransformer(method='yeo-johnson')
qt = QuantileTransformer(output_distribution='normal', random_state=rng)
size = (N_SAMPLES, 1)
# lognormal distribution
X_lognormal = rng.lognormal(size=size)
# chi-squared distribution
df = 3
X_chisq = rng.chisquare(df=df, size=size)
# weibull distribution
a = 50
X_weibull = rng.weibull(a=a, size=size)
# gaussian distribution
loc = 100
X_gaussian = rng.normal(loc=loc, size=size)
# uniform distribution
X_uniform = rng.uniform(low=0, high=1, size=size)
# bimodal distribution
loc_a, loc_b = 100, 105
X_a, X_b = rng.normal(loc=loc_a, size=size), rng.normal(loc=loc_b, size=size)
X_bimodal = np.concatenate([X_a, X_b], axis=0)
# create plots
distributions = [
('Lognormal', X_lognormal),
('Chi-squared', X_chisq),
('Weibull', X_weibull),
('Gaussian', X_gaussian),
('Uniform', X_uniform),
('Bimodal', X_bimodal)
]
colors = ['firebrick', 'darkorange', 'goldenrod',

'seagreen', 'royalblue', 'darkorchid']

fig, axes = plt.subplots(nrows=8, ncols=3, figsize=plt.figaspect(2))

axes = axes.flatten()
axes_idxs = [(0, 3, 6, 9), (1, 4, 7, 10), (2, 5, 8, 11), (12, 15, 18, 21),
(13, 16, 19, 22), (14, 17, 20, 23)]
axes_list = [(axes[i], axes[j], axes[k], axes[l])
for (i, j, k, l) in axes_idxs]
for distribution, color, axes in zip(distributions, colors, axes_list):

name, X = distribution
X_train, X_test = train_test_split(X, test_size=.5)
# perform power transforms and quantile transform

X_trans_bc = bc.fit(X_train).transform(X_test)
lmbda_bc = round(bc.lambdas_[0], 2)
X_trans_yj = yj.fit(X_train).transform(X_test)
lmbda_yj = round(yj.lambdas_[0], 2)
X_trans_qt = qt.fit(X_train).transform(X_test)
ax_original, ax_bc, ax_yj, ax_qt = axes
ax_original.hist(X_train, color=color, bins=BINS)

ax_original.set_title(name, fontsize=FONT_SIZE)
ax_original.tick_params(axis='both', which='major', labelsize=FONT_SIZE)
for ax, X_trans, meth_name, lmbda in zip(

(ax_bc, ax_yj, ax_qt),
(X_trans_bc, X_trans_yj, X_trans_qt),
('Box-Cox', 'Yeo-Johnson', 'Quantile transform'),
(lmbda_bc, lmbda_yj, None)):
ax.hist(X_trans, color=color, bins=BINS)
title = 'After {}'.format(meth_name)
if lmbda is not None:
title += r'\n$\lambda$ = {}'.format(lmbda)
ax.set_title(title, fontsize=FONT_SIZE)
ax.tick_params(axis='both', which='major', labelsize=FONT_SIZE)
ax.set_xlim([-3.5, 3.5])
plt.tight_layout()
plt.show()
5.23.5 Importance of Feature Scaling
Feature scaling through standardization (or Z-score normalization) can be an important preprocessing step for many
machine learning algorithms. Standardization involves rescaling the features such that they have the properties of a
standard normal distribution with a mean of zero and a standard deviation of one.
While many algorithms (such as SVM, K-nearest neighbors, and logistic regression) require features to be normalized,
intuitively we can think of Principle Component Analysis (PCA) as being a prime example of when normalization is
important. In PCA we are interested in the components that maximize the variance. If one component (e.g. human

height) varies less than another (e.g. weight) because of their respective scales (meters vs. kilos), PCA might determine
that the direction of maximal variance more closely corresponds with the ‘weight’ axis, if those features are not scaled.
As a change in height of one meter can be considered much more important than the change in weight of one kilogram,
this is clearly incorrect.
To illustrate this, PCA is performed comparing the use of data with StandardScaler applied, to unscaled data.
The results are visualized and a clear difference noted. The 1st principal component in the unscaled set can be seen. It
can be seen that feature #13 dominates the direction, being a whole two orders of magnitude above the other features.
This is contrasted when observing the principal component for the scaled version of the data. In the scaled version,
the orders of magnitude are roughly the same across all the features.
The dataset used is the Wine Dataset available at UCI. This dataset has continuous features that are heterogeneous in
scale due to differing properties that they measure (i.e alcohol content, and malic acid).
The transformed data is then used to train a naive Bayes classifier, and a clear difference in prediction accuracies is
observed wherein the dataset which is scaled before PCA vastly outperforms the unscaled version.
Out:
Prediction accuracy for the normal test dataset with PCA
81.48%
Prediction accuracy for the standardized test dataset with PCA

98.15%
PC 1 without scaling:
[ 1.76e-03 -8.36e-04 1.55e-04 -5.31e-03 2.02e-02 1.02e-03 1.53e-03

-1.12e-04 6.31e-04 2.33e-03 1.54e-04 7.43e-04 1.00e+00]
PC 1 with scaling:
[ 0.13 -0.26 -0.01 -0.23 0.16 0.39 0.42 -0.28 0.33 -0.11 0.3 0.38
0.28]

from sklearn.datasets import load_wine
print(__doc__)
# Code source: Tyler Lanigan <tylerlanigan@gmail.com>

# Sebastian Raschka <mail@sebastianraschka.com>
RANDOM_STATE = 42
FIG_SIZE = (10, 7)
features, target = load_wine(return_X_y=True)
# Make a train/test split using 30% test size

X_train, X_test, y_train, y_test = train_test_split(features, target,
test_size=0.30,
random_state=RANDOM_STATE)
# Fit to data and predict using pipelined GNB and PCA.

unscaled_clf = make_pipeline(PCA(n_components=2), GaussianNB())
unscaled_clf.fit(X_train, y_train)
pred_test = unscaled_clf.predict(X_test)
# Fit to data and predict using pipelined scaling, GNB and PCA.
std_clf = make_pipeline(StandardScaler(), PCA(n_components=2), GaussianNB())
std_clf.fit(X_train, y_train)
pred_test_std = std_clf.predict(X_test)
# Show prediction accuracies in scaled and unscaled data.

print('\nPrediction accuracy for the normal test dataset with PCA')
print('{:.2%}\n'.format(metrics.accuracy_score(y_test, pred_test)))
print('\nPrediction accuracy for the standardized test dataset with PCA')

print('{:.2%}\n'.format(metrics.accuracy_score(y_test, pred_test_std)))
# Extract PCA from pipeline

pca = unscaled_clf.named_steps['pca']
pca_std = std_clf.named_steps['pca']

# Show first principal components

print('\nPC 1 without scaling:\n', pca.components_[0])
print('\nPC 1 with scaling:\n', pca_std.components_[0])
# Use PCA without and with scale on X_train data for visualization.
X_train_transformed = pca.transform(X_train)
scaler = std_clf.named_steps['standardscaler']
X_train_std_transformed = pca_std.transform(scaler.transform(X_train))
# visualize standardized vs. untouched dataset with PCA performed

fig, (ax1, ax2) = plt.subplots(ncols=2, figsize=FIG_SIZE)
for l, c, m in zip(range(0, 3), ('blue', 'red', 'green'), ('^', 's', 'o')):

ax1.scatter(X_train_transformed[y_train == l, 0],
X_train_transformed[y_train == l, 1],
color=c,
label='class %s' % l,
alpha=0.5,
marker=m
)
for l, c, m in zip(range(0, 3), ('blue', 'red', 'green'), ('^', 's', 'o')):

ax2.scatter(X_train_std_transformed[y_train == l, 0],
X_train_std_transformed[y_train == l, 1],
color=c,
label='class %s' % l,
alpha=0.5,
marker=m
)
ax1.set_title('Training dataset after PCA')

ax2.set_title('Standardized training dataset after PCA')
for ax in (ax1, ax2):

ax.set_xlabel('1st principal component')
ax.set_ylabel('2nd principal component')
ax.legend(loc='upper right')
ax.grid()
plt.tight_layout()
plt.show()
5.23.6 Feature discretization
A demonstration of feature discretization on synthetic classification datasets. Feature discretization decomposes each
feature into a set of bins, here equally distributed in width. The discrete values are then one-hot encoded, and given to
a linear classifier. This preprocessing enables a non-linear behavior even though the classifier is linear.

On this example, the first two rows represent linearly non-separable datasets (moons and concentric circles) while the
third is approximately linearly separable. On the two linearly non-separable datasets, feature discretization largely
increases the performance of linear classifiers. On the linearly separable dataset, feature discretization decreases the
performance of linear classifiers. Two non-linear classifiers are also shown for comparison.
This example should be taken with a grain of salt, as the intuition conveyed does not necessarily carry over to real
datasets. Particularly in high-dimensional spaces, data can more easily be separated linearly. Moreover, using feature
discretization and one-hot encoding increases the number of features, which easily lead to overfitting when the number
of samples is small.
The plots show training points in solid colors and testing points semi-transparent. The lower right shows the classifi-
cation accuracy on the test set.
Out:
dataset 0
---------
LogisticRegression: 0.86
LinearSVC: 0.86
KBinsDiscretizer + LogisticRegression: 0.94
KBinsDiscretizer + LinearSVC: 0.92
GradientBoostingClassifier: 0.90
SVC: 0.94
dataset 1
---------
LinearSVC: 0.40
SVC: 0.84
dataset 2
---------
LinearSVC: 0.98

SVC: 0.98
# Code source: Tom Dupré la Tour

# Adapted from plot_classifier_comparison by Gaël Varoquaux and Andreas Müller
#
import numpy as np
from sklearn.svm import SVC, LinearSVC
from sklearn.ensemble import GradientBoostingClassifier
from sklearn.utils.testing import ignore_warnings
from sklearn.exceptions import ConvergenceWarning
print(__doc__)
def get_name(estimator):
name = estimator.__class__.__name__
if name == 'Pipeline':
name = [get_name(est[1]) for est in estimator.steps]
name = ' + '.join(name)
return name
# list of (estimator, param_grid), where param_grid is used in GridSearchCV

classifiers = [
(LogisticRegression(solver='lbfgs', random_state=0), {
'C': np.logspace(-2, 7, 10)
}),
(LinearSVC(random_state=0), {
}),
(make_pipeline(
KBinsDiscretizer(encode='onehot'),
LogisticRegression(solver='lbfgs', random_state=0)), {
'kbinsdiscretizer__n_bins': np.arange(2, 10),
'logisticregression__C': np.logspace(-2, 7, 10),
}),
(make_pipeline(
KBinsDiscretizer(encode='onehot'), LinearSVC(random_state=0)), {
'kbinsdiscretizer__n_bins': np.arange(2, 10),
'linearsvc__C': np.logspace(-2, 7, 10),

}),
(GradientBoostingClassifier(n_estimators=50, random_state=0), {
'learning_rate': np.logspace(-4, 0, 10)
}),
(SVC(random_state=0, gamma='scale'), {
}),
]
names = [get_name(e) for e, g in classifiers]
n_samples = 100
datasets = [
make_moons(n_samples=n_samples, noise=0.2, random_state=0),
make_circles(n_samples=n_samples, noise=0.2, factor=0.5, random_state=1),
make_classification(n_samples=n_samples, n_features=2, n_redundant=0,
n_informative=2, random_state=2,
]
fig, axes = plt.subplots(nrows=len(datasets), ncols=len(classifiers) + 1,

figsize=(21, 9))
cm = plt.cm.PiYG
cm_bright = ListedColormap(['#b30065', '#178000'])

for ds_cnt, (X, y) in enumerate(datasets):
print('\ndataset %d\n---------' % ds_cnt)

X, y, test_size=.5, random_state=42)
# create the grid for background colors

xx, yy = np.meshgrid(
np.arange(x_min, x_max, h), np.arange(y_min, y_max, h))
# plot the dataset first

ax = axes[ds_cnt, 0]
if ds_cnt == 0:
ax.set_title("Input data")
# plot the training points
edgecolors='k')
ax.scatter(X_test[:, 0], X_test[:, 1], c=y_test, cmap=cm_bright, alpha=0.6,
edgecolors='k')
ax.set_xticks(())
ax.set_yticks(())

for est_idx, (name, (estimator, param_grid)) in \

enumerate(zip(names, classifiers)):
ax = axes[ds_cnt, est_idx + 1]
clf = GridSearchCV(estimator=estimator, param_grid=param_grid, cv=5,

iid=False)
with ignore_warnings(category=ConvergenceWarning):
print('%s: %.2f' % (name, score))
# plot the decision boundary. For that, we will assign a color to each
# point in the mesh [x_min, x_max]*[y_min, y_max].
else:
# put the result into a color plot

# plot the training points

edgecolors='k')
edgecolors='k', alpha=0.6)
ax.set_xticks(())
ax.set_yticks(())
if ds_cnt == 0:
ax.set_title(name.replace(' + ', '\n'))
ax.text(0.95, 0.06, ('%.2f' % score).lstrip('0'), size=15,
bbox=dict(boxstyle='round', alpha=0.8, facecolor='white'),
transform=ax.transAxes, horizontalalignment='right')
plt.tight_layout()
# Add suptitles above the figure

plt.subplots_adjust(top=0.90)
suptitles = [
'Linear classifiers',
'Feature discretization and linear classifiers',
'Non-linear classifiers',
]
for i, suptitle in zip([1, 3, 5], suptitles):
ax = axes[0, i]
ax.text(1.05, 1.25, suptitle, transform=ax.transAxes,
horizontalalignment='center', size='x-large')
plt.show()

5.23.7 Compare the effect of different scalers on data with outliers
Feature 0 (median income in a block) and feature 5 (number of households) of the California housing dataset have
very different scales and contain some very large outliers. These two characteristics lead to difficulties to visualize
the data and, more importantly, they can degrade the predictive performance of many machine learning algorithms.
Unscaled data can also slow down or even prevent the convergence of many gradient-based estimators.
Indeed many estimators are designed with the assumption that each feature takes values close to zero or more im-
portantly that all features vary on comparable scales. In particular, metric-based and gradient-based estimators often
assume approximately standardized data (centered features with unit variances). A notable exception are decision
tree-based estimators that are robust to arbitrary scaling of the data.
This example uses different scalers, transformers, and normalizers to bring the data within a pre-defined range.
Scalers are linear (or more precisely affine) transformers and differ from each other in the way to estimate the param-
eters used to shift and scale each feature.
QuantileTransformer provides non-linear transformations in which distances between marginal outliers and
inliers are shrunk. PowerTransformer provides non-linear transformations in which data is mapped to a normal
distribution to stabilize variance and minimize skewness.
Unlike the previous transformations, normalization refers to a per sample transformation instead of a per feature
transformation.
The following code is a bit verbose, feel free to jump directly to the analysis of the results.
# Author: Raghav RV <rvraghav93@gmail.com>

# Guillaume Lemaitre <g.lemaitre58@gmail.com>
# Thomas Unterthiner
import numpy as np

from sklearn.preprocessing import MinMaxScaler

from sklearn.preprocessing import minmax_scale
from sklearn.preprocessing import MaxAbsScaler
from sklearn.preprocessing import RobustScaler
from sklearn.preprocessing import Normalizer
from sklearn.preprocessing import QuantileTransformer
from sklearn.preprocessing import PowerTransformer
from sklearn.datasets import fetch_california_housing
print(__doc__)
dataset = fetch_california_housing()
X_full, y_full = dataset.data, dataset.target

# Take only 2 features to make visualization easier

# Feature of 0 has a long tail distribution.
# Feature 5 has a few but very large outliers.
X = X_full[:, [0, 5]]
distributions = [
('Unscaled data', X),
('Data after standard scaling',
StandardScaler().fit_transform(X)),
('Data after min-max scaling',
MinMaxScaler().fit_transform(X)),
('Data after max-abs scaling',
MaxAbsScaler().fit_transform(X)),
('Data after robust scaling',
RobustScaler(quantile_range=(25, 75)).fit_transform(X)),
('Data after power transformation (Yeo-Johnson)',
PowerTransformer(method='yeo-johnson').fit_transform(X)),
('Data after power transformation (Box-Cox)',
PowerTransformer(method='box-cox').fit_transform(X)),
('Data after quantile transformation (gaussian pdf)',
QuantileTransformer(output_distribution='normal')
.fit_transform(X)),
('Data after quantile transformation (uniform pdf)',
QuantileTransformer(output_distribution='uniform')
.fit_transform(X)),
('Data after sample-wise L2 normalizing',
Normalizer().fit_transform(X)),
]
# scale the output between 0 and 1 for the colorbar

y = minmax_scale(y_full)
# plasma does not exist in matplotlib < 1.5

cmap = getattr(cm, 'plasma_r', cm.hot_r)
def create_axes(title, figsize=(16, 6)):

fig = plt.figure(figsize=figsize)
fig.suptitle(title)
# define the axis for the first plot

left, width = 0.1, 0.22
bottom, height = 0.1, 0.7
bottom_h = height + 0.15
left_h = left + width + 0.02
rect_scatter = [left, bottom, width, height]

rect_histx = [left, bottom_h, width, 0.1]
rect_histy = [left_h, bottom, 0.05, height]
ax_scatter = plt.axes(rect_scatter)
ax_histx = plt.axes(rect_histx)
ax_histy = plt.axes(rect_histy)
# define the axis for the zoomed-in plot

left = width + left + 0.2
left_h = left + width + 0.02

rect_scatter = [left, bottom, width, height]

rect_histx = [left, bottom_h, width, 0.1]
rect_histy = [left_h, bottom, 0.05, height]
ax_scatter_zoom = plt.axes(rect_scatter)
ax_histx_zoom = plt.axes(rect_histx)
ax_histy_zoom = plt.axes(rect_histy)
# define the axis for the colorbar

left, width = width + left + 0.13, 0.01
rect_colorbar = [left, bottom, width, height]

ax_colorbar = plt.axes(rect_colorbar)
return ((ax_scatter, ax_histy, ax_histx),

(ax_scatter_zoom, ax_histy_zoom, ax_histx_zoom),
ax_colorbar)
def plot_distribution(axes, X, y, hist_nbins=50, title="",

x0_label="", x1_label=""):
ax, hist_X1, hist_X0 = axes
ax.set_title(title)
ax.set_xlabel(x0_label)
ax.set_ylabel(x1_label)
# The scatter plot

colors = cmap(y)
ax.scatter(X[:, 0], X[:, 1], alpha=0.5, marker='o', s=5, lw=0, c=colors)
# Removing the top and the right spine for aesthetics

# make nice axis layout
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.get_xaxis().tick_bottom()
ax.get_yaxis().tick_left()
ax.spines['left'].set_position(('outward', 10))
ax.spines['bottom'].set_position(('outward', 10))
# Histogram for axis X1 (feature 5)

hist_X1.set_ylim(ax.get_ylim())
hist_X1.hist(X[:, 1], bins=hist_nbins, orientation='horizontal',
color='grey', ec='grey')
hist_X1.axis('off')
# Histogram for axis X0 (feature 0)

hist_X0.set_xlim(ax.get_xlim())
hist_X0.hist(X[:, 0], bins=hist_nbins, orientation='vertical',
color='grey', ec='grey')
hist_X0.axis('off')
Two plots will be shown for each scaler/normalizer/transformer. The left figure will show a scatter plot of the full data
set while the right figure will exclude the extreme values considering only 99 % of the data set, excluding marginal
outliers. In addition, the marginal distributions for each feature will be shown on the side of the scatter plot.

def make_plot(item_idx):
title, X = distributions[item_idx]
ax_zoom_out, ax_zoom_in, ax_colorbar = create_axes(title)
axarr = (ax_zoom_out, ax_zoom_in)
plot_distribution(axarr[0], X, y, hist_nbins=200,
x0_label="Median Income",
x1_label="Number of households",
title="Full data")
# zoom-in
zoom_in_percentile_range = (0, 99)
cutoffs_X0 = np.percentile(X[:, 0], zoom_in_percentile_range)
cutoffs_X1 = np.percentile(X[:, 1], zoom_in_percentile_range)
non_outliers_mask = (
np.all(X > [cutoffs_X0[0], cutoffs_X1[0]], axis=1) &
np.all(X < [cutoffs_X0[1], cutoffs_X1[1]], axis=1))
plot_distribution(axarr[1], X[non_outliers_mask], y[non_outliers_mask],
hist_nbins=50,
x0_label="Median Income",
x1_label="Number of households",
title="Zoom-in")
norm = mpl.colors.Normalize(y_full.min(), y_full.max())

mpl.colorbar.ColorbarBase(ax_colorbar, cmap=cmap,
norm=norm, orientation='vertical',
label='Color mapping for values of y')
Original data
Each transformation is plotted showing two transformed features, with the left plot showing the entire dataset, and
the right zoomed-in to show the dataset without the marginal outliers. A large majority of the samples are compacted
to a specific range, [0, 10] for the median income and [0, 6] for the number of households. Note that there are
some marginal outliers (some blocks have more than 1200 households). Therefore, a specific pre-processing can
be very beneficial depending of the application. In the following, we present some insights and behaviors of those
pre-processing methods in the presence of marginal outliers.
make_plot(0)

StandardScaler
StandardScaler removes the mean and scales the data to unit variance. However, the outliers have an influence
when computing the empirical mean and standard deviation which shrink the range of the feature values as shown in
the left figure below. Note in particular that because the outliers on each feature have different magnitudes, the spread
of the transformed data on each feature is very different: most of the data lie in the [-2, 4] range for the transformed
median income feature while the same data is squeezed in the smaller [-0.2, 0.2] range for the transformed number of
households.
StandardScaler therefore cannot guarantee balanced feature scales in the presence of outliers.
make_plot(1)
MinMaxScaler
MinMaxScaler rescales the data set such that all feature values are in the range [0, 1] as shown in the right panel
below. However, this scaling compress all inliers in the narrow range [0, 0.005] for the transformed number of
households.
As StandardScaler, MinMaxScaler is very sensitive to the presence of outliers.
make_plot(2)

MaxAbsScaler
MaxAbsScaler differs from the previous scaler such that the absolute values are mapped in the range [0, 1]. On
positive only data, this scaler behaves similarly to MinMaxScaler and therefore also suffers from the presence of
large outliers.
make_plot(3)
RobustScaler
Unlike the previous scalers, the centering and scaling statistics of this scaler are based on percentiles and are therefore
not influenced by a few number of very large marginal outliers. Consequently, the resulting range of the transformed
feature values is larger than for the previous scalers and, more importantly, are approximately similar: for both features
most of the transformed values lie in a [-2, 3] range as seen in the zoomed-in figure. Note that the outliers themselves
are still present in the transformed data. If a separate outlier clipping is desirable, a non-linear transformation is
required (see below).
make_plot(4)

PowerTransformer
PowerTransformer applies a power transformation to each feature to make the data more Gaussian-like. Cur-
rently, PowerTransformer implements the Yeo-Johnson and Box-Cox transforms. The power transform finds
the optimal scaling factor to stabilize variance and mimimize skewness through maximum likelihood estimation. By
default, PowerTransformer also applies zero-mean, unit variance normalization to the transformed output. Note
that Box-Cox can only be applied to strictly positive data. Income and number of households happen to be strictly
positive, but if negative values are present the Yeo-Johnson transformed is to be preferred.
make_plot(5)
make_plot(6)
QuantileTransformer (Gaussian output)
QuantileTransformer has an additional output_distribution parameter allowing to match a Gaussian

distribution instead of a uniform distribution. Note that this non-parametetric transformer introduces saturation arti-
facts for extreme values.
make_plot(7)

QuantileTransformer (uniform output)
QuantileTransformer applies a non-linear transformation such that the probability density function of each
feature will be mapped to a uniform distribution. In this case, all the data will be mapped in the range [0, 1], even the
outliers which cannot be distinguished anymore from the inliers.
As RobustScaler, QuantileTransformer is robust to outliers in the sense that adding or removing outliers in
the training set will yield approximately the same transformation on held out data. But contrary to RobustScaler,
QuantileTransformer will also automatically collapse any outlier by setting them to the a priori defined range
boundaries (0 and 1).
make_plot(8)
Normalizer
The Normalizer rescales the vector for each sample to have unit norm, independently of the distribution of the
samples. It can be seen on both figures below where all samples are mapped onto the unit circle. In our example the
two selected features have only positive values; therefore the transformed data only lie in the positive quadrant. This
would not be the case if some original features had a mix of positive and negative values.

make_plot(9)
plt.show()
5.24 Semi Supervised Classification
Examples concerning the sklearn.semi_supervised module.
5.24.1 Decision boundary of label propagation versus SVM on the Iris dataset
Comparison for decision boundary generated on iris dataset between Label Propagation and SVM.
This demonstrates Label Propagation learning a good boundary even with a small amount of labeled data.

print(__doc__)
# Authors: Clay Woolam <clay@woolam.org>

# License: BSD
import numpy as np
from sklearn.semi_supervised import label_propagation
y = iris.target
# step size in the mesh

h = .02
y_30 = np.copy(y)
y_30[rng.rand(len(y)) < 0.3] = -1
y_50 = np.copy(y)
y_50[rng.rand(len(y)) < 0.5] = -1
# we create an instance of SVM and fit out data. We do not scale our
5.24. Semi Supervised Classification 1323

# data since we want to plot the support vectors

ls30 = (label_propagation.LabelSpreading().fit(X, y_30),
y_30)
ls50 = (label_propagation.LabelSpreading().fit(X, y_50),
y_50)
ls100 = (label_propagation.LabelSpreading().fit(X, y), y)
rbf_svc = (svm.SVC(kernel='rbf', gamma=.5).fit(X, y), y)


titles = ['Label Spreading 30% data',
'Label Spreading 50% data',
'Label Spreading 100% data',
'SVC with rbf kernel']
color_map = {-1: (1, 1, 1), 0: (0, 0, .9), 1: (1, 0, 0), 2: (.8, .6, 0)}
for i, (clf, y_train) in enumerate((ls30, ls50, ls100, rbf_svc)):


plt.contourf(xx, yy, Z, cmap=plt.cm.Paired)
plt.axis('off')

colors = [color_map[y] for y in y_train]
plt.scatter(X[:, 0], X[:, 1], c=colors, edgecolors='black')
plt.title(titles[i])
plt.suptitle("Unlabeled points are colored white", y=0.1)

plt.show()
5.24.2 Label Propagation learning a complex structure
Example of LabelPropagation learning a complex internal structure to demonstrate “manifold learning”. The outer
circle should be labeled “red” and the inner circle “blue”. Because both label groups lie inside their own distinct
shape, we can see that the labels propagate correctly around the circle.

print(__doc__)

# License: BSD
import numpy as np
# generate ring with inner box

n_samples = 200
X, y = make_circles(n_samples=n_samples, shuffle=False)
outer, inner = 0, 1
labels = np.full(n_samples, -1.)
labels[0] = outer
labels[-1] = inner
# #############################################################################
# Learn with LabelSpreading
label_spread = label_propagation.LabelSpreading(kernel='knn', alpha=0.8)
label_spread.fit(X, labels)
# #############################################################################
# Plot output labels
output_labels = label_spread.transduction_
plt.scatter(X[labels == outer, 0], X[labels == outer, 1], color='navy',
marker='s', lw=0, label="outer labeled", s=10)
plt.scatter(X[labels == inner, 0], X[labels == inner, 1], color='c',
marker='s', lw=0, label='inner labeled', s=10)
plt.scatter(X[labels == -1, 0], X[labels == -1, 1], color='darkorange',
marker='.', label='unlabeled')
plt.legend(scatterpoints=1, shadow=False, loc='upper right')
plt.title("Raw data (2 classes=outer and inner)")

output_label_array = np.asarray(output_labels)
outer_numbers = np.where(output_label_array == outer)[0]
inner_numbers = np.where(output_label_array == inner)[0]
plt.scatter(X[outer_numbers, 0], X[outer_numbers, 1], color='navy',
marker='s', lw=0, s=10, label="outer learned")
plt.scatter(X[inner_numbers, 0], X[inner_numbers, 1], color='c',
marker='s', lw=0, s=10, label="inner learned")
plt.legend(scatterpoints=1, shadow=False, loc='upper right')
plt.title("Labels learned with Label Spreading (KNN)")
plt.subplots_adjust(left=0.07, bottom=0.07, right=0.93, top=0.92)

plt.show()
5.24.3 Label Propagation digits: Demonstrating performance
This example demonstrates the power of semisupervised learning by training a Label Spreading model to classify
handwritten digits with sets of very few labels.
The handwritten digit dataset has 1797 total points. The model will be trained using all points, but only 30 will be
labeled. Results in the form of a confusion matrix and a series of metrics over each class will be very good.
At the end, the top 10 most uncertain predictions will be shown.

Out:
Label Spreading model: 30 labeled & 300 unlabeled points (330 total)
0 1.00 1.00 1.00 23

1 0.58 0.50 0.54 28
2 0.93 0.93 0.93 29
3 0.00 0.00 0.00 28
4 0.92 0.88 0.90 25
5 0.96 0.76 0.85 33
6 0.97 0.97 0.97 36
7 0.89 1.00 0.94 34
8 0.51 0.79 0.62 29
9 0.51 0.80 0.62 35
micro avg 0.77 0.77 0.77 300

macro avg 0.73 0.76 0.74 300
weighted avg 0.73 0.77 0.74 300
Confusion matrix
[[23 0 0 0 0 0 0 0 0]
[ 0 14 2 0 0 1 0 11 0]
[ 0 0 27 0 0 0 2 0 0]
[ 0 3 0 22 0 0 0 0 0]
[ 0 0 0 0 25 0 0 0 8]
[ 0 1 0 0 0 35 0 0 0]

[ 0 0 0 0 0 0 34 0 0]
[ 0 6 0 0 0 0 0 23 0]
[ 0 0 0 2 1 0 2 2 28]]
print(__doc__)

# License: BSD
import numpy as np

from sklearn.metrics import confusion_matrix, classification_report
indices = np.arange(len(digits.data))
rng.shuffle(indices)
X = digits.data[indices[:330]]
y = digits.target[indices[:330]]
images = digits.images[indices[:330]]
n_total_samples = len(y)
n_labeled_points = 30
indices = np.arange(n_total_samples)
unlabeled_set = indices[n_labeled_points:]
# #############################################################################
# Shuffle everything around
y_train = np.copy(y)
y_train[unlabeled_set] = -1
# #############################################################################
# Learn with LabelSpreading
lp_model = label_propagation.LabelSpreading(gamma=0.25, max_iter=5)
lp_model.fit(X, y_train)
predicted_labels = lp_model.transduction_[unlabeled_set]
true_labels = y[unlabeled_set]
cm = confusion_matrix(true_labels, predicted_labels, labels=lp_model.classes_)
print("Label Spreading model: %d labeled & %d unlabeled points (%d total)" %

(n_labeled_points, n_total_samples - n_labeled_points, n_total_samples))
print(classification_report(true_labels, predicted_labels))

print("Confusion matrix")
print(cm)
# #############################################################################
# Calculate uncertainty values for each transduced distribution
pred_entropies = stats.distributions.entropy(lp_model.label_distributions_.T)
# #############################################################################
# Pick the top 10 most uncertain labels
uncertainty_index = np.argsort(pred_entropies)[-10:]
# #############################################################################
# Plot
f = plt.figure(figsize=(7, 5))
for index, image_index in enumerate(uncertainty_index):
image = images[image_index]
sub = f.add_subplot(2, 5, index + 1)

sub.imshow(image, cmap=plt.cm.gray_r)
plt.xticks([])
plt.yticks([])
sub.set_title('predict: %i\ntrue: %i' % (
lp_model.transduction_[image_index], y[image_index]))
f.suptitle('Learning with small amount of labeled data')

plt.show()
5.24.4 Label Propagation digits active learning
Demonstrates an active learning technique to learn handwritten digits using label propagation.
We start by training a label propagation model with only 10 labeled points, then we select the top five most uncertain
points to label. Next, we train with 15 labeled points (original 10 + 5 new ones). We repeat this process four times
to have a model trained with 30 labeled examples. Note you can increase this to label more than 30 by changing
max_iterations. Labeling more than 30 can be useful to get a sense for the speed of convergence of this active learning
technique.
A plot will appear showing the top 5 most uncertain digits for each iteration of training. These may or may not contain
mistakes, but we will train the next model with their true labels.

Out:
Iteration 0 ______________________________________________________________________
Label Spreading model: 10 labeled & 320 unlabeled (330 total)
0 0.00 0.00 0.00 24

1 0.51 0.86 0.64 29
2 0.83 0.97 0.90 31
3 0.00 0.00 0.00 28
4 0.00 0.00 0.00 27
5 0.85 0.49 0.62 35
6 0.84 0.95 0.89 40
7 0.70 0.92 0.80 36
8 0.57 0.76 0.65 33
9 0.41 0.86 0.55 37
micro avg 0.62 0.62 0.62 320

macro avg 0.47 0.58 0.50 320
weighted avg 0.51 0.62 0.54 320
Confusion matrix
[[25 3 0 0 0 0 1]
[ 1 30 0 0 0 0 0]
[ 0 0 17 7 0 1 10]
[ 2 0 0 38 0 0 0]

[ 0 3 0 0 33 0 0]
[ 8 0 0 0 0 25 0]
[ 0 0 3 0 0 2 32]]
Iteration 1 ______________________________________________________________________
0 0.00 0.00 0.00 24

1 0.51 0.75 0.61 28
2 0.91 0.97 0.94 31
3 0.00 0.00 0.00 28
4 0.00 0.00 0.00 27
5 0.84 0.97 0.90 33
6 1.00 0.95 0.97 40
7 0.75 0.92 0.83 36
8 0.46 0.81 0.59 31
9 0.43 0.78 0.56 37
micro avg 0.66 0.66 0.66 315

macro avg 0.49 0.61 0.54 315
weighted avg 0.53 0.66 0.58 315
Confusion matrix
[[21 0 0 0 0 6 1]
[ 1 30 0 0 0 0 0]
[ 0 0 32 0 0 0 1]
[ 2 0 0 38 0 0 0]
[ 0 3 0 0 33 0 0]
[ 6 0 0 0 0 25 0]
[ 0 0 6 0 0 2 29]]
Iteration 2 ______________________________________________________________________
0 1.00 1.00 1.00 22

1 0.67 0.71 0.69 28
2 0.94 0.97 0.95 31
3 0.00 0.00 0.00 28
4 0.85 0.92 0.88 24
5 0.89 0.97 0.93 33
6 1.00 0.95 0.97 40
7 1.00 0.92 0.96 36
8 0.50 0.81 0.62 31
9 0.67 0.78 0.72 37
micro avg 0.81 0.81 0.81 310

macro avg 0.75 0.80 0.77 310
weighted avg 0.76 0.81 0.78 310
Confusion matrix
[[22 0 0 0 0 0 0 0 0]
[ 0 20 0 1 0 0 0 6 1]
[ 0 1 30 0 0 0 0 0 0]
[ 0 1 0 22 0 0 0 1 0]
[ 0 0 0 0 32 0 0 0 1]
[ 0 2 0 0 0 38 0 0 0]
[ 0 0 2 1 0 0 33 0 0]
[ 0 6 0 0 0 0 0 25 0]

[ 0 0 0 2 4 0 0 2 29]]
Iteration 3 ______________________________________________________________________
0 1.00 1.00 1.00 22

1 0.68 0.85 0.75 27
2 1.00 0.90 0.95 31
3 1.00 0.77 0.87 26
4 1.00 0.92 0.96 24
5 0.89 0.97 0.93 33
6 1.00 0.97 0.99 39
7 0.95 1.00 0.97 35
8 0.66 0.81 0.72 31
9 0.97 0.78 0.87 37
micro avg 0.90 0.90 0.90 305

macro avg 0.91 0.90 0.90 305
weighted avg 0.91 0.90 0.90 305
Confusion matrix
[[22 0 0 0 0 0 0 0 0 0]
[ 0 23 0 0 0 0 0 0 4 0]
[ 0 1 28 0 0 0 0 2 0 0]
[ 0 0 0 20 0 0 0 0 6 0]
[ 0 1 0 0 22 0 0 0 1 0]
[ 0 0 0 0 0 32 0 0 0 1]
[ 0 1 0 0 0 0 38 0 0 0]
[ 0 0 0 0 0 0 0 35 0 0]
[ 0 6 0 0 0 0 0 0 25 0]
[ 0 2 0 0 0 4 0 0 2 29]]
Iteration 4 ______________________________________________________________________
0 1.00 1.00 1.00 22

1 0.68 0.85 0.75 27
2 1.00 0.87 0.93 31
3 0.92 1.00 0.96 23
4 1.00 0.92 0.96 24
5 0.97 0.94 0.95 33
6 1.00 0.97 0.99 39
7 0.95 1.00 0.97 35
8 0.81 0.81 0.81 31
9 0.94 0.86 0.90 35
micro avg 0.92 0.92 0.92 300

macro avg 0.93 0.92 0.92 300
weighted avg 0.93 0.92 0.92 300
Confusion matrix
[[22 0 0 0 0 0 0 0 0 0]
[ 0 23 0 0 0 0 0 0 4 0]
[ 0 1 27 1 0 0 0 2 0 0]
[ 0 0 0 23 0 0 0 0 0 0]
[ 0 1 0 0 22 0 0 0 1 0]
[ 0 0 0 0 0 31 0 0 0 2]
[ 0 1 0 0 0 0 38 0 0 0]

[ 0 0 0 0 0 0 0 35 0 0]
[ 0 6 0 0 0 0 0 0 25 0]
[ 0 2 0 1 0 1 0 0 1 30]]
print(__doc__)

# License: BSD
import numpy as np

from sklearn.metrics import classification_report, confusion_matrix
indices = np.arange(len(digits.data))
rng.shuffle(indices)
X = digits.data[indices[:330]]
y = digits.target[indices[:330]]
images = digits.images[indices[:330]]
n_total_samples = len(y)
n_labeled_points = 10
max_iterations = 5
unlabeled_indices = np.arange(n_total_samples)[n_labeled_points:]
f = plt.figure()
for i in range(max_iterations):
if len(unlabeled_indices) == 0:
print("No unlabeled items left to label.")
break
y_train = np.copy(y)
y_train[unlabeled_indices] = -1
lp_model = label_propagation.LabelSpreading(gamma=0.25, max_iter=5)

lp_model.fit(X, y_train)
predicted_labels = lp_model.transduction_[unlabeled_indices]
true_labels = y[unlabeled_indices]
cm = confusion_matrix(true_labels, predicted_labels,
labels=lp_model.classes_)
print("Iteration %i %s" % (i, 70 * "_"))

print("Label Spreading model: %d labeled & %d unlabeled (%d total)"
% (n_labeled_points, n_total_samples - n_labeled_points,
n_total_samples))

print(classification_report(true_labels, predicted_labels))
print("Confusion matrix")
print(cm)
# compute the entropies of transduced label distributions

pred_entropies = stats.distributions.entropy(
lp_model.label_distributions_.T)
# select up to 5 digit examples that the classifier is most uncertain about

uncertainty_index = np.argsort(pred_entropies)[::-1]
uncertainty_index = uncertainty_index[
np.in1d(uncertainty_index, unlabeled_indices)][:5]
# keep track of indices that we get labels for

delete_indices = np.array([])
# for more than 5 iterations, visualize the gain only on the first 5
if i < 5:
f.text(.05, (1 - (i + 1) * .183),
"model %d\n\nfit with\n%d labels" %
((i + 1), i * 5 + 10), size=10)
for index, image_index in enumerate(uncertainty_index):
image = images[image_index]
# for more than 5 iterations, visualize the gain only on the first 5
if i < 5:
sub = f.add_subplot(5, 5, index + 1 + (5 * i))
sub.imshow(image, cmap=plt.cm.gray_r, interpolation='none')
sub.set_title("predict: %i\ntrue: %i" % (
lp_model.transduction_[image_index], y[image_index]), size=10)
sub.axis('off')
# labeling 5 points, remote from labeled set

delete_index, = np.where(unlabeled_indices == image_index)
delete_indices = np.concatenate((delete_indices, delete_index))
unlabeled_indices = np.delete(unlabeled_indices, delete_indices)

n_labeled_points += len(uncertainty_index)
f.suptitle("Active learning with Label Propagation.\nRows show 5 most "

"uncertain labels to learn with the next model.", y=1.15)
plt.subplots_adjust(left=0.2, bottom=0.03, right=0.9, top=0.9, wspace=0.2,
hspace=0.85)
plt.show()
5.25 Support Vector Machines
Examples concerning the sklearn.svm module.

5.25.1 Non-linear SVM
Perform binary classification using non-linear SVC with RBF kernel. The target to predict is a XOR of the inputs.
The color map illustrates the decision function learned by the SVC.
print(__doc__)
import numpy as np

np.random.seed(0)
X = np.random.randn(300, 2)
Y = np.logical_xor(X[:, 0] > 0, X[:, 1] > 0)
# fit the model

clf = svm.NuSVC()
clf.fit(X, Y)
# plot the decision function for each datapoint on the grid

5.25. Support Vector Machines 1335

plt.imshow(Z, interpolation='nearest',
extent=(xx.min(), xx.max(), yy.min(), yy.max()), aspect='auto',
origin='lower', cmap=plt.cm.PuOr_r)
contours = plt.contour(xx, yy, Z, levels=[0], linewidths=2,
linetypes='--')
plt.scatter(X[:, 0], X[:, 1], s=30, c=Y, cmap=plt.cm.Paired,
edgecolors='k')
plt.xticks(())
plt.yticks(())
plt.axis([-3, 3, -3, 3])
plt.show()
5.25.2 SVM: Maximum margin separating hyperplane
Plot the maximum margin separating hyperplane within a two-class separable dataset using a Support Vector Machine
classifier with linear kernel.
print(__doc__)

import numpy as np

X, y = make_blobs(n_samples=40, centers=2, random_state=6)
# fit the model, don't regularize for illustration purposes

clf = svm.SVC(kernel='linear', C=1000)
clf.fit(X, y)
plt.scatter(X[:, 0], X[:, 1], c=y, s=30, cmap=plt.cm.Paired)
# plot the decision function

ax = plt.gca()
xlim = ax.get_xlim()
ylim = ax.get_ylim()
# create grid to evaluate model

xx = np.linspace(xlim[0], xlim[1], 30)
yy = np.linspace(ylim[0], ylim[1], 30)
YY, XX = np.meshgrid(yy, xx)
xy = np.vstack([XX.ravel(), YY.ravel()]).T
Z = clf.decision_function(xy).reshape(XX.shape)
# plot decision boundary and margins

ax.contour(XX, YY, Z, colors='k', levels=[-1, 0, 1], alpha=0.5,
linestyles=['--', '-', '--'])
# plot support vectors
ax.scatter(clf.support_vectors_[:, 0], clf.support_vectors_[:, 1], s=100,
linewidth=1, facecolors='none', edgecolors='k')
plt.show()
5.25.3 SVM with custom kernel
Simple usage of Support Vector Machines to classify a sample. It will plot the decision surface and the support vectors.

print(__doc__)
import numpy as np

X = iris.data[:, :2] # we only take the first two features. We could
# avoid this ugly slicing by using a two-dim dataset
Y = iris.target
def my_kernel(X, Y):

"""
We create a custom kernel:
(2 0)
k(X, Y) = X ( ) Y.T
(0 1)
"""
M = np.array([[2, 0], [0, 1.0]])
return np.dot(np.dot(X, M), Y.T)

# we create an instance of SVM and fit out data.

clf = svm.SVC(kernel=my_kernel)
clf.fit(X, Y)

plt.pcolormesh(xx, yy, Z, cmap=plt.cm.Paired)

plt.scatter(X[:, 0], X[:, 1], c=Y, cmap=plt.cm.Paired, edgecolors='k')
plt.title('3-Class classification using Support Vector Machine with custom'
' kernel')
plt.axis('tight')
plt.show()
5.25.4 SVM: Weighted samples
Plot decision function of a weighted dataset, where the size of points is proportional to its weight.
The sample weighting rescales the C parameter, which means that the classifier puts more emphasis on getting these
points right. The effect might often be subtle. To emphasize the effect here, we particularly weight outliers, making
the deformation of the decision boundary very visible.

print(__doc__)
import numpy as np
def plot_decision_function(classifier, sample_weight, axis, title):

# plot the decision function
Z = classifier.decision_function(np.c_[xx.ravel(), yy.ravel()])
axis.contourf(xx, yy, Z, alpha=0.75, cmap=plt.cm.bone)
axis.scatter(X[:, 0], X[:, 1], c=y, s=100 * sample_weight, alpha=0.9,
cmap=plt.cm.bone, edgecolors='black')
axis.axis('off')
axis.set_title(title)
# we create 20 points
np.random.seed(0)
X = np.r_[np.random.randn(10, 2) + [1, 1], np.random.randn(10, 2)]
y = [1] * 10 + [-1] * 10
sample_weight_last_ten = abs(np.random.randn(len(X)))
sample_weight_constant = np.ones(len(X))
# and bigger weights to some outliers
sample_weight_last_ten[15:] *= 5
sample_weight_last_ten[9] *= 15
# for reference, first fit without sample weights
# fit the model

clf_weights = svm.SVC(gamma=1)
clf_weights.fit(X, y, sample_weight=sample_weight_last_ten)
clf_no_weights = svm.SVC(gamma=1)
clf_no_weights.fit(X, y)
fig, axes = plt.subplots(1, 2, figsize=(14, 6))

plot_decision_function(clf_no_weights, sample_weight_constant, axes[0],
"Constant weights")
plot_decision_function(clf_weights, sample_weight_last_ten, axes[1],
"Modified weights")
plt.show()

5.25.5 SVM: Separating hyperplane for unbalanced classes
Find the optimal separating hyperplane using an SVC for classes that are unbalanced.
We first find the separating plane with a plain SVC and then plot (dashed) the separating hyperplane with automatically
correction for unbalanced classes.
Note: This example will also work by replacing SVC(kernel="linear") with

SGDClassifier(loss="hinge"). Setting the loss parameter of the SGDClassifier equal to hinge will
yield behaviour such as that of a SVC with a linear kernel.
For example try instead of the SVC:
clf = SGDClassifier(n_iter=100, alpha=0.01)
print(__doc__)
import numpy as np
# we create two clusters of random points

n_samples_1 = 1000

n_samples_2 = 100
centers = [[0.0, 0.0], [2.0, 2.0]]
clusters_std = [1.5, 0.5]
X, y = make_blobs(n_samples=[n_samples_1, n_samples_2],
centers=centers,
cluster_std=clusters_std,
random_state=0, shuffle=False)
# fit the model and get the separating hyperplane

clf = svm.SVC(kernel='linear', C=1.0)
clf.fit(X, y)
# fit the model and get the separating hyperplane using weighted classes
wclf = svm.SVC(kernel='linear', class_weight={1: 10})
wclf.fit(X, y)
# plot the samples

plt.scatter(X[:, 0], X[:, 1], c=y, cmap=plt.cm.Paired, edgecolors='k')
# plot the decision functions for both classifiers

ax = plt.gca()
xlim = ax.get_xlim()
ylim = ax.get_ylim()
# create grid to evaluate model

xx = np.linspace(xlim[0], xlim[1], 30)
yy = np.linspace(ylim[0], ylim[1], 30)
YY, XX = np.meshgrid(yy, xx)
xy = np.vstack([XX.ravel(), YY.ravel()]).T

Z = clf.decision_function(xy).reshape(XX.shape)
# plot decision boundary and margins

a = ax.contour(XX, YY, Z, colors='k', levels=[0], alpha=0.5, linestyles=['-'])
# get the separating hyperplane for weighted classes

Z = wclf.decision_function(xy).reshape(XX.shape)
# plot decision boundary and margins for weighted classes

b = ax.contour(XX, YY, Z, colors='r', levels=[0], alpha=0.5, linestyles=['-'])
plt.legend([a.collections[0], b.collections[0]], ["non weighted", "weighted"],

loc="upper right")
plt.show()
5.25.6 SVM-Kernels
Three different types of SVM-Kernels are displayed below. The polynomial and RBF are especially useful when the
data-points are not linearly separable.


•
print(__doc__)

import numpy as np
# Our dataset and targets

X = np.c_[(.4, -.7),
(-1.5, -1),
(-1.4, -.9),
(-1.3, -1.2),
(-1.1, -.2),
(-1.2, -.4),
(-.5, 1.2),
(-1.5, 2.1),
(1, 1),
# --
(1.3, .8),
(1.2, .5),
(.2, -2),
(.5, -2.4),
(.2, -2.3),
(0, -2.7),
(1.3, 2.1)].T
Y = [0] * 8 + [1] * 8
# figure number
fignum = 1
# fit the model

for kernel in ('linear', 'poly', 'rbf'):
clf = svm.SVC(kernel=kernel, gamma=2)
clf.fit(X, Y)

plt.figure(fignum, figsize=(4, 3))
plt.clf()
plt.scatter(clf.support_vectors_[:, 0], clf.support_vectors_[:, 1], s=80,

facecolors='none', zorder=10, edgecolors='k')
plt.scatter(X[:, 0], X[:, 1], c=Y, zorder=10, cmap=plt.cm.Paired,
edgecolors='k')
plt.axis('tight')
x_min = -3
x_max = 3
y_min = -3
y_max = 3

Z = clf.decision_function(np.c_[XX.ravel(), YY.ravel()])

plt.pcolormesh(XX, YY, Z > 0, cmap=plt.cm.Paired)
plt.contour(XX, YY, Z, colors=['k', 'k', 'k'], linestyles=['--', '-', '--'],
levels=[-.5, 0, .5])
plt.xticks(())
plt.yticks(())
fignum = fignum + 1
plt.show()
5.25.7 SVM-Anova: SVM with univariate feature selection
This example shows how to perform univariate feature selection before running a SVC (support vector classifier) to
improve the classification scores. We use the iris dataset (4 features) and add 36 non-informative features. We can find
that our model achieves best performance when we select around 10% of features.

print(__doc__)
import numpy as np
from sklearn.feature_selection import SelectPercentile, chi2
# #############################################################################
X, y = load_iris(return_X_y=True)
# Add non-informative features
np.random.seed(0)
X = np.hstack((X, 2 * np.random.random((X.shape[0], 36))))
# #############################################################################
# Create a feature-selection transform, a scaler and an instance of SVM that we
# combine together to have an full-blown estimator
clf = Pipeline([('anova', SelectPercentile(chi2)),
('scaler', StandardScaler()),
('svc', SVC(gamma="auto"))])

# #############################################################################
# Plot the cross-validation score as a function of percentile of features
score_means = list()
score_stds = list()
percentiles = (1, 3, 6, 10, 15, 20, 30, 40, 60, 80, 100)
for percentile in percentiles:

clf.set_params(anova__percentile=percentile)
this_scores = cross_val_score(clf, X, y, cv=5)
score_means.append(this_scores.mean())
score_stds.append(this_scores.std())
plt.errorbar(percentiles, score_means, np.array(score_stds))

plt.title(
'Performance of the SVM-Anova varying the percentile of features selected')
plt.xticks(np.linspace(0, 100, 11, endpoint=True))
plt.xlabel('Percentile')
plt.ylabel('Accuracy Score')
plt.axis('tight')
plt.show()
5.25.8 SVM Margins Example
The plots below illustrate the effect the parameter C has on the separation line. A large value of C basically tells our
model that we do not have that much faith in our data’s distribution, and will only consider points close to line of
separation.
A small value of C includes more/all the observations, allowing the margins to be calculated using all the data in the
area.

•
print(__doc__)

import numpy as np

np.random.seed(0)
X = np.r_[np.random.randn(20, 2) - [2, 2], np.random.randn(20, 2) + [2, 2]]
Y = [0] * 20 + [1] * 20
# figure number
fignum = 1
# fit the model

for name, penalty in (('unreg', 1), ('reg', 0.05)):
clf = svm.SVC(kernel='linear', C=penalty)

clf.fit(X, Y)

w = clf.coef_[0]
a = -w[0] / w[1]
xx = np.linspace(-5, 5)
yy = a * xx - (clf.intercept_[0]) / w[1]
# plot the parallels to the separating hyperplane that pass through the
# support vectors (margin away from hyperplane in direction
# perpendicular to hyperplane). This is sqrt(1+a^2) away vertically in
# 2-d.
margin = 1 / np.sqrt(np.sum(clf.coef_ ** 2))
yy_down = yy - np.sqrt(1 + a ** 2) * margin
yy_up = yy + np.sqrt(1 + a ** 2) * margin

plt.clf()
plt.plot(xx, yy, 'k-')
plt.plot(xx, yy_down, 'k--')
plt.plot(xx, yy_up, 'k--')
plt.scatter(clf.support_vectors_[:, 0], clf.support_vectors_[:, 1], s=80,

facecolors='none', zorder=10, edgecolors='k')
plt.scatter(X[:, 0], X[:, 1], c=Y, zorder=10, cmap=plt.cm.Paired,
edgecolors='k')
plt.axis('tight')
x_min = -4.8
x_max = 4.2
y_min = -6
y_max = 6

Z = clf.predict(np.c_[XX.ravel(), YY.ravel()])

plt.pcolormesh(XX, YY, Z, cmap=plt.cm.Paired)
plt.xticks(())
plt.yticks(())
fignum = fignum + 1
plt.show()
5.25.9 One-class SVM with non-linear kernel (RBF)
An example using a one-class SVM for novelty detection.

One-class SVM is an unsupervised algorithm that learns a decision function for novelty detection: classifying new
data as similar or different to the training set.

print(__doc__)
import numpy as np

X_train = np.r_[X + 2, X - 2]
# Generate some regular novel observations
X_test = np.r_[X + 2, X - 2]
# fit the model

clf = svm.OneClassSVM(nu=0.1, kernel="rbf", gamma=0.1)
clf.fit(X_train)
y_pred_train = clf.predict(X_train)
n_error_train = y_pred_train[y_pred_train == -1].size
n_error_test = y_pred_test[y_pred_test == -1].size
n_error_outliers = y_pred_outliers[y_pred_outliers == 1].size

plt.title("Novelty Detection")
plt.contourf(xx, yy, Z, levels=np.linspace(Z.min(), 0, 7), cmap=plt.cm.PuBu)
a = plt.contour(xx, yy, Z, levels=[0], linewidths=2, colors='darkred')
plt.contourf(xx, yy, Z, levels=[0, Z.max()], colors='palevioletred')
s = 40
b1 = plt.scatter(X_train[:, 0], X_train[:, 1], c='white', s=s, edgecolors='k')
b2 = plt.scatter(X_test[:, 0], X_test[:, 1], c='blueviolet', s=s,
edgecolors='k')
c = plt.scatter(X_outliers[:, 0], X_outliers[:, 1], c='gold', s=s,
edgecolors='k')
plt.axis('tight')
plt.xlim((-5, 5))
plt.ylim((-5, 5))
plt.legend([a.collections[0], b1, b2, c],
["learned frontier", "training observations",
loc="upper left",
plt.xlabel(
"error train: %d/200 ; errors novel regular: %d/40 ; "
"errors novel abnormal: %d/40"
% (n_error_train, n_error_test, n_error_outliers))
plt.show()
5.25.10 Support Vector Regression (SVR) using linear and non-linear kernels
Toy example of 1D regression using linear, polynomial and RBF kernels.

print(__doc__)
import numpy as np
from sklearn.svm import SVR
# #############################################################################
X = np.sort(5 * np.random.rand(40, 1), axis=0)
# #############################################################################
y[::5] += 3 * (0.5 - np.random.rand(8))
# #############################################################################
svr_rbf = SVR(kernel='rbf', C=100, gamma=0.1, epsilon=.1)
svr_lin = SVR(kernel='linear', C=100, gamma='auto')
svr_poly = SVR(kernel='poly', C=100, gamma='auto', degree=3, epsilon=.1,
coef0=1)
y_rbf = svr_rbf.fit(X, y).predict(X)
y_lin = svr_lin.fit(X, y).predict(X)
y_poly = svr_poly.fit(X, y).predict(X)
# #############################################################################
# Look at the results
lw = 2

svrs = [svr_rbf, svr_lin, svr_poly]

kernel_label = ['RBF', 'Linear', 'Polynomial']
model_color = ['m', 'c', 'g']
fig, axes = plt.subplots(nrows=1, ncols=3, figsize=(15, 10), sharey=True)

for ix, svr in enumerate(svrs):
axes[ix].plot(X, svr.fit(X, y).predict(X), color=model_color[ix], lw=lw,
label='{} model'.format(kernel_label[ix]))
axes[ix].scatter(X[svr.support_], y[svr.support_], facecolor="none",
edgecolor=model_color[ix], s=50,
label='{} support vectors'.format(kernel_label[ix]))
axes[ix].scatter(X[np.setdiff1d(np.arange(len(X)), svr.support_)],
y[np.setdiff1d(np.arange(len(X)), svr.support_)],
facecolor="none", edgecolor="k", s=50,
label='other training data')
axes[ix].legend(loc='upper center', bbox_to_anchor=(0.5, 1.1),
ncol=1, fancybox=True, shadow=True)
fig.text(0.5, 0.04, 'data', ha='center', va='center')

fig.text(0.06, 0.5, 'target', ha='center', va='center', rotation='vertical')
fig.suptitle("Support Vector Regression", fontsize=14)
plt.show()
5.25.11 Plot different SVM classifiers in the iris dataset
Comparison of different linear SVM classifiers on a 2D projection of the iris dataset. We only consider the first 2
features of this dataset:
• Sepal length
• Sepal width
This example shows how to plot the decision surface for four SVM classifiers with different kernels.
The linear models LinearSVC() and SVC(kernel='linear') yield slightly different decision boundaries.
This can be a consequence of the following differences:
• LinearSVC minimizes the squared hinge loss while SVC minimizes the regular hinge loss.
• LinearSVC uses the One-vs-All (also known as One-vs-Rest) multiclass reduction while SVC uses the One-
vs-One multiclass reduction.
Both linear models have linear decision boundaries (intersecting hyperplanes) while the non-linear kernel models
(polynomial or Gaussian RBF) have more flexible non-linear decision boundaries with shapes that depend on the kind
of kernel and its parameters.
Note: while plotting the decision function of classifiers for toy 2D datasets can help get an intuitive understanding
of their respective expressive power, be aware that those intuitions don’t always generalize to more realistic high-
dimensional problems.

print(__doc__)
import numpy as np
def make_meshgrid(x, y, h=.02):

"""Create a mesh of points to plot in
Parameters
----------
x: data to base x-axis meshgrid on
y: data to base y-axis meshgrid on
h: stepsize for meshgrid, optional
Returns
-------
xx, yy : ndarray
"""
x_min, x_max = x.min() - 1, x.max() + 1
y_min, y_max = y.min() - 1, y.max() + 1
return xx, yy

def plot_contours(ax, clf, xx, yy, **params):

"""Plot the decision boundaries for a classifier.
Parameters
----------
ax: matplotlib axes object
clf: a classifier
xx: meshgrid ndarray
yy: meshgrid ndarray
params: dictionary of params to pass to contourf, optional
"""
out = ax.contourf(xx, yy, Z, **params)
return out

# Take the first two features. We could avoid this by using a two-dim dataset
y = iris.target
# we create an instance of SVM and fit out data. We do not scale our
# data since we want to plot the support vectors
C = 1.0 # SVM regularization parameter
models = (svm.SVC(kernel='linear', C=C),
svm.LinearSVC(C=C),
svm.SVC(kernel='rbf', gamma=0.7, C=C),
svm.SVC(kernel='poly', degree=3, C=C))
models = (clf.fit(X, y) for clf in models)

titles = ('SVC with linear kernel',
'LinearSVC (linear kernel)',
'SVC with RBF kernel',
'SVC with polynomial (degree 3) kernel')
# Set-up 2x2 grid for plotting.

fig, sub = plt.subplots(2, 2)
plt.subplots_adjust(wspace=0.4, hspace=0.4)
X0, X1 = X[:, 0], X[:, 1]

xx, yy = make_meshgrid(X0, X1)
for clf, title, ax in zip(models, titles, sub.flatten()):

plot_contours(ax, clf, xx, yy,
cmap=plt.cm.coolwarm, alpha=0.8)
ax.scatter(X0, X1, c=y, cmap=plt.cm.coolwarm, s=20, edgecolors='k')
ax.set_xlabel('Sepal length')
ax.set_ylabel('Sepal width')
ax.set_xticks(())
ax.set_yticks(())
ax.set_title(title)

plt.show()
5.25.12 Scaling the regularization parameter for SVCs
The following example illustrates the effect of scaling the regularization parameter when using Support Vector Ma-
chines for classification. For SVC classification, we are interested in a risk minimization for the equation:
∑︁
𝐶 ℒ(𝑓 (𝑥𝑖 ), 𝑦𝑖 ) + Ω(𝑤)
𝑖=1,𝑛
where
• 𝐶 is used to set the amount of regularization
• ℒ is a loss function of our samples and our model parameters.
• Ω is a penalty function of our model parameters
If we consider the loss function to be the individual error per sample, then the data-fit term, or the sum of the error for
each sample, will increase as we add more samples. The penalization term, however, will not increase.
When using, for example, cross validation, to set the amount of regularization with C, there will be a different amount
of samples between the main problem and the smaller problems within the folds of the cross validation.
Since our loss function is dependent on the amount of samples, the latter will influence the selected value of C. The
question that arises is How do we optimally adjust C to account for the different amount of training samples?
The figures below are used to illustrate the effect of scaling our C to compensate for the change in the number of
samples, in the case of using an l1 penalty, as well as the l2 penalty.
l1-penalty case
In the l1 case, theory says that prediction consistency (i.e. that under given hypothesis, the estimator learned predicts
as well as a model knowing the true distribution) is not possible because of the bias of the l1. It does say, however,
that model consistency, in terms of finding the right set of non-zero parameters as well as their signs, can be achieved
by scaling C1.
l2-penalty case
The theory says that in order to achieve prediction consistency, the penalty parameter should be kept constant as the
number of samples grow.
Simulations
The two figures below plot the values of C on the x-axis and the corresponding cross-validation scores on the y-axis,
for several different fractions of a generated data-set.
In the l1 penalty case, the cross-validation-error correlates best with the test-error, when scaling our C with the number
of samples, n, which can be seen in the first figure.
For the l2 penalty case, the best result comes from the case where C is not scaled.

Note:
Two separate datasets are used for the two different plots. The reason behind this is the l1 case works better on
sparse data, while l2 is better suited to the non-sparse case.

•
print(__doc__)
# Author: Andreas Mueller <amueller@ais.uni-bonn.de>

# Jaques Grobler <jaques.grobler@inria.fr>
import numpy as np

from sklearn.model_selection import ShuffleSplit


rnd = check_random_state(1)
# set up dataset
n_samples = 100
n_features = 300
# l1 data (only 5 informative features)

X_1, y_1 = datasets.make_classification(n_samples=n_samples,
n_features=n_features, n_informative=5,
random_state=1)
# l2 data: non sparse, but less features

y_2 = np.sign(.5 - rnd.rand(n_samples))
X_2 = rnd.randn(n_samples, n_features // 5) + y_2[:, np.newaxis]
X_2 += 5 * rnd.randn(n_samples, n_features // 5)
clf_sets = [(LinearSVC(penalty='l1', loss='squared_hinge', dual=False,

tol=1e-3),
np.logspace(-2.3, -1.3, 10), X_1, y_1),
(LinearSVC(penalty='l2', loss='squared_hinge', dual=True,
tol=1e-4),
np.logspace(-4.5, -2, 10), X_2, y_2)]
colors = ['navy', 'cyan', 'darkorange']

lw = 2
for clf, cs, X, y in clf_sets:

# set up the plot for each regressor
fig, axes = plt.subplots(nrows=2, sharey=True, figsize=(9, 10))
for k, train_size in enumerate(np.linspace(0.3, 0.7, 3)[::-1]):

param_grid = dict(C=cs)
# To get nice curve, we need a large number of iterations to
# reduce the variance
grid = GridSearchCV(clf, refit=False, param_grid=param_grid,
cv=ShuffleSplit(train_size=train_size,
test_size=.3,
n_splits=250, random_state=1))
grid.fit(X, y)
scores = grid.cv_results_['mean_test_score']
scales = [(1, 'No scaling'),

((n_samples * train_size), '1/n_samples'),
]
for ax, (scaler, name) in zip(axes, scales):

ax.set_xlabel('C')
ax.set_ylabel('CV Score')
grid_cs = cs * float(scaler) # scale the C's
ax.semilogx(grid_cs, scores, label="fraction %.2f" %
train_size, color=colors[k], lw=lw)
ax.set_title('scaling=%s, penalty=%s, loss=%s' %
(name, clf.penalty, clf.loss))

plt.show()
5.25.13 RBF SVM parameters
This example illustrates the effect of the parameters gamma and C of the Radial Basis Function (RBF) kernel SVM.
Intuitively, the gamma parameter defines how far the influence of a single training example reaches, with low values
meaning ‘far’ and high values meaning ‘close’. The gamma parameters can be seen as the inverse of the radius of
influence of samples selected by the model as support vectors.
The C parameter trades off correct classification of training examples against maximization of the decision function’s
margin. For larger values of C, a smaller margin will be accepted if the decision function is better at classifying all
training points correctly. A lower C will encourage a larger margin, therefore a simpler decision function, at the cost
of training accuracy. In other words‘‘C‘‘ behaves as a regularization parameter in the SVM.
The first plot is a visualization of the decision function for a variety of parameter values on a simplified classification
problem involving only 2 input features and 2 possible target classes (binary classification). Note that this kind of plot
is not possible to do for problems with more features or target classes.
The second plot is a heatmap of the classifier’s cross-validation accuracy as a function of C and gamma. For this
example we explore a relatively large grid for illustration purposes. In practice, a logarithmic grid from 10−3 to 103
is usually sufficient. If the best parameters lie on the boundaries of the grid, it can be extended in that direction in a
subsequent search.
Note that the heat map plot has a special colorbar with a midpoint value close to the score values of the best performing
models so as to make it easy to tell them apart in the blink of an eye.
The behavior of the model is very sensitive to the gamma parameter. If gamma is too large, the radius of the area of
influence of the support vectors only includes the support vector itself and no amount of regularization with C will be
able to prevent overfitting.
When gamma is very small, the model is too constrained and cannot capture the complexity or “shape” of the data.
The region of influence of any selected support vector would include the whole training set. The resulting model will
behave similarly to a linear model with a set of hyperplanes that separate the centers of high density of any pair of two
classes.
For intermediate values, we can see on the second plot that good models can be found on a diagonal of C and gamma.
Smooth models (lower gamma values) can be made more complex by increasing the importance of classifying each
point correctly (larger C values) hence the diagonal of good performing models.
Finally one can also observe that for some intermediate values of gamma we get equally performing models when C
becomes very large: it is not necessary to regularize by enforcing a larger margin. The radius of the RBF kernel alone
acts as a good structural regularizer. In practice though it might still be interesting to simplify the decision function
with a lower value of C so as to favor models that use less memory and that are faster to predict.
We should also note that small differences in scores results from the random splits of the cross-validation procedure.
Those spurious variations can be smoothed out by increasing the number of CV iterations n_splits at the expense
of compute time. Increasing the value number of C_range and gamma_range steps will increase the resolution of
the hyper-parameter heat map.


•
Out:
The best parameters are {'C': 1.0, 'gamma': 0.1} with a score of 0.97
print(__doc__)
import numpy as np
from matplotlib.colors import Normalize

from sklearn.model_selection import StratifiedShuffleSplit
# Utility function to move the midpoint of a colormap to be around

# the values of interest.
class MidpointNormalize(Normalize):

def __init__(self, vmin=None, vmax=None, midpoint=None, clip=False):

self.midpoint = midpoint
Normalize.__init__(self, vmin, vmax, clip)
def __call__(self, value, clip=None):

x, y = [self.vmin, self.midpoint, self.vmax], [0, 0.5, 1]
return np.ma.masked_array(np.interp(value, x, y))
# #############################################################################
# Load and prepare data set
#
# dataset for grid search
iris = load_iris()
X = iris.data
y = iris.target
# Dataset for decision function visualization: we only keep the first two
# features in X and sub-sample the dataset to keep only 2 classes and
# make it a binary classification problem.
X_2d = X[:, :2]

X_2d = X_2d[y > 0]
y_2d = y[y > 0]
y_2d -= 1
# It is usually a good idea to scale the data for SVM training.

# We are cheating a bit in this example in scaling all of the data,
# instead of fitting the transformation on the training set and
# just applying it on the test set.
X = scaler.fit_transform(X)
X_2d = scaler.fit_transform(X_2d)
# #############################################################################
# Train classifiers
#
# For an initial search, a logarithmic grid with basis
# 10 is often helpful. Using a basis of 2, a finer
# tuning can be achieved but at a much higher cost.
C_range = np.logspace(-2, 10, 13)

gamma_range = np.logspace(-9, 3, 13)
param_grid = dict(gamma=gamma_range, C=C_range)
cv = StratifiedShuffleSplit(n_splits=5, test_size=0.2, random_state=42)
grid = GridSearchCV(SVC(), param_grid=param_grid, cv=cv)
grid.fit(X, y)
print("The best parameters are %s with a score of %0.2f"

% (grid.best_params_, grid.best_score_))
# Now we need to fit a classifier for all parameters in the 2d version

# (we use a smaller set of parameters here because it takes a while to train)
C_2d_range = [1e-2, 1, 1e2]

gamma_2d_range = [1e-1, 1, 1e1]
classifiers = []

for C in C_2d_range:
for gamma in gamma_2d_range:
clf = SVC(C=C, gamma=gamma)
clf.fit(X_2d, y_2d)
classifiers.append((C, gamma, clf))
# #############################################################################
# Visualization
#
# draw visualization of parameter effects
for (k, (C, gamma, clf)) in enumerate(classifiers):
# evaluate decision function in a grid
# visualize decision function for these parameters

plt.subplot(len(C_2d_range), len(gamma_2d_range), k + 1)
plt.title("gamma=10^%d, C=10^%d" % (np.log10(gamma), np.log10(C)),
size='medium')
# visualize parameter's effect on decision function

plt.pcolormesh(xx, yy, -Z, cmap=plt.cm.RdBu)
plt.scatter(X_2d[:, 0], X_2d[:, 1], c=y_2d, cmap=plt.cm.RdBu_r,
edgecolors='k')
plt.xticks(())
plt.yticks(())
plt.axis('tight')
scores = grid.cv_results_['mean_test_score'].reshape(len(C_range),
len(gamma_range))
# Draw heatmap of the validation accuracy as a function of gamma and C

#
# The score are encoded as colors with the hot colormap which varies from dark
# red to bright yellow. As the most interesting scores are all located in the
# 0.92 to 0.97 range we use a custom normalizer to set the mid-point to 0.92 so
# as to make it easier to visualize the small variations of score values in the
# interesting range while not brutally collapsing all the low score values to
# the same color.
plt.subplots_adjust(left=.2, right=0.95, bottom=0.15, top=0.95)
plt.imshow(scores, interpolation='nearest', cmap=plt.cm.hot,
norm=MidpointNormalize(vmin=0.2, midpoint=0.92))
plt.xlabel('gamma')
plt.ylabel('C')
plt.colorbar()
plt.xticks(np.arange(len(gamma_range)), gamma_range, rotation=45)
plt.yticks(np.arange(len(C_range)), C_range)
plt.title('Validation accuracy')
plt.show()

5.26 Working with text documents
Examples concerning the sklearn.feature_extraction.text module.
5.26.1 FeatureHasher and DictVectorizer Comparison
Compares FeatureHasher and DictVectorizer by using both to vectorize text documents.

The example demonstrates syntax and speed only; it doesn’t actually do anything useful with the extracted vectors.
See the example scripts {document_classification_20newsgroups,clustering}.py for actual learning on text documents.
A discrepancy between the number of terms reported for DictVectorizer and for FeatureHasher is to be expected due
to hash collisions.
Out:
Usage: /home/circleci/project/examples/text/plot_hashing_vs_dict_vectorizer.py [n_

˓→features_for_hashing]
The default number of features is 2**18.
Loading 20 newsgroups training data

3803 documents - 6.245MB
DictVectorizer
done in 1.005914s at 6.208MB/s
Found 47928 unique terms
FeatureHasher on frequency dicts

FeatureHasher on raw tokens

# Author: Lars Buitinck


import re
import sys
import numpy as np

from sklearn.feature_extraction import DictVectorizer, FeatureHasher
5.26. Working with text documents 1365

def n_nonzero_columns(X):
"""Returns the number of non-zero columns in a CSR matrix X."""
return len(np.unique(X.nonzero()[1]))
def tokens(doc):
"""Extract tokens from doc.
This uses a simple regex to break strings into tokens. For a more
principled approach, see CountVectorizer or TfidfVectorizer.
"""
return (tok.lower() for tok in re.findall(r"\w+", doc))
def token_freqs(doc):
"""Extract a dict mapping tokens from doc to their frequencies."""
freq = defaultdict(int)
for tok in tokens(doc):
freq[tok] += 1
return freq
categories = [
'alt.atheism',
'comp.graphics',
'comp.sys.ibm.pc.hardware',
'misc.forsale',
'rec.autos',
'sci.space',
]
# Uncomment the following line to use a larger set (11k+ documents)
# categories = None
print(__doc__)
print("Usage: %s [n_features_for_hashing]" % sys.argv[0])
print(" The default number of features is 2**18.")
print()
try:
n_features = int(sys.argv[1])
except IndexError:
n_features = 2 ** 18
except ValueError:
print("not a valid number of features: %r" % sys.argv[1])
sys.exit(1)
print("Loading 20 newsgroups training data")

raw_data = fetch_20newsgroups(subset='train', categories=categories).data
data_size_mb = sum(len(s.encode('utf-8')) for s in raw_data) / 1e6
print("%d documents - %0.3fMB" % (len(raw_data), data_size_mb))
print()
print("DictVectorizer")
t0 = time()

vectorizer = DictVectorizer()
vectorizer.fit_transform(token_freqs(d) for d in raw_data)
duration = time() - t0
print("done in %fs at %0.3fMB/s" % (duration, data_size_mb / duration))
print("Found %d unique terms" % len(vectorizer.get_feature_names()))
print()
print("FeatureHasher on frequency dicts")

t0 = time()
hasher = FeatureHasher(n_features=n_features)
X = hasher.transform(token_freqs(d) for d in raw_data)
print("Found %d unique terms" % n_nonzero_columns(X))
print()
print("FeatureHasher on raw tokens")

t0 = time()
hasher = FeatureHasher(n_features=n_features, input_type="string")
X = hasher.transform(tokens(d) for d in raw_data)
print("Found %d unique terms" % n_nonzero_columns(X))
5.26.2 Clustering text documents using k-means
This is an example showing how the scikit-learn can be used to cluster documents by topics using a bag-of-words
approach. This example uses a scipy.sparse matrix to store the features instead of standard numpy arrays.
Two feature extraction methods can be used in this example:
• TfidfVectorizer uses a in-memory vocabulary (a python dict) to map the most frequent words to features indices
and hence compute a word occurrence frequency (sparse) matrix. The word frequencies are then reweighted
using the Inverse Document Frequency (IDF) vector collected feature-wise over the corpus.
• HashingVectorizer hashes word occurrences to a fixed dimensional space, possibly with collisions. The word
count vectors are then normalized to each have l2-norm equal to one (projected to the euclidean unit-ball) which
seems to be important for k-means to work in high dimensional space.
HashingVectorizer does not provide IDF weighting as this is a stateless model (the fit method does nothing).
When IDF weighting is needed it can be added by pipelining its output to a TfidfTransformer instance.
Two algorithms are demoed: ordinary k-means and its more scalable cousin minibatch k-means.
Additionally, latent semantic analysis can also be used to reduce dimensionality and discover latent patterns in the
data.
It can be noted that k-means (and minibatch k-means) are very sensitive to feature scaling and that in this case the IDF
weighting helps improve the quality of the clustering by quite a lot as measured against the “ground truth” provided
by the class label assignments of the 20 newsgroups dataset.
This improvement is not visible in the Silhouette Coefficient which is small for both as this measure seem to suffer
from the phenomenon called “Concentration of Measure” or “Curse of Dimensionality” for high dimensional datasets

such as text data. Other measures such as V-measure and Adjusted Rand Index are information theoretic based eval-
uation scores: as they are only based on cluster assignments rather than distances, hence not affected by the curse of
dimensionality.
Note: as k-means is optimizing a non-convex objective function, it will likely end up in a local optimum. Several runs
with independent random init might be necessary to get a good convergence.
Out:
Usage: plot_document_clustering.py [options]
Options:
-h, --help show this help message and exit
--lsa=N_COMPONENTS Preprocess documents with latent semantic analysis.
--no-minibatch Use ordinary k-means algorithm (in batch mode).
--no-idf Disable Inverse Document Frequency feature weighting.
--use-hashing Use a hashing feature vectorizer
--n-features=N_FEATURES
Maximum number of features (dimensions) to extract
from text.
--verbose Print progress reports inside k-means algorithm.
['alt.atheism', 'talk.religion.misc', 'comp.graphics', 'sci.space']
3387 documents
4 categories
Extracting features from the training dataset using a sparse vectorizer

done in 0.697900s
n_samples: 3387, n_features: 10000
Clustering sparse data with MiniBatchKMeans(batch_size=1000, compute_labels=True,

˓→init='k-means++',
init_size=1000, max_iter=100, max_no_improvement=10, n_clusters=4,

n_init=1, random_state=None, reassignment_ratio=0.01, tol=0.0,
verbose=False)
done in 0.053s
Homogeneity: 0.576
Completeness: 0.594
V-measure: 0.585
Adjusted Rand-Index: 0.513
Top terms per cluster:

Cluster 0: graphics file thanks image university 3d ac files format program
Cluster 1: com access digex pat nasa gov posting nntp host net
Cluster 2: god com people sandvik keith don jesus article say think
Cluster 3: space henry alaska toronto nasa moon shuttle zoo launch spencer

# Lars Buitinck


from sklearn.feature_extraction.text import TfidfTransformer
from sklearn.preprocessing import Normalizer
from sklearn.cluster import KMeans, MiniBatchKMeans
import logging
import sys
import numpy as np

# parse commandline arguments

op = OptionParser()
op.add_option("--lsa",
dest="n_components", type="int",
help="Preprocess documents with latent semantic analysis.")
op.add_option("--no-minibatch",
action="store_false", dest="minibatch", default=True,
help="Use ordinary k-means algorithm (in batch mode).")
op.add_option("--no-idf",
action="store_false", dest="use_idf", default=True,
help="Disable Inverse Document Frequency feature weighting.")
op.add_option("--use-hashing",
action="store_true", default=False,
help="Use a hashing feature vectorizer")
op.add_option("--n-features", type=int, default=10000,
help="Maximum number of features (dimensions)"
" to extract from text.")
op.add_option("--verbose",
action="store_true", dest="verbose", default=False,
help="Print progress reports inside k-means algorithm.")
print(__doc__)
op.print_help()
def is_interactive():
return not hasattr(sys.modules['__main__'], '__file__')
# work-around for Jupyter notebook and IPython console

argv = [] if is_interactive() else sys.argv[1:]
(opts, args) = op.parse_args(argv)
if len(args) > 0:
op.error("this script takes no arguments.")

sys.exit(1)
# #############################################################################
categories = [
'alt.atheism',
'comp.graphics',
'sci.space',
]
# Uncomment the following to do the analysis on all the categories
# categories = None

print(categories)
dataset = fetch_20newsgroups(subset='all', categories=categories,

shuffle=True, random_state=42)
print("%d documents" % len(dataset.data))

print("%d categories" % len(dataset.target_names))
print()
labels = dataset.target
true_k = np.unique(labels).shape[0]
print("Extracting features from the training dataset "

"using a sparse vectorizer")
t0 = time()
if opts.use_hashing:
if opts.use_idf:
# Perform an IDF normalization on the output of HashingVectorizer
hasher = HashingVectorizer(n_features=opts.n_features,
stop_words='english', alternate_sign=False,
norm=None, binary=False)
vectorizer = make_pipeline(hasher, TfidfTransformer())
else:
vectorizer = HashingVectorizer(n_features=opts.n_features,
stop_words='english',
alternate_sign=False, norm='l2',
binary=False)
else:
vectorizer = TfidfVectorizer(max_df=0.5, max_features=opts.n_features,
min_df=2, stop_words='english',
use_idf=opts.use_idf)
X = vectorizer.fit_transform(dataset.data)
print("done in %fs" % (time() - t0))

print("n_samples: %d, n_features: %d" % X.shape)
print()
if opts.n_components:
print("Performing dimensionality reduction using LSA")
t0 = time()
# Vectorizer results are normalized, which makes KMeans behave as
# spherical k-means for better results. Since LSA/SVD results are
# not normalized, we have to redo the normalization.

svd = TruncatedSVD(opts.n_components)
normalizer = Normalizer(copy=False)
lsa = make_pipeline(svd, normalizer)
X = lsa.fit_transform(X)
explained_variance = svd.explained_variance_ratio_.sum()
print("Explained variance of the SVD step: {}%".format(
int(explained_variance * 100)))
print()
# #############################################################################
# Do the actual clustering
if opts.minibatch:
km = MiniBatchKMeans(n_clusters=true_k, init='k-means++', n_init=1,
init_size=1000, batch_size=1000, verbose=opts.verbose)
else:
km = KMeans(n_clusters=true_k, init='k-means++', max_iter=100, n_init=1,
verbose=opts.verbose)
print("Clustering sparse data with %s" % km)

t0 = time()
km.fit(X)
print()
print("Homogeneity: %0.3f" % metrics.homogeneity_score(labels, km.labels_))

print("Completeness: %0.3f" % metrics.completeness_score(labels, km.labels_))
print("V-measure: %0.3f" % metrics.v_measure_score(labels, km.labels_))
print("Adjusted Rand-Index: %.3f"
% metrics.adjusted_rand_score(labels, km.labels_))
% metrics.silhouette_score(X, km.labels_, sample_size=1000))
print()
if not opts.use_hashing:
print("Top terms per cluster:")
if opts.n_components:
original_space_centroids = svd.inverse_transform(km.cluster_centers_)
order_centroids = original_space_centroids.argsort()[:, ::-1]
else:
order_centroids = km.cluster_centers_.argsort()[:, ::-1]
terms = vectorizer.get_feature_names()
for i in range(true_k):
print("Cluster %d:" % i, end='')
for ind in order_centroids[i, :10]:
print(' %s' % terms[ind], end='')
print()

5.26.3 Classification of text documents using sparse features
This is an example showing how scikit-learn can be used to classify documents by topics using a bag-of-words ap-
proach. This example uses a scipy.sparse matrix to store the features and demonstrates various classifiers that can
efficiently handle sparse matrices.
The dataset used in this example is the 20 newsgroups dataset. It will be automatically downloaded, then cached.
The bar plot indicates the accuracy, training time (normalized) and test time (normalized) of each classifier.
Out:
Usage: plot_document_classification_20newsgroups.py [options]
Options:
-h, --help show this help message and exit
--report Print a detailed classification report.
--chi2_select=SELECT_CHI2
Select some number of features using a chi-squared
test
--confusion_matrix Print the confusion matrix.
--top10 Print ten most discriminative terms per class for
every classifier.

--all_categories Whether to use all categories or not.

--use_hashing Use a hashing vectorizer.
--n_features=N_FEATURES
n_features when using the hashing vectorizer.
--filtered Remove newsgroup information that is easily overfit:
headers, signatures, and quoting.

['alt.atheism', 'talk.religion.misc', 'comp.graphics', 'sci.space']
data loaded
2034 documents - 3.980MB (training set)
1353 documents - 2.867MB (test set)
4 categories
Extracting features from the training data using a sparse vectorizer

Extracting features from the test data using the same vectorizer
================================================================================
Ridge Classifier
________________________________________________________________________________
Training:
RidgeClassifier(alpha=1.0, class_weight=None, copy_X=True, fit_intercept=True,
max_iter=None, normalize=False, random_state=None, solver='sag',
tol=0.01)
train time: 0.211s
test time: 0.002s
accuracy: 0.897
dimensionality: 33809
density: 1.000000
================================================================================
Perceptron
________________________________________________________________________________
Training:
Perceptron(alpha=0.0001, class_weight=None, early_stopping=False, eta0=1.0,
fit_intercept=True, max_iter=50, n_iter=None, n_iter_no_change=5,
n_jobs=None, penalty=None, random_state=0, shuffle=True, tol=0.001,
train time: 0.084s
test time: 0.002s
accuracy: 0.888
density: 0.240114
================================================================================
Passive-Aggressive
________________________________________________________________________________
Training:
PassiveAggressiveClassifier(C=1.0, average=False, class_weight=None,
early_stopping=False, fit_intercept=True, loss='hinge',
max_iter=50, n_iter=None, n_iter_no_change=5, n_jobs=None,

random_state=None, shuffle=True, tol=0.001,

train time: 0.028s
test time: 0.002s
accuracy: 0.905
density: 0.690615
================================================================================
kNN
________________________________________________________________________________
Training:
KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',
metric_params=None, n_jobs=None, n_neighbors=10, p=2,
weights='uniform')
train time: 0.042s
test time: 0.248s
accuracy: 0.858
================================================================================
Random forest
________________________________________________________________________________
Training:
RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',
max_depth=None, max_features='auto', max_leaf_nodes=None,
oob_score=False, random_state=None, verbose=0,
warm_start=False)
train time: 1.417s
test time: 0.075s
accuracy: 0.840
================================================================================
L2 penalty
________________________________________________________________________________
Training:
LinearSVC(C=1.0, class_weight=None, dual=False, fit_intercept=True,
verbose=0)
train time: 0.107s
test time: 0.002s
accuracy: 0.900
density: 1.000000
________________________________________________________________________________
Training:

train time: 0.196s

test time: 0.002s
accuracy: 0.900
density: 0.667381
================================================================================
L1 penalty
________________________________________________________________________________
Training:
LinearSVC(C=1.0, class_weight=None, dual=False, fit_intercept=True,
verbose=0)
train time: 0.268s
test time: 0.002s
accuracy: 0.874
density: 0.005553
________________________________________________________________________________
Training:
train time: 0.656s
test time: 0.002s
accuracy: 0.888
density: 0.020350
================================================================================
Elastic-Net penalty
________________________________________________________________________________
Training:
n_iter=None, n_iter_no_change=5, n_jobs=None, penalty='elasticnet',
train time: 0.881s
test time: 0.002s
accuracy: 0.902
density: 0.186718
================================================================================
NearestCentroid (aka Rocchio classifier)
________________________________________________________________________________
Training:

train time: 0.007s
test time: 0.002s
accuracy: 0.855
================================================================================
Naive Bayes
________________________________________________________________________________
Training:
train time: 0.003s
test time: 0.001s
accuracy: 0.899
density: 1.000000
________________________________________________________________________________
Training:
BernoulliNB(alpha=0.01, binarize=0.0, class_prior=None, fit_prior=True)
train time: 0.005s
test time: 0.003s
accuracy: 0.884
density: 1.000000
________________________________________________________________________________
Training:
ComplementNB(alpha=0.1, class_prior=None, fit_prior=True, norm=False)
train time: 0.004s
test time: 0.001s
accuracy: 0.911
density: 1.000000
================================================================================
LinearSVC with L1-based feature selection
________________________________________________________________________________
Training:
steps=[('feature_selection', SelectFromModel(estimator=LinearSVC(C=1.0, class_
˓→weight=None, dual=False, fit_intercept=True,

verbose=0),
max_features=None, no...ax_iter=1000,
verbose=0))])
train time: 0.219s
test time: 0.003s
accuracy: 0.881


# Lars Buitinck
import logging
import numpy as np
import sys

from sklearn.feature_selection import SelectFromModel
from sklearn.feature_selection import SelectKBest, chi2
from sklearn.linear_model import RidgeClassifier
from sklearn.linear_model import Perceptron
from sklearn.naive_bayes import BernoulliNB, ComplementNB, MultinomialNB
from sklearn.neighbors import NearestCentroid
from sklearn.utils.extmath import density

# parse commandline arguments

op = OptionParser()
op.add_option("--report",
action="store_true", dest="print_report",
help="Print a detailed classification report.")
op.add_option("--chi2_select",
action="store", type="int", dest="select_chi2",
help="Select some number of features using a chi-squared test")
op.add_option("--confusion_matrix",
action="store_true", dest="print_cm",
help="Print the confusion matrix.")
op.add_option("--top10",
action="store_true", dest="print_top10",
help="Print ten most discriminative terms per class"
" for every classifier.")
op.add_option("--all_categories",
action="store_true", dest="all_categories",
help="Whether to use all categories or not.")
op.add_option("--use_hashing",

action="store_true",
help="Use a hashing vectorizer.")
op.add_option("--n_features",
action="store", type=int, default=2 ** 16,
help="n_features when using the hashing vectorizer.")
op.add_option("--filtered",
action="store_true",
help="Remove newsgroup information that is easily overfit: "
"headers, signatures, and quoting.")
def is_interactive():
return not hasattr(sys.modules['__main__'], '__file__')
# work-around for Jupyter notebook and IPython console

argv = [] if is_interactive() else sys.argv[1:]
(opts, args) = op.parse_args(argv)
if len(args) > 0:
op.error("this script takes no arguments.")
sys.exit(1)
print(__doc__)
op.print_help()
print()
# #############################################################################
if opts.all_categories:
categories = None
else:
categories = [
'alt.atheism',
'comp.graphics',
'sci.space',
]
if opts.filtered:
remove = ('headers', 'footers', 'quotes')
else:
remove = ()

print(categories if categories else "all")
data_train = fetch_20newsgroups(subset='train', categories=categories,

shuffle=True, random_state=42,
remove=remove)
data_test = fetch_20newsgroups(subset='test', categories=categories,

shuffle=True, random_state=42,
remove=remove)
print('data loaded')
# order of labels in `target_names` can be different from `categories`

target_names = data_train.target_names

def size_mb(docs):
return sum(len(s.encode('utf-8')) for s in docs) / 1e6
data_train_size_mb = size_mb(data_train.data)
data_test_size_mb = size_mb(data_test.data)
print("%d documents - %0.3fMB (training set)" % (

len(data_train.data), data_train_size_mb))
print("%d documents - %0.3fMB (test set)" % (
len(data_test.data), data_test_size_mb))
print("%d categories" % len(target_names))
print()
# split a training set and a test set

y_train, y_test = data_train.target, data_test.target
print("Extracting features from the training data using a sparse vectorizer")

t0 = time()
vectorizer = HashingVectorizer(stop_words='english', alternate_sign=False,
n_features=opts.n_features)
X_train = vectorizer.transform(data_train.data)
else:
vectorizer = TfidfVectorizer(sublinear_tf=True, max_df=0.5,
X_train = vectorizer.fit_transform(data_train.data)
print("done in %fs at %0.3fMB/s" % (duration, data_train_size_mb / duration))
print("n_samples: %d, n_features: %d" % X_train.shape)
print()
print("Extracting features from the test data using the same vectorizer")
t0 = time()
X_test = vectorizer.transform(data_test.data)
print("done in %fs at %0.3fMB/s" % (duration, data_test_size_mb / duration))
print("n_samples: %d, n_features: %d" % X_test.shape)
print()
# mapping from integer feature name to original token string

feature_names = None
else:
feature_names = vectorizer.get_feature_names()
if opts.select_chi2:
print("Extracting %d best features by a chi-squared test" %
opts.select_chi2)
t0 = time()
ch2 = SelectKBest(chi2, k=opts.select_chi2)
X_train = ch2.fit_transform(X_train, y_train)
X_test = ch2.transform(X_test)
if feature_names:
# keep selected feature names
feature_names = [feature_names[i] for i

in ch2.get_support(indices=True)]
print()
if feature_names:
feature_names = np.asarray(feature_names)
def trim(s):
"""Trim string to fit on terminal (assuming 80-column display)"""
return s if len(s) <= 80 else s[:77] + "..."
# #############################################################################
# Benchmark classifiers
def benchmark(clf):
print('_' * 80)
print("Training: ")
print(clf)
t0 = time()
train_time = time() - t0
print("train time: %0.3fs" % train_time)
t0 = time()
pred = clf.predict(X_test)
test_time = time() - t0
print("test time: %0.3fs" % test_time)
score = metrics.accuracy_score(y_test, pred)

print("accuracy: %0.3f" % score)
if hasattr(clf, 'coef_'):
print("dimensionality: %d" % clf.coef_.shape[1])
print("density: %f" % density(clf.coef_))
if opts.print_top10 and feature_names is not None:

print("top 10 keywords per class:")
for i, label in enumerate(target_names):
top10 = np.argsort(clf.coef_[i])[-10:]
print(trim("%s: %s" % (label, " ".join(feature_names[top10]))))
print()
if opts.print_report:
print("classification report:")
print(metrics.classification_report(y_test, pred,
target_names=target_names))
if opts.print_cm:
print("confusion matrix:")
print(metrics.confusion_matrix(y_test, pred))
print()
clf_descr = str(clf).split('(')[0]
return clf_descr, score, train_time, test_time
results = []

for clf, name in (

(RidgeClassifier(tol=1e-2, solver="sag"), "Ridge Classifier"),
(Perceptron(max_iter=50, tol=1e-3), "Perceptron"),
(PassiveAggressiveClassifier(max_iter=50, tol=1e-3),
"Passive-Aggressive"),
(KNeighborsClassifier(n_neighbors=10), "kNN"),
(RandomForestClassifier(n_estimators=100), "Random forest")):
print('=' * 80)
print(name)
results.append(benchmark(clf))
for penalty in ["l2", "l1"]:

print('=' * 80)
print("%s penalty" % penalty.upper())
# Train Liblinear model
results.append(benchmark(LinearSVC(penalty=penalty, dual=False,
tol=1e-3)))
# Train SGD model

results.append(benchmark(SGDClassifier(alpha=.0001, max_iter=50,
penalty=penalty)))
# Train SGD with Elastic Net penalty

print('=' * 80)
print("Elastic-Net penalty")
results.append(benchmark(SGDClassifier(alpha=.0001, max_iter=50,
penalty="elasticnet")))
# Train NearestCentroid without threshold

print('=' * 80)
print("NearestCentroid (aka Rocchio classifier)")
results.append(benchmark(NearestCentroid()))
# Train sparse Naive Bayes classifiers

print('=' * 80)
print("Naive Bayes")
results.append(benchmark(MultinomialNB(alpha=.01)))
results.append(benchmark(BernoulliNB(alpha=.01)))
results.append(benchmark(ComplementNB(alpha=.1)))
print('=' * 80)
print("LinearSVC with L1-based feature selection")
# The smaller C, the stronger the regularization.
# The more regularization, the more sparsity.
results.append(benchmark(Pipeline([
('feature_selection', SelectFromModel(LinearSVC(penalty="l1", dual=False,
tol=1e-3))),
('classification', LinearSVC(penalty="l2"))])))
# make some plots
indices = np.arange(len(results))
results = [[x[i] for x in results] for i in range(4)]
clf_names, score, training_time, test_time = results

training_time = np.array(training_time) / np.max(training_time)
test_time = np.array(test_time) / np.max(test_time)

plt.title("Score")
plt.barh(indices, score, .2, label="score", color='navy')
plt.barh(indices + .3, training_time, .2, label="training time",
color='c')
plt.barh(indices + .6, test_time, .2, label="test time", color='darkorange')
plt.yticks(())
plt.subplots_adjust(left=.25)
plt.subplots_adjust(top=.95)
plt.subplots_adjust(bottom=.05)
for i, c in zip(indices, clf_names):

plt.text(-.3, i, c)
plt.show()
5.27 Decision Trees
Examples concerning the sklearn.tree module.
5.27.1 Decision Tree Regression
A 1D regression with decision tree.

The decision trees is used to fit a sine curve with addition noisy observation. As a result, it learns local linear regres-
sions approximating the sine curve.
We can see that if the maximum depth of the tree (controlled by the max_depth parameter) is set too high, the decision
trees learn too fine details of the training data and learn from the noise, i.e. they overfit.

print(__doc__)
# Import the necessary modules and libraries

import numpy as np

X = np.sort(5 * rng.rand(80, 1), axis=0)
y[::5] += 3 * (0.5 - rng.rand(16))

regr_1.fit(X, y)
regr_2.fit(X, y)
# Predict
X_test = np.arange(0.0, 5.0, 0.01)[:, np.newaxis]
y_1 = regr_1.predict(X_test)
# Plot the results

plt.figure()
5.27. Decision Trees 1383

plt.scatter(X, y, s=20, edgecolor="black",

c="darkorange", label="data")
plt.plot(X_test, y_1, color="cornflowerblue",
label="max_depth=2", linewidth=2)
plt.plot(X_test, y_2, color="yellowgreen", label="max_depth=5", linewidth=2)
plt.xlabel("data")
plt.ylabel("target")
plt.title("Decision Tree Regression")
plt.legend()
plt.show()
5.27.2 Multi-output Decision Tree Regression
An example to illustrate multi-output regression with decision tree.

The decision trees is used to predict simultaneously the noisy x and y observations of a circle given a single underlying
feature. As a result, it learns local linear regressions approximating the circle.
We can see that if the maximum depth of the tree (controlled by the max_depth parameter) is set too high, the decision
trees learn too fine details of the training data and learn from the noise, i.e. they overfit.

print(__doc__)
import numpy as np

X = np.sort(200 * rng.rand(100, 1) - 100, axis=0)
y = np.array([np.pi * np.sin(X).ravel(), np.pi * np.cos(X).ravel()]).T
y[::5, :] += (0.5 - rng.rand(20, 2))

regr_1.fit(X, y)
regr_2.fit(X, y)
regr_3.fit(X, y)
# Predict
X_test = np.arange(-100.0, 100.0, 0.01)[:, np.newaxis]

# Plot the results

plt.figure()
s = 25
plt.scatter(y[:, 0], y[:, 1], c="navy", s=s,
edgecolor="black", label="data")
plt.scatter(y_1[:, 0], y_1[:, 1], c="cornflowerblue", s=s,
edgecolor="black", label="max_depth=2")
plt.scatter(y_2[:, 0], y_2[:, 1], c="red", s=s,
plt.scatter(y_3[:, 0], y_3[:, 1], c="orange", s=s,
plt.xlim([-6, 6])
plt.ylim([-6, 6])
plt.xlabel("target 1")
plt.ylabel("target 2")
plt.title("Multi-output Decision Tree Regression")
plt.show()
5.27.3 Plot the decision surface of a decision tree on the iris dataset
Plot the decision surface of a decision tree trained on pairs of features of the iris dataset.
See decision tree for more information on the estimator.
For each pair of iris features, the decision tree learns decision boundaries made of combinations of simple thresholding
rules inferred from the training samples.

print(__doc__)
import numpy as np

# Parameters
n_classes = 3
plot_colors = "ryb"
plot_step = 0.02
# Load data
iris = load_iris()
for pairidx, pair in enumerate([[0, 1], [0, 2], [0, 3],

[1, 2], [1, 3], [2, 3]]):
# We only take the two corresponding features
X = iris.data[:, pair]
y = iris.target
# Train
clf = DecisionTreeClassifier().fit(X, y)
# Plot the decision boundary

plt.subplot(2, 3, pairidx + 1)

plt.tight_layout(h_pad=0.5, w_pad=0.5, pad=2.5)
cs = plt.contourf(xx, yy, Z, cmap=plt.cm.RdYlBu)
plt.xlabel(iris.feature_names[pair[0]])
plt.ylabel(iris.feature_names[pair[1]])

for i, color in zip(range(n_classes), plot_colors):
plt.scatter(X[idx, 0], X[idx, 1], c=color, label=iris.target_names[i],
cmap=plt.cm.RdYlBu, edgecolor='black', s=15)
plt.suptitle("Decision surface of a decision tree using paired features")

plt.legend(loc='lower right', borderpad=0, handletextpad=0)
plt.axis("tight")
plt.show()
5.27.4 Understanding the decision tree structure
The decision tree structure can be analysed to gain further insight on the relation between the features and the target
to predict. In this example, we show how to retrieve:
• the binary tree structure;
• the depth of each node and whether or not it’s a leaf;
• the nodes that were reached by a sample using the decision_path method;
• the leaf that was reached by a sample using the apply method;
• the rules that were used to predict a sample;
• the decision path shared by a group of samples.
Out:
The binary tree structure has 5 nodes and has the following tree structure:
node=0 test node: go to node 1 if X[:, 3] <= 0.800000011920929 else to node 2.
node=1 leaf node.
node=2 test node: go to node 3 if X[:, 2] <= 4.950000047683716 else to node 4.
node=3 leaf node.
node=4 leaf node.
Rules used to predict sample 0:

decision id node 0 : (X_test[0, 3] (= 2.4) > 0.800000011920929)

decision id node 2 : (X_test[0, 2] (= 5.1) > 4.950000047683716)
The following samples [0, 1] share the node [0 2] in the tree

It is 40.0 % of all nodes.
import numpy as np

iris = load_iris()
X = iris.data
y = iris.target
estimator = DecisionTreeClassifier(max_leaf_nodes=3, random_state=0)

# The decision estimator has an attribute called tree_ which stores the entire
# tree structure and allows access to low level attributes. The binary tree
# tree_ is represented as a number of parallel arrays. The i-th element of each
# array holds information about the node ì`. Node 0 is the tree's root. NOTE:
# Some of the arrays only apply to either leaves or split nodes, resp. In this
# case the values of nodes of the other type are arbitrary!
#
# Among those arrays, we have:
# - left_child, id of the left child of the node
# - right_child, id of the right child of the node
# - feature, feature used for splitting the node
# - threshold, threshold value at the node
#
# Using those arrays, we can parse the tree structure:
n_nodes = estimator.tree_.node_count
children_left = estimator.tree_.children_left
children_right = estimator.tree_.children_right
feature = estimator.tree_.feature
threshold = estimator.tree_.threshold
# The tree structure can be traversed to compute various properties such

# as the depth of each node and whether or not it is a leaf.
node_depth = np.zeros(shape=n_nodes, dtype=np.int64)
is_leaves = np.zeros(shape=n_nodes, dtype=bool)
stack = [(0, -1)] # seed is the root node id and its parent depth
while len(stack) > 0:
node_id, parent_depth = stack.pop()
node_depth[node_id] = parent_depth + 1
# If we have a test node

if (children_left[node_id] != children_right[node_id]):
stack.append((children_left[node_id], parent_depth + 1))
stack.append((children_right[node_id], parent_depth + 1))
else:
is_leaves[node_id] = True
print("The binary tree structure has %s nodes and has "

"the following tree structure:"
% n_nodes)
for i in range(n_nodes):
if is_leaves[i]:
print("%snode=%s leaf node." % (node_depth[i] * "\t", i))
else:
print("%snode=%s test node: go to node %s if X[:, %s] <= %s else to "
"node %s."
% (node_depth[i] * "\t",
i,
children_left[i],
feature[i],
threshold[i],
children_right[i],
))
print()
# First let's retrieve the decision path of each sample. The decision_path
# method allows to retrieve the node indicator functions. A non zero element of
# indicator matrix at the position (i, j) indicates that the sample i goes
# through the node j.
node_indicator = estimator.decision_path(X_test)
# Similarly, we can also have the leaves ids reached by each sample.
leave_id = estimator.apply(X_test)
# Now, it's possible to get the tests that were used to predict a sample or
# a group of samples. First, let's make it for the sample.
sample_id = 0
node_index = node_indicator.indices[node_indicator.indptr[sample_id]:
node_indicator.indptr[sample_id + 1]]
print('Rules used to predict sample %s: ' % sample_id)

for node_id in node_index:
if leave_id[sample_id] == node_id:
continue
if (X_test[sample_id, feature[node_id]] <= threshold[node_id]):

threshold_sign = "<="
else:
threshold_sign = ">"
print("decision id node %s : (X_test[%s, %s] (= %s) %s %s)"

% (node_id,
sample_id,
feature[node_id],
X_test[sample_id, feature[node_id]],
threshold_sign,

threshold[node_id]))
# For a group of samples, we have the following common node.

sample_ids = [0, 1]
common_nodes = (node_indicator.toarray()[sample_ids].sum(axis=0) ==
len(sample_ids))
common_node_id = np.arange(n_nodes)[common_nodes]
print("\nThe following samples %s share the node %s in the tree"

% (sample_ids, common_node_id))
print("It is %s %% of all nodes." % (100 * len(common_node_id) / n_nodes,))


CHAPTER
SIX
API REFERENCE
This is the class and function reference of scikit-learn. Please refer to the full user guide for further details, as the class
and function raw specifications may not be enough to give full guidelines on their uses. For reference on concepts
repeated across the API, see Glossary of Common Terms and API Elements.
6.1 sklearn.base: Base classes and utility functions
Base classes for all estimators.
6.1.1 Base classes
base.BaseEstimator Base class for all estimators in scikit-learn

base.BiclusterMixin Mixin class for all bicluster estimators in scikit-learn
base.ClassifierMixin Mixin class for all classifiers in scikit-learn.
base.ClusterMixin Mixin class for all cluster estimators in scikit-learn.
base.DensityMixin Mixin class for all density estimators in scikit-learn.
base.RegressorMixin Mixin class for all regression estimators in scikit-learn.
base.TransformerMixin Mixin class for all transformers in scikit-learn.
sklearn.base.BaseEstimator
class sklearn.base.BaseEstimator
Base class for all estimators in scikit-learn
Notes
All estimators should specify all the parameters that can be set at the class level in their __init__ as explicit
keyword arguments (no *args or **kwargs).
Methods

1393
__init__($self, /, *args, **kwargs)

Initialize self. See help(type(self)) for accurate signature.
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.base.BaseEstimator
sklearn.base.BiclusterMixin
class sklearn.base.BiclusterMixin
Mixin class for all bicluster estimators in scikit-learn
Attributes
biclusters_ Convenient way to get row and column indicators together.
Methods
get_indices(i) Row and column indices of the i’th bicluster.

get_shape(i) Shape of the i’th bicluster.
get_submatrix(i, data) Returns the submatrix corresponding to bicluster i.

biclusters_
Convenient way to get row and column indicators together.
Returns the rows_ and columns_ members.
get_indices(i)
Row and column indices of the i’th bicluster.
Only works if rows_ and columns_ attributes exist.
1394 Chapter 6. API Reference

Parameters
i [int] The index of the cluster.
Returns
row_ind [np.array, dtype=np.intp] Indices of rows in the dataset that belong to the bicluster.
col_ind [np.array, dtype=np.intp] Indices of columns in the dataset that belong to the biclus-
ter.
get_shape(i)
Shape of the i’th bicluster.
Parameters
Returns
shape [(int, int)] Number of rows and columns (resp.) in the bicluster.
get_submatrix(i, data)
Returns the submatrix corresponding to bicluster i.
Parameters
data [array] The data.
Returns
submatrix [array] The submatrix corresponding to bicluster i.
Notes
Works with sparse matrices. Only works if rows_ and columns_ attributes exist.
sklearn.base.ClassifierMixin
class sklearn.base.ClassifierMixin
Mixin class for all classifiers in scikit-learn.
Methods
labels.

Parameters
6.1. sklearn.base: Base classes and utility functions 1395


Returns
sklearn.base.ClusterMixin
class sklearn.base.ClusterMixin
Mixin class for all cluster estimators in scikit-learn.
Methods
fit_predict(X[, y]) Performs clustering on X and returns cluster labels.

fit_predict(X, y=None)
Performs clustering on X and returns cluster labels.
Parameters
X [ndarray, shape (n_samples, n_features)] Input data.
y [Ignored] not used, present for API consistency by convention.
Returns
labels [ndarray, shape (n_samples,)] cluster labels
sklearn.base.DensityMixin
class sklearn.base.DensityMixin
Mixin class for all density estimators in scikit-learn.
Methods
score(X[, y]) Returns the score of the model on the data X

score(X, y=None)
Returns the score of the model on the data X
Parameters
X [array-like, shape = (n_samples, n_features)]
Returns

score [float]
sklearn.base.RegressorMixin
class sklearn.base.RegressorMixin
Mixin class for all regression estimators in scikit-learn.
Methods
diction.

Parameters
Returns
sklearn.base.TransformerMixin
class sklearn.base.TransformerMixin
Mixin class for all transformers in scikit-learn.
Methods
fit_transform(X[, y]) Fit to data, then transform it.

fit_transform(X, y=None, **fit_params)
Fit to data, then transform it.
Fits transformer to X and y with optional parameters fit_params and returns a transformed version of X.

Parameters
X [numpy array of shape [n_samples, n_features]] Training set.
y [numpy array of shape [n_samples]] Target values.
Returns
X_new [numpy array of shape [n_samples, n_features_new]] Transformed array.
Examples using sklearn.base.TransformerMixin
6.1.2 Functions
base.clone(estimator[, safe]) Constructs a new estimator with the same parameters.

base.is_classifier(estimator) Returns True if the given estimator is (probably) a classi-
fier.
base.is_regressor(estimator) Returns True if the given estimator is (probably) a regres-
sor.
config_context(**new_config) Context manager for global scikit-learn configuration
get_config() Retrieve current values for configuration set by
set_config
set_config([assume_finite, working_memory]) Set global scikit-learn configuration
show_versions() Print useful debugging information
sklearn.base.clone
sklearn.base.clone(estimator, safe=True)
Constructs a new estimator with the same parameters.
Clone does a deep copy of the model in an estimator without actually copying attached data. It yields a new
estimator with the same parameters that has not been fit on any data.
Parameters
estimator [estimator object, or list, tuple or set of objects] The estimator or group of estimators
to be cloned
safe [boolean, optional] If safe is false, clone will fall back to a deep copy on objects that are
not estimators.
sklearn.base.is_classifier
sklearn.base.is_classifier(estimator)
Returns True if the given estimator is (probably) a classifier.
Parameters
estimator [object] Estimator object to test.
Returns
out [bool] True if estimator is a classifier and False otherwise.

sklearn.base.is_regressor
sklearn.base.is_regressor(estimator)
Returns True if the given estimator is (probably) a regressor.
Parameters
estimator [object] Estimator object to test.
Returns
out [bool] True if estimator is a regressor and False otherwise.
sklearn.config_context
sklearn.config_context(**new_config)
Context manager for global scikit-learn configuration
Parameters
assume_finite [bool, optional] If True, validation for finiteness will be skipped, saving time, but
leading to potential crashes. If False, validation for finiteness will be performed, avoiding
error. Global default: False.
working_memory [int, optional] If set, scikit-learn will attempt to limit the size of temporary
arrays to this number of MiB (per job when parallelised), often saving both computation
time and memory on expensive operations that can be performed in chunks. Global default:
1024.
Notes
All settings, not just those presently modified, will be returned to their previous values when the context manager
is exited. This is not thread-safe.
Examples
>>> import sklearn

>>> from sklearn.utils.validation import assert_all_finite
... assert_all_finite([float('nan')])
... with sklearn.config_context(assume_finite=False):
... assert_all_finite([float('nan')])
...
Traceback (most recent call last):
...
ValueError: Input contains NaN, ...
sklearn.get_config
sklearn.get_config()
Retrieve current values for configuration set by set_config
Returns

config [dict] Keys are parameter names that can be passed to set_config.
sklearn.set_config
sklearn.set_config(assume_finite=None, working_memory=None)
Set global scikit-learn configuration
Parameters
assume_finite [bool, optional] If True, validation for finiteness will be skipped, saving time, but
leading to potential crashes. If False, validation for finiteness will be performed, avoiding
error. Global default: False.
working_memory [int, optional] If set, scikit-learn will attempt to limit the size of temporary
arrays to this number of MiB (per job when parallelised), often saving both computation
time and memory on expensive operations that can be performed in chunks. Global default:
1024.
sklearn.show_versions
sklearn.show_versions()
Print useful debugging information
6.2 sklearn.calibration: Probability Calibration
Calibration of predicted probabilities.

User guide: See the Probability calibration section for further details.
calibration.CalibratedClassifierCV ([. . . ]) Probability calibration with isotonic regression or sigmoid.
6.2.1 sklearn.calibration.CalibratedClassifierCV
class sklearn.calibration.CalibratedClassifierCV(base_estimator=None,
method=’sigmoid’, cv=’warn’)
Probability calibration with isotonic regression or sigmoid.
With this class, the base_estimator is fit on the train set of the cross-validation generator and the test set is used
for calibration. The probabilities for each of the folds are then averaged for prediction. In case that cv=”prefit”
is passed to __init__, it is assumed that base_estimator has been fitted already and all data is used for calibration.
Note that data for fitting the classifier and for calibrating it must be disjoint.
Parameters
base_estimator [instance BaseEstimator] The classifier whose output decision function needs

to be calibrated to offer more accurate predict_proba outputs. If cv=prefit, the classifier must
have been fit already on data.
method [‘sigmoid’ or ‘isotonic’] The method to use for calibration. Can be ‘sigmoid’ which
corresponds to Platt’s method or ‘isotonic’ which is a non-parametric approach. It is not
advised to use isotonic calibration with too few calibration samples (<<1000) since it
tends to overfit. Use sigmoids (Platt’s calibration) in this case.
cv [integer, cross-validation generator, iterable or “prefit”, optional] Determines the cross-
validation splitting strategy. Possible inputs for cv are:
• CV splitter,
StratifiedKFold is used. If y is neither binary nor multiclass, sklearn.
model_selection.KFold is used.
If “prefit” is passed, it is assumed that base_estimator has been fitted already and all data is
used for calibration.
v0.22.
Attributes
classes_ [array, shape (n_classes)] The class labels.
calibrated_classifiers_ [list (len() equal to cv or 1 if cv == “prefit”)] The list of calibrated
classifiers, one for each crossvalidation fold, which has been fitted on all but the validation
fold and calibrated on the validation fold.
References
[1], [2], [3], [4]
Methods
fit(X, y[, sample_weight]) Fit the calibrated model

predict(X) Predict the target of new samples.
predict_proba(X) Posterior probabilities of classification
labels.
__init__(base_estimator=None, method=’sigmoid’, cv=’warn’)

Fit the calibrated model
6.2. sklearn.calibration: Probability Calibration 1401

Parameters
samples are equally weighted.
Returns
self [object] Returns an instance of self.
Parameters
Returns
predict(X)
Predict the target of new samples. Can be different from the prediction of the uncalibrated classifier.
Parameters
X [array-like, shape (n_samples, n_features)] The samples.
Returns
C [array, shape (n_samples,)] The predicted class.
predict_proba(X)
Posterior probabilities of classification
This function returns posterior probabilities of classification according to each class on an array of test
vectors X.
Parameters
X [array-like, shape (n_samples, n_features)] The samples.
Returns
C [array, shape (n_samples, n_classes)] The predicted probas.
Parameters
Returns

of a nested object.
Returns
self
Examples using sklearn.calibration.CalibratedClassifierCV
• Probability Calibration curves

• Probability calibration of classifiers
calibration.calibration_curve(y_true, Compute true and predicted probabilities for a calibration

y_prob) curve.
6.2.2 sklearn.calibration.calibration_curve
sklearn.calibration.calibration_curve(y_true, y_prob, normalize=False, n_bins=5)

Compute true and predicted probabilities for a calibration curve.
The method assumes the inputs come from a binary classifier.
Calibration curves may also be referred to as reliability diagrams.
Parameters
y_true [array, shape (n_samples,)] True targets.
y_prob [array, shape (n_samples,)] Probabilities of the positive class.
normalize [bool, optional, default=False] Whether y_prob needs to be normalized into the bin
[0, 1], i.e. is not a proper probability. If True, the smallest value in y_prob is mapped onto
0 and the largest one onto 1.
n_bins [int] Number of bins. A bigger number requires more data. Bins with no data points
(i.e. without corresponding values in y_prob) will not be returned, thus there may be fewer
than n_bins in the return value.
Returns
prob_true [array, shape (n_bins,) or smaller] The true probability in each bin (fraction of posi-
tives).
prob_pred [array, shape (n_bins,) or smaller] The mean predicted probability in each bin.
References
Alexandru Niculescu-Mizil and Rich Caruana (2005) Predicting Good Probabilities With Supervised Learning,
in Proceedings of the 22nd International Conference on Machine Learning (ICML). See section 4 (Qualitative
Analysis of Predictions).
6.2. sklearn.calibration: Probability Calibration 1403

Examples using sklearn.calibration.calibration_curve

6.3 sklearn.cluster: Clustering
The sklearn.cluster module gathers popular unsupervised clustering algorithms.

User guide: See the Clustering section for further details.
6.3.1 Classes
cluster.AffinityPropagation([damping, . . . ]) Perform Affinity Propagation Clustering of data.

cluster.AgglomerativeClustering([. . . ]) Agglomerative Clustering
cluster.Birch([threshold, branching_factor, . . . ]) Implements the Birch clustering algorithm.
cluster.DBSCAN ([eps, min_samples, metric, . . . ]) Perform DBSCAN clustering from vector array or distance
matrix.
cluster.FeatureAgglomeration([n_clusters, Agglomerate features.
. . . ])
cluster.KMeans([n_clusters, init, n_init, . . . ]) K-Means clustering
cluster.MiniBatchKMeans([n_clusters, init, . . . ]) Mini-Batch K-Means clustering
cluster.MeanShift([bandwidth, seeds, . . . ]) Mean shift clustering using a flat kernel.
cluster.SpectralClustering([n_clusters, . . . ]) Apply clustering to a projection to the normalized lapla-
cian.
sklearn.cluster.AffinityPropagation
class sklearn.cluster.AffinityPropagation(damping=0.5, max_iter=200, conver-

gence_iter=15, copy=True, preference=None,
affinity=’euclidean’, verbose=False)
Perform Affinity Propagation Clustering of data.
Parameters
damping [float, optional, default: 0.5] Damping factor (between 0.5 and 1) is the extent to
which the current value is maintained relative to incoming values (weighted 1 - damping).
This in order to avoid numerical oscillations when updating these values (messages).
max_iter [int, optional, default: 200] Maximum number of iterations.
convergence_iter [int, optional, default: 15] Number of iterations with no change in the number
of estimated clusters that stops the convergence.
copy [boolean, optional, default: True] Make a copy of input data.
preference [array-like, shape (n_samples,) or float, optional] Preferences for each point - points
with larger values of preferences are more likely to be chosen as exemplars. The number of
exemplars, ie of clusters, is influenced by the input preferences value. If the preferences are
not passed as arguments, they will be set to the median of the input similarities.

affinity [string, optional, default=‘‘euclidean‘‘] Which affinity to use. At the moment

precomputed and euclidean are supported. euclidean uses the negative squared
euclidean distance between points.
verbose [boolean, optional, default: False] Whether to be verbose.
Attributes
cluster_centers_indices_ [array, shape (n_clusters,)] Indices of cluster centers
cluster_centers_ [array, shape (n_clusters, n_features)] Cluster centers (if affinity !=
precomputed).
labels_ [array, shape (n_samples,)] Labels of each point
affinity_matrix_ [array, shape (n_samples, n_samples)] Stores the affinity matrix used in fit.
n_iter_ [int] Number of iterations taken to converge.
Notes
For an example, see examples/cluster/plot_affinity_propagation.py.

The algorithmic complexity of affinity propagation is quadratic in the number of points.
When fit does not converge, cluster_centers_ becomes an empty array and all training samples will be
labelled as -1. In addition, predict will then label every sample as -1.
When all training samples have equal similarities and equal preferences, the assignment of cluster centers and
labels depends on the preference. If the preference is smaller than the similarities, fit will result in a single
cluster center and label 0 for every sample. Otherwise, every training sample becomes its own cluster center
and is assigned a unique label.
References
Brendan J. Frey and Delbert Dueck, “Clustering by Passing Messages Between Data Points”, Science Feb. 2007
Examples
>>> from sklearn.cluster import AffinityPropagation

>>> X = np.array([[1, 2], [1, 4], [1, 0],
... [4, 2], [4, 4], [4, 0]])
>>> clustering = AffinityPropagation().fit(X)
>>> clustering
AffinityPropagation(affinity='euclidean', convergence_iter=15, copy=True,
damping=0.5, max_iter=200, preference=None, verbose=False)
>>> clustering.labels_
array([0, 0, 0, 1, 1, 1])
>>> clustering.predict([[0, 0], [4, 4]])
array([0, 1])
>>> clustering.cluster_centers_
array([[1, 2],
[4, 2]])
6.3. sklearn.cluster: Clustering 1405

Methods
fit(X[, y]) Create affinity matrix from negative euclidean dis-

tances, then apply affinity propagation clustering.
predict(X) Predict the closest cluster each sample in X belongs to.
__init__(damping=0.5, max_iter=200, convergence_iter=15, copy=True, preference=None, affin-

ity=’euclidean’, verbose=False)
fit(X, y=None)
Create affinity matrix from negative euclidean distances, then apply affinity propagation clustering.
Parameters
X [array-like, shape (n_samples, n_features) or (n_samples, n_samples)] Data matrix or, if
affinity is precomputed, matrix of similarities / affinities.
y [Ignored]
Parameters
Returns
Parameters
Returns
predict(X)
Predict the closest cluster each sample in X belongs to.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] New data to predict.
Returns
labels [array, shape (n_samples,)] Index of the cluster each sample belongs to.
of a nested object.

Returns
self
Examples using sklearn.cluster.AffinityPropagation
• Demo of affinity propagation clustering algorithm

• Comparing different clustering algorithms on toy datasets
sklearn.cluster.AgglomerativeClustering
class sklearn.cluster.AgglomerativeClustering(n_clusters=2, affinity=’euclidean’, mem-

ory=None, connectivity=None, com-
pute_full_tree=’auto’, linkage=’ward’,
pooling_func=’deprecated’)
Agglomerative Clustering
Recursively merges the pair of clusters that minimally increases a given linkage distance.
Parameters
n_clusters [int, default=2] The number of clusters to find.
affinity [string or callable, default: “euclidean”] Metric used to compute the linkage. Can be
“euclidean”, “l1”, “l2”, “manhattan”, “cosine”, or ‘precomputed’. If linkage is “ward”, only
“euclidean” is accepted.
memory [None, str or object with the joblib.Memory interface, optional] Used to cache the
output of the computation of the tree. By default, no caching is done. If a string is given, it
is the path to the caching directory.
connectivity [array-like or callable, optional] Connectivity matrix. Defines for each sample the
neighboring samples following a given structure of the data. This can be a connectivity
matrix itself or a callable that transforms the data into a connectivity matrix, such as de-
rived from kneighbors_graph. Default is None, i.e, the hierarchical clustering algorithm is
unstructured.
compute_full_tree [bool or ‘auto’ (optional)] Stop early the construction of the tree at
n_clusters. This is useful to decrease computation time if the number of clusters is not
small compared to the number of samples. This option is useful only when specifying a
connectivity matrix. Note also that when varying the number of clusters and using caching,
it may be advantageous to compute the full tree.
linkage [{“ward”, “complete”, “average”, “single”}, optional (default=”ward”)] Which linkage
criterion to use. The linkage criterion determines which distance to use between sets of
observation. The algorithm will merge the pairs of cluster that minimize this criterion.
• ward minimizes the variance of the clusters being merged.
• average uses the average of the distances of each observation of the two sets.
• complete or maximum linkage uses the maximum distances between all observations of
the two sets.
• single uses the minimum of the distances between all observations of the two sets.

pooling_func [callable, default=’deprecated’] Ignored.

Deprecated since version 0.20: pooling_func has been deprecated in 0.20 and will be
removed in 0.22.
Attributes
labels_ [array [n_samples]] cluster labels for each point
n_leaves_ [int] Number of leaves in the hierarchical tree.
n_components_ [int] The estimated number of connected components in the graph.
children_ [array-like, shape (n_samples-1, 2)] The children of each non-leaf node. Values
less than n_samples correspond to leaves of the tree which are the original samples. A
node i greater than or equal to n_samples is a non-leaf node and has children children_[i -
n_samples]. Alternatively at the i-th iteration, children[i][0] and children[i][1] are merged
to form node n_samples + i
Examples
>>> from sklearn.cluster import AgglomerativeClustering

>>> X = np.array([[1, 2], [1, 4], [1, 0],
... [4, 2], [4, 4], [4, 0]])
>>> clustering = AgglomerativeClustering().fit(X)
>>> clustering
AgglomerativeClustering(affinity='euclidean', compute_full_tree='auto',
connectivity=None, linkage='ward', memory=None, n_clusters=2,
pooling_func='deprecated')
array([1, 1, 1, 0, 0, 0])
Methods
fit(X[, y]) Fit the hierarchical clustering on the data

__init__(n_clusters=2, affinity=’euclidean’, memory=None, connectivity=None, com-

pute_full_tree=’auto’, linkage=’ward’, pooling_func=’deprecated’)
fit(X, y=None)
Fit the hierarchical clustering on the data
Parameters
X [array-like, shape = [n_samples, n_features]] Training data. Shape [n_samples,
n_features], or [n_samples, n_samples] if affinity==’precomputed’.
y [Ignored]
Returns
self

Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.cluster.AgglomerativeClustering
• A demo of structured Ward hierarchical clustering on an image of coins

• Various Agglomerative Clustering on a 2D embedding of digits
• Hierarchical clustering: structured vs unstructured ward
• Comparing different hierarchical linkage methods on toy datasets
sklearn.cluster.Birch
class sklearn.cluster.Birch(threshold=0.5, branching_factor=50, n_clusters=3, com-

pute_labels=True, copy=True)
Implements the Birch clustering algorithm.
It is a memory-efficient, online-learning algorithm provided as an alternative to MiniBatchKMeans. It con-
structs a tree data structure with the cluster centroids being read off the leaf. These can be either the final cluster
centroids or can be provided as input to another clustering algorithm such as AgglomerativeClustering.

Parameters
threshold [float, default 0.5] The radius of the subcluster obtained by merging a new sample
and the closest subcluster should be lesser than the threshold. Otherwise a new subcluster is
started. Setting this value to be very low promotes splitting and vice-versa.
branching_factor [int, default 50] Maximum number of CF subclusters in each node. If a new
samples enters such that the number of subclusters exceed the branching_factor then that
node is split into two nodes with the subclusters redistributed in each. The parent subcluster
of that node is removed and two new subclusters are added as parents of the 2 split nodes.
n_clusters [int, instance of sklearn.cluster model, default 3] Number of clusters after the final
clustering step, which treats the subclusters from the leaves as new samples.
• None : the final clustering step is not performed and the subclusters are returned as they
are.
• sklearn.cluster Estimator : If a model is provided, the model is fit treating the subclusters
as new samples and the initial data is mapped to the label of the closest subcluster.
• int : the model fit is AgglomerativeClustering with n_clusters set to be equal to
the int.
compute_labels [bool, default True] Whether or not to compute labels for each fit.
copy [bool, default True] Whether or not to make a copy of the given data. If set to False, the
initial data will be overwritten.
Attributes
root_ [_CFNode] Root of the CFTree.
dummy_leaf_ [_CFNode] Start pointer to all the leaves.
subcluster_centers_ [ndarray,] Centroids of all subclusters read directly from the leaves.
subcluster_labels_ [ndarray,] Labels assigned to the centroids of the subclusters after they are
clustered globally.
labels_ [ndarray, shape (n_samples,)] Array of labels assigned to the input data. if partial_fit is
used instead of fit, they are assigned to the last batch of data.
Notes
The tree data structure consists of nodes with each node consisting of a number of subclusters. The maximum
number of subclusters in a node is determined by the branching factor. Each subcluster maintains a linear sum,
squared sum and the number of samples in that subcluster. In addition, each subcluster can also have a node as
its child, if the subcluster is not a member of a leaf node.
For a new point entering the root, it is merged with the subcluster closest to it and the linear sum, squared sum
and the number of samples of that subcluster are updated. This is done recursively till the properties of the leaf
node are updated.
References
• Tian Zhang, Raghu Ramakrishnan, Maron Livny BIRCH: An efficient data clustering method for large
databases. http://www.cs.sfu.ca/CourseCentral/459/han/papers/zhang96.pdf
• Roberto Perdisci JBirch - Java implementation of BIRCH clustering algorithm https://code.google.com/
archive/p/jbirch

Examples
>>> from sklearn.cluster import Birch

>>> X = [[0, 1], [0.3, 1], [-0.3, 1], [0, -1], [0.3, -1], [-0.3, -1]]
>>> brc = Birch(branching_factor=50, n_clusters=None, threshold=0.5,
... compute_labels=True)
>>> brc.fit(X)
Birch(branching_factor=50, compute_labels=True, copy=True, n_clusters=None,
threshold=0.5)
>>> brc.predict(X)
array([0, 0, 0, 1, 1, 1])
Methods
fit(X[, y]) Build a CF Tree for the input data.

partial_fit([X, y]) Online learning.
predict(X) Predict data using the centroids_ of subclusters.
transform(X) Transform X into subcluster centroids dimension.
__init__(threshold=0.5, branching_factor=50, n_clusters=3, compute_labels=True, copy=True)

fit(X, y=None)
Build a CF Tree for the input data.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Input data.
y [Ignored]
Parameters
Returns
Parameters
Returns


Parameters
Returns
partial_fit(X=None, y=None)
Online learning. Prevents rebuilding of CFTree from scratch.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features), None] Input data. If X is not
provided, only the global clustering step is done.
y [Ignored]
predict(X)
Predict data using the centroids_ of subclusters.
Avoid computation of the row norms of X.
Parameters
Returns
labels [ndarray, shape(n_samples)] Labelled data.
of a nested object.
Returns
self
transform(X)
Transform X into subcluster centroids dimension.
Each dimension represents the distance from the sample point to each cluster centroid.
Parameters
Returns
X_trans [{array-like, sparse matrix}, shape (n_samples, n_clusters)] Transformed data.
Examples using sklearn.cluster.Birch
• Compare BIRCH and MiniBatchKMeans


sklearn.cluster.DBSCAN
class sklearn.cluster.DBSCAN(eps=0.5, min_samples=5, metric=’euclidean’, metric_params=None,

algorithm=’auto’, leaf_size=30, p=None, n_jobs=None)
Perform DBSCAN clustering from vector array or distance matrix.
DBSCAN - Density-Based Spatial Clustering of Applications with Noise. Finds core samples of high density
and expands clusters from them. Good for data which contains clusters of similar density.
Parameters
eps [float, optional] The maximum distance between two samples for them to be considered as
in the same neighborhood.
min_samples [int, optional] The number of samples (or total weight) in a neighborhood for a
point to be considered as a core point. This includes the point itself.
metric [string, or callable] The metric to use when calculating distance between instances in
a feature array. If metric is a string or callable, it must be one of the options allowed
by sklearn.metrics.pairwise_distances for its metric parameter. If metric
is “precomputed”, X is assumed to be a distance matrix and must be square. X may be
a sparse matrix, in which case only “nonzero” elements may be considered neighbors for
DBSCAN.
New in version 0.17: metric precomputed to accept precomputed sparse matrix.
metric_params [dict, optional] Additional keyword arguments for the metric function.
algorithm [{‘auto’, ‘ball_tree’, ‘kd_tree’, ‘brute’}, optional] The algorithm to be used by the
NearestNeighbors module to compute pointwise distances and find nearest neighbors. See
NearestNeighbors module documentation for details.
leaf_size [int, optional (default = 30)] Leaf size passed to BallTree or cKDTree. This can affect
the speed of the construction and query, as well as the memory required to store the tree.
The optimal value depends on the nature of the problem.
p [float, optional] The power of the Minkowski metric to be used to calculate distance between
points.
n_jobs [int or None, optional (default=None)] The number of parallel jobs to run. None means
1 unless in a joblib.parallel_backend context. -1 means using all processors. See
Attributes
core_sample_indices_ [array, shape = [n_core_samples]] Indices of core samples.
components_ [array, shape = [n_core_samples, n_features]] Copy of each core sample found
by training.
labels_ [array, shape = [n_samples]] Cluster labels for each point in the dataset given to fit().
Noisy samples are given the label -1.
Notes
For an example, see examples/cluster/plot_dbscan.py.

This implementation bulk-computes all neighborhood queries, which increases the memory complexity to
O(n.d) where d is the average number of neighbors, while original DBSCAN had memory complexity O(n).
It may attract a higher memory complexity when querying these nearest neighborhoods, depending on the
algorithm.
One way to avoid the query complexity is to pre-compute sparse neighborhoods in chunks
using NearestNeighbors.radius_neighbors_graph with mode='distance', then using
metric='precomputed' here.
Another way to reduce memory and computation time is to remove (near-)duplicate points and use
sample_weight instead.
References
Ester, M., H. P. Kriegel, J. Sander, and X. Xu, “A Density-Based Algorithm for Discovering Clusters in Large
Spatial Databases with Noise”. In: Proceedings of the 2nd International Conference on Knowledge Discovery
and Data Mining, Portland, OR, AAAI Press, pp. 226-231. 1996
Examples
>>> from sklearn.cluster import DBSCAN

>>> X = np.array([[1, 2], [2, 2], [2, 3],
... [8, 7], [8, 8], [25, 80]])
>>> clustering = DBSCAN(eps=3, min_samples=2).fit(X)
array([ 0, 0, 0, 1, 1, -1])
>>> clustering
DBSCAN(algorithm='auto', eps=3, leaf_size=30, metric='euclidean',
metric_params=None, min_samples=2, n_jobs=None, p=None)
Methods
fit(X[, y, sample_weight]) Perform DBSCAN clustering from features or distance

matrix.
fit_predict(X[, y, sample_weight]) Performs clustering on X and returns cluster labels.
__init__(eps=0.5, min_samples=5, metric=’euclidean’, metric_params=None, algorithm=’auto’,

leaf_size=30, p=None, n_jobs=None)
fit(X, y=None, sample_weight=None)
Perform DBSCAN clustering from features or distance matrix.
Parameters
X [array or sparse (CSR) matrix of shape (n_samples, n_features), or array of shape
(n_samples, n_samples)] A feature array, or array of distances between samples if
metric='precomputed'.
sample_weight [array, shape (n_samples,), optional] Weight of each sample, such that a
sample with a weight of at least min_samples is by itself a core sample; a sample

with negative weight may inhibit its eps-neighbor from being core. Note that weights are
absolute, and default to 1.
y [Ignored]
fit_predict(X, y=None, sample_weight=None)
Parameters
sample_weight [array, shape (n_samples,), optional] Weight of each sample, such that a
sample with a weight of at least min_samples is by itself a core sample; a sample
with negative weight may inhibit its eps-neighbor from being core. Note that weights are
absolute, and default to 1.
y [Ignored]
Returns
y [ndarray, shape (n_samples,)] cluster labels
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.cluster.DBSCAN

sklearn.cluster.FeatureAgglomeration
class sklearn.cluster.FeatureAgglomeration(n_clusters=2, affinity=’euclidean’, mem-

ory=None, connectivity=None, com-
pute_full_tree=’auto’, linkage=’ward’, pool-
ing_func=<function mean>)
Agglomerate features.

Similar to AgglomerativeClustering, but recursively merges features instead of samples.

Parameters
n_clusters [int, default 2] The number of clusters to find.
affinity [string or callable, default “euclidean”] Metric used to compute the linkage. Can be
“euclidean”, “l1”, “l2”, “manhattan”, “cosine”, or ‘precomputed’. If linkage is “ward”, only
“euclidean” is accepted.
memory [None, str or object with the joblib.Memory interface, optional] Used to cache the
output of the computation of the tree. By default, no caching is done. If a string is given, it
is the path to the caching directory.
connectivity [array-like or callable, optional] Connectivity matrix. Defines for each feature the
neighboring features following a given structure of the data. This can be a connectivity
matrix itself or a callable that transforms the data into a connectivity matrix, such as de-
rived from kneighbors_graph. Default is None, i.e, the hierarchical clustering algorithm is
unstructured.
compute_full_tree [bool or ‘auto’, optional, default “auto”] Stop early the construction of the
tree at n_clusters. This is useful to decrease computation time if the number of clusters is
not small compared to the number of features. This option is useful only when specifying a
connectivity matrix. Note also that when varying the number of clusters and using caching,
it may be advantageous to compute the full tree.
linkage [{“ward”, “complete”, “average”, “single”}, optional (default=”ward”)] Which linkage
criterion to use. The linkage criterion determines which distance to use between sets of
features. The algorithm will merge the pairs of cluster that minimize this criterion.
• ward minimizes the variance of the clusters being merged.
• average uses the average of the distances of each feature of the two sets.
• complete or maximum linkage uses the maximum distances between all features of the
two sets.
• single uses the minimum of the distances between all observations of the two sets.
pooling_func [callable, default np.mean] This combines the values of agglomerated features
into a single value, and should accept an array of shape [M, N] and the keyword argument
axis=1, and reduce it to an array of size [M].
Attributes
labels_ [array-like, (n_features,)] cluster labels for each feature.
n_leaves_ [int] Number of leaves in the hierarchical tree.
n_components_ [int] The estimated number of connected components in the graph.
children_ [array-like, shape (n_nodes-1, 2)] The children of each non-leaf node. Values less
than n_features correspond to leaves of the tree which are the original samples. A node
i greater than or equal to n_features is a non-leaf node and has children children_[i -
n_features]. Alternatively at the i-th iteration, children[i][0] and children[i][1] are merged
to form node n_features + i

Examples

>>> from sklearn import datasets, cluster
>>> images = digits.images
>>> X = np.reshape(images, (len(images), -1))
>>> agglo = cluster.FeatureAgglomeration(n_clusters=32)
>>> agglo.fit(X)
FeatureAgglomeration(affinity='euclidean', compute_full_tree='auto',
connectivity=None, linkage='ward', memory=None, n_clusters=32,
pooling_func=...)
>>> X_reduced = agglo.transform(X)
>>> X_reduced.shape
(1797, 32)
Methods
fit(X[, y]) Fit the hierarchical clustering on the data

inverse_transform(Xred) Inverse the transformation.
pooling_func(a[, axis, dtype, out, keepdims]) Compute the arithmetic mean along the specified axis.
transform(X) Transform a new matrix using the built clustering
__init__(n_clusters=2, affinity=’euclidean’, memory=None, connectivity=None, com-

pute_full_tree=’auto’, linkage=’ward’, pooling_func=<function mean>)
fit(X, y=None, **params)
Fit the hierarchical clustering on the data
Parameters
X [array-like, shape = [n_samples, n_features]] The data
y [Ignored]
Returns
self
Parameters
Returns

Parameters
Returns
inverse_transform(Xred)
Inverse the transformation. Return a vector of size nb_features with the values of Xred assigned to each
group of features
Parameters
Xred [array-like, shape=[n_samples, n_clusters] or [n_clusters,]] The values to be assigned
to each cluster of samples
Returns
X [array, shape=[n_samples, n_features] or [n_features]] A vector of size n_samples with
the values of Xred assigned to each of the cluster of samples.
pooling_func(a, axis=None, dtype=None, out=None, keepdims=<no value>)
Compute the arithmetic mean along the specified axis.
Returns the average of the array elements. The average is taken over the flattened array by default, other-
wise over the specified axis. float64 intermediate and return values are used for integer inputs.
Parameters
a [array_like] Array containing numbers whose mean is desired. If a is not an array, a
conversion is attempted.
axis [None or int or tuple of ints, optional] Axis or axes along which the means are com-
puted. The default is to compute the mean of the flattened array.
New in version 1.7.0.
If this is a tuple of ints, a mean is performed over multiple axes, instead of a single axis or
all the axes as before.
dtype [data-type, optional] Type to use in computing the mean. For integer inputs, the
default is float64; for floating point inputs, it is the same as the input dtype.
out [ndarray, optional] Alternate output array in which to place the result. The default is
None; if provided, it must have the same shape as the expected output, but the type will
be cast if necessary. See doc.ufuncs for details.
keepdims [bool, optional] If this is set to True, the axes which are reduced are left in the
result as dimensions with size one. With this option, the result will broadcast correctly
against the input array.
If the default value is passed, then keepdims will not be passed through to the mean method
of sub-classes of ndarray, however any non-default value will be. If the sub-class’ method
does not implement keepdims any exceptions will be raised.
Returns
m [ndarray, see dtype parameter above] If out=None, returns a new array containing the
mean values, otherwise a reference to the output array is returned.
See also:
average Weighted average

std, var, nanmean, nanstd, nanvar
Notes
The arithmetic mean is the sum of the elements along the axis divided by the number of elements.
Note that for floating-point input, the mean is computed using the same precision the input has. Depending
on the input data, this can cause the results to be inaccurate, especially for float32 (see example below).
Specifying a higher-precision accumulator using the dtype keyword can alleviate this issue.
By default, float16 results are computed using float32 intermediates for extra precision.
Examples
>>> a = np.array([[1, 2], [3, 4]])

>>> np.mean(a)
2.5
>>> np.mean(a, axis=0)
array([ 2., 3.])
>>> np.mean(a, axis=1)
array([ 1.5, 3.5])
In single precision, mean can be inaccurate:
>>> a = np.zeros((2, 512*512), dtype=np.float32)

>>> a[0, :] = 1.0
>>> a[1, :] = 0.1
>>> np.mean(a)
0.54999924
Computing the mean in float64 is more accurate:
>>> np.mean(a, dtype=np.float64)

0.55000000074505806
of a nested object.
Returns
self
transform(X)
Transform a new matrix using the built clustering
Parameters
X [array-like, shape = [n_samples, n_features] or [n_features]] A M by N array of M obser-
vations in N dimensions or a length M array of M one-dimensional observations.
Returns
Y [array, shape = [n_samples, n_clusters] or [n_clusters]] The pooled values for each feature
cluster.

Examples using sklearn.cluster.FeatureAgglomeration
sklearn.cluster.KMeans
class sklearn.cluster.KMeans(n_clusters=8, init=’k-means++’, n_init=10, max_iter=300,

tol=0.0001, precompute_distances=’auto’, verbose=0, ran-
dom_state=None, copy_x=True, n_jobs=None, algorithm=’auto’)
K-Means clustering
Parameters
n_clusters [int, optional, default: 8] The number of clusters to form as well as the number of
centroids to generate.
init [{‘k-means++’, ‘random’ or an ndarray}] Method for initialization, defaults to ‘k-
means++’:
‘k-means++’ : selects initial cluster centers for k-mean clustering in a smart way to speed
up convergence. See section Notes in k_init for more details.
‘random’: choose k observations (rows) at random from data for the initial centroids.
If an ndarray is passed, it should be of shape (n_clusters, n_features) and gives the initial
centers.
n_init [int, default: 10] Number of time the k-means algorithm will be run with different cen-
troid seeds. The final results will be the best output of n_init consecutive runs in terms of
inertia.
max_iter [int, default: 300] Maximum number of iterations of the k-means algorithm for a
single run.
tol [float, default: 1e-4] Relative tolerance with regards to inertia to declare convergence
precompute_distances [{‘auto’, True, False}] Precompute distances (faster but takes more
memory).
‘auto’ : do not precompute distances if n_samples * n_clusters > 12 million. This corre-
sponds to about 100MB overhead per job using double precision.
True : always precompute distances
False : never precompute distances
verbose [int, default 0] Verbosity mode.
random_state [int, RandomState instance or None (default)] Determines random number gen-
eration for centroid initialization. Use an int to make the randomness deterministic. See
Glossary.
copy_x [boolean, optional] When pre-computing distances it is more numerically accurate to
center the data first. If copy_x is True (default), then the original data is not modified,
ensuring X is C-contiguous. If False, the original data is modified, and put back before the
function returns, but small numerical differences may be introduced by subtracting and then
adding the data mean, in this case it will also not ensure that data is C-contiguous which
may cause a significant slowdown.

n_jobs [int or None, optional (default=None)] The number of jobs to use for the computation.
This works by computing each of the n_init runs in parallel.
None means 1 unless in a joblib.parallel_backend context. -1 means using all
processors. See Glossary for more details.
algorithm [“auto”, “full” or “elkan”, default=”auto”] K-means algorithm to use. The classical
EM-style algorithm is “full”. The “elkan” variation is more efficient by using the triangle
inequality, but currently doesn’t support sparse data. “auto” chooses “elkan” for dense data
and “full” for sparse data.
Attributes
cluster_centers_ [array, [n_clusters, n_features]] Coordinates of cluster centers. If the algo-
rithm stops before fully converging (see tol and max_iter), these will not be consistent
with labels_.
labels_ : Labels of each point
inertia_ [float] Sum of squared distances of samples to their closest cluster center.
n_iter_ [int] Number of iterations run.
See also:
MiniBatchKMeans Alternative online implementation that does incremental updates of the centers positions
using mini-batches. For large scale learning (say n_samples > 10k) MiniBatchKMeans is probably much
faster than the default batch implementation.
Notes
The k-means problem is solved using either Lloyd’s or Elkan’s algorithm.

The average complexity is given by O(k n T), were n is the number of samples and T is the number of iteration.
The worst case complexity is given by O(n^(k+2/p)) with n = n_samples, p = n_features. (D. Arthur and S.
Vassilvitskii, ‘How slow is the k-means method?’ SoCG2006)
In practice, the k-means algorithm is very fast (one of the fastest clustering algorithms available), but it falls in
local minima. That’s why it can be useful to restart it several times.
If the algorithm stops before fully converging (because of tol or max_iter), labels_ and
cluster_centers_ will not be consistent, i.e. the cluster_centers_ will not be the means of the
points in each cluster. Also, the estimator will reassign labels_ after the last iteration to make labels_
consistent with predict on the training set.
Examples

>>> X = np.array([[1, 2], [1, 4], [1, 0],
... [10, 2], [10, 4], [10, 0]])
>>> kmeans = KMeans(n_clusters=2, random_state=0).fit(X)
>>> kmeans.labels_
array([1, 1, 1, 0, 0, 0], dtype=int32)
>>> kmeans.predict([[0, 0], [12, 3]])
array([1, 0], dtype=int32)
>>> kmeans.cluster_centers_

array([[10., 2.],
[ 1., 2.]])
Methods
fit(X[, y, sample_weight]) Compute k-means clustering.

fit_predict(X[, y, sample_weight]) Compute cluster centers and predict cluster index for
each sample.
fit_transform(X[, y, sample_weight]) Compute clustering and transform X to cluster-distance
space.
predict(X[, sample_weight]) Predict the closest cluster each sample in X belongs to.
score(X[, y, sample_weight]) Opposite of the value of X on the K-means objective.
transform(X) Transform X to a cluster-distance space.
__init__(n_clusters=8, init=’k-means++’, n_init=10, max_iter=300, tol=0.0001, precom-

pute_distances=’auto’, verbose=0, random_state=None, copy_x=True, n_jobs=None,
algorithm=’auto’)
Compute k-means clustering.
Parameters
X [array-like or sparse matrix, shape=(n_samples, n_features)] Training instances to cluster.
It must be noted that the data will be converted to C ordering, which will cause a memory
copy if the given data is not C-contiguous.
y [Ignored] not used, present here for API consistency by convention.
sample_weight [array-like, shape (n_samples,), optional] The weights for each observation
in X. If None, all observations are assigned equal weight (default: None)
Compute cluster centers and predict cluster index for each sample.
Convenience method; equivalent to calling fit(X) followed by predict(X).
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] New data to transform.
Returns
labels [array, shape [n_samples,]] Index of the cluster each sample belongs to.
fit_transform(X, y=None, sample_weight=None)
Compute clustering and transform X to cluster-distance space.
Equivalent to fit(X).transform(X), but more efficiently implemented.
Parameters


Returns
X_new [array, shape [n_samples, k]] X transformed in the new space.
Parameters
Returns
predict(X, sample_weight=None)
In the vector quantization literature, cluster_centers_ is called the code book and each value returned by
predict is the index of the closest code in the code book.
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] New data to predict.
Returns
score(X, y=None, sample_weight=None)
Opposite of the value of X on the K-means objective.
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] New data.
Returns
score [float] Opposite of the value of X on the K-means objective.
of a nested object.
Returns
self

transform(X)
Transform X to a cluster-distance space.
In the new space, each dimension is the distance to the cluster centers. Note that even if X is sparse, the
array returned by transform will typically be dense.
Parameters
Returns
Examples using sklearn.cluster.KMeans
• Demonstration of k-means assumptions

• Vector Quantization Example
• K-means Clustering
• Empirical evaluation of the impact of k-means initialization
• A demo of K-Means clustering on the handwritten digits data
• Comparison of the K-Means and MiniBatchKMeans clustering algorithms
• Selecting the number of clusters with silhouette analysis on KMeans clustering
sklearn.cluster.MiniBatchKMeans
class sklearn.cluster.MiniBatchKMeans(n_clusters=8, init=’k-means++’, max_iter=100,

batch_size=100, verbose=0, compute_labels=True,
random_state=None, tol=0.0, max_no_improvement=10,
init_size=None, n_init=3, reassignment_ratio=0.01)
Mini-Batch K-Means clustering
Parameters
n_clusters [int, optional, default: 8] The number of clusters to form as well as the number of
centroids to generate.
init [{‘k-means++’, ‘random’ or an ndarray}, default: ‘k-means++’] Method for initialization,
defaults to ‘k-means++’:
centers.
max_iter [int, optional] Maximum number of iterations over the complete dataset before stop-
ping independently of any early stopping criterion heuristics.

batch_size [int, optional, default: 100] Size of the mini batches.

verbose [boolean, optional] Verbosity mode.
compute_labels [boolean, default=True] Compute label assignment and inertia for the com-
plete dataset once the minibatch optimization has converged in fit.
eration for centroid initialization and random reassignment. Use an int to make the random-
ness deterministic. See Glossary.
tol [float, default: 0.0] Control early stopping based on the relative center changes as measured
by a smoothed, variance-normalized of the mean center squared position changes. This
early stopping heuristics is closer to the one used for the batch variant of the algorithms but
induces a slight computational and memory overhead over the inertia heuristic.
To disable convergence detection based on normalized center change, set tol to 0.0 (default).
max_no_improvement [int, default: 10] Control early stopping based on the consecutive num-
ber of mini batches that does not yield an improvement on the smoothed inertia.
To disable convergence detection based on inertia, set max_no_improvement to None.
init_size [int, optional, default: 3 * batch_size] Number of samples to randomly sample for
speeding up the initialization (sometimes at the expense of accuracy): the only algorithm
is initialized by running a batch KMeans on a random subset of the data. This needs to be
larger than n_clusters.
n_init [int, default=3] Number of random initializations that are tried. In contrast to KMeans,
the algorithm is only run once, using the best of the n_init initializations as measured by
inertia.
reassignment_ratio [float, default: 0.01] Control the fraction of the maximum number of
counts for a center to be reassigned. A higher value means that low count centers are more
easily reassigned, which means that the model will take longer to converge, but should con-
verge in a better clustering.
Attributes
cluster_centers_ [array, [n_clusters, n_features]] Coordinates of cluster centers
labels_ : Labels of each point (if compute_labels is set to True).
inertia_ [float] The value of the inertia criterion associated with the chosen partition (if com-
pute_labels is set to True). The inertia is defined as the sum of square distances of samples
to their nearest neighbor.
See also:
KMeans The classic implementation of the clustering method based on the Lloyd’s algorithm. It consumes the
whole set of input data at each iteration.
Notes
See http://www.eecs.tufts.edu/~dsculley/papers/fastkmeans.pdf

Examples
>>> from sklearn.cluster import MiniBatchKMeans

>>> X = np.array([[1, 2], [1, 4], [1, 0],
... [4, 2], [4, 0], [4, 4],
... [4, 5], [0, 1], [2, 2],
... [3, 2], [5, 5], [1, -1]])
>>> # manually fit on batches
>>> kmeans = MiniBatchKMeans(n_clusters=2,
... random_state=0,
... batch_size=6)
>>> kmeans = kmeans.partial_fit(X[0:6,:])
>>> kmeans = kmeans.partial_fit(X[6:12,:])
array([[1, 1],
[3, 4]])
>>> kmeans.predict([[0, 0], [4, 4]])
>>> # fit on the whole data
>>> kmeans = MiniBatchKMeans(n_clusters=2,
... random_state=0,
... batch_size=6,
... max_iter=10).fit(X)
array([[3.95918367, 2.40816327],
[1.12195122, 1.3902439 ]])
>>> kmeans.predict([[0, 0], [4, 4]])
Methods
fit(X[, y, sample_weight]) Compute the centroids on X by chunking it into mini-

batches.
fit_predict(X[, y, sample_weight]) Compute cluster centers and predict cluster index for
each sample.
fit_transform(X[, y, sample_weight]) Compute clustering and transform X to cluster-distance
space.
partial_fit(X[, y, sample_weight]) Update k means estimate on a single mini-batch X.
predict(X[, sample_weight]) Predict the closest cluster each sample in X belongs to.
score(X[, y, sample_weight]) Opposite of the value of X on the K-means objective.
transform(X) Transform X to a cluster-distance space.
__init__(n_clusters=8, init=’k-means++’, max_iter=100, batch_size=100, verbose=0,

compute_labels=True, random_state=None, tol=0.0, max_no_improvement=10,
init_size=None, n_init=3, reassignment_ratio=0.01)
Compute the centroids on X by chunking it into mini-batches.
Parameters

X [array-like or sparse matrix, shape=(n_samples, n_features)] Training instances to cluster.

It must be noted that the data will be converted to C ordering, which will cause a memory
Compute cluster centers and predict cluster index for each sample.
Convenience method; equivalent to calling fit(X) followed by predict(X).
Parameters
Returns
Compute clustering and transform X to cluster-distance space.
Equivalent to fit(X).transform(X), but more efficiently implemented.
Parameters
Returns
Parameters
Returns
partial_fit(X, y=None, sample_weight=None)
Update k means estimate on a single mini-batch X.
Parameters
X [array-like, shape = [n_samples, n_features]] Coordinates of the data points to cluster. It
must be noted that X will be copied if it is not C-contiguous.

predict(X, sample_weight=None)
In the vector quantization literature, cluster_centers_ is called the code book and each value returned by
predict is the index of the closest code in the code book.
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] New data to predict.
Returns
Opposite of the value of X on the K-means objective.
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] New data.
Returns
score [float] Opposite of the value of X on the K-means objective.
of a nested object.
Returns
self
transform(X)
Transform X to a cluster-distance space.
In the new space, each dimension is the distance to the cluster centers. Note that even if X is sparse, the
array returned by transform will typically be dense.
Parameters
Returns
Examples using sklearn.cluster.MiniBatchKMeans
• Biclustering documents with the Spectral Co-clustering algorithm


• Empirical evaluation of the impact of k-means initialization
sklearn.cluster.MeanShift
class sklearn.cluster.MeanShift(bandwidth=None, seeds=None, bin_seeding=False,

min_bin_freq=1, cluster_all=True, n_jobs=None)
Mean shift clustering using a flat kernel.
Mean shift clustering aims to discover “blobs” in a smooth density of samples. It is a centroid-based algo-
rithm, which works by updating candidates for centroids to be the mean of the points within a given region.
These candidates are then filtered in a post-processing stage to eliminate near-duplicates to form the final set of
centroids.
Seeding is performed using a binning technique for scalability.
Parameters
bandwidth [float, optional] Bandwidth used in the RBF kernel.
If not given, the bandwidth is estimated using sklearn.cluster.estimate_bandwidth; see the
documentation for that function for hints on scalability (see also the Notes, below).
seeds [array, shape=[n_samples, n_features], optional] Seeds used to initialize kernels. If not
set, the seeds are calculated by clustering.get_bin_seeds with bandwidth as the grid size and
default values for other parameters.
bin_seeding [boolean, optional] If true, initial kernel locations are not locations of all points,
but rather the location of the discretized version of points, where points are binned onto a
grid whose coarseness corresponds to the bandwidth. Setting this option to True will speed
up the algorithm because fewer seeds will be initialized. default value: False Ignored if
seeds argument is not None.
min_bin_freq [int, optional] To speed up the algorithm, accept only those bins with at least
min_bin_freq points as seeds. If not defined, set to 1.
cluster_all [boolean, default True] If true, then all points are clustered, even those orphans that
are not within any kernel. Orphans are assigned to the nearest kernel. If false, then orphans
are given cluster label -1.
Attributes
cluster_centers_ [array, [n_clusters, n_features]] Coordinates of cluster centers.
labels_ : Labels of each point.

Notes
Scalability:
Because this implementation uses a flat kernel and a Ball Tree to look up members of each kernel, the complexity
will tend towards O(T*n*log(n)) in lower dimensions, with n the number of samples and T the number of points.
In higher dimensions the complexity will tend towards O(T*n^2).
Scalability can be boosted by using fewer seeds, for example by using a higher value of min_bin_freq in the
get_bin_seeds function.
Note that the estimate_bandwidth function is much less scalable than the mean shift algorithm and will be the
bottleneck if it is used.
References
Dorin Comaniciu and Peter Meer, “Mean Shift: A robust approach toward feature space analysis”. IEEE Trans-
actions on Pattern Analysis and Machine Intelligence. 2002. pp. 603-619.
Examples
>>> from sklearn.cluster import MeanShift

>>> X = np.array([[1, 1], [2, 1], [1, 0],
... [4, 7], [3, 5], [3, 6]])
>>> clustering = MeanShift(bandwidth=2).fit(X)
array([1, 1, 1, 0, 0, 0])
>>> clustering.predict([[0, 0], [5, 5]])
array([1, 0])
>>> clustering
MeanShift(bandwidth=2, bin_seeding=False, cluster_all=True, min_bin_freq=1,
n_jobs=None, seeds=None)
Methods
fit(X[, y]) Perform clustering.

predict(X) Predict the closest cluster each sample in X belongs to.
__init__(bandwidth=None, seeds=None, bin_seeding=False, min_bin_freq=1, cluster_all=True,

n_jobs=None)
fit(X, y=None)
Perform clustering.
Parameters
X [array-like, shape=[n_samples, n_features]] Samples to cluster.
y [Ignored]

Parameters
Returns
Parameters
Returns
predict(X)
Parameters
X [{array-like, sparse matrix}, shape=[n_samples, n_features]] New data to predict.
Returns
of a nested object.
Returns
self
Examples using sklearn.cluster.MeanShift
• A demo of the mean-shift clustering algorithm

sklearn.cluster.SpectralClustering
class sklearn.cluster.SpectralClustering(n_clusters=8, eigen_solver=None, ran-

dom_state=None, n_init=10, gamma=1.0, affin-
ity=’rbf’, n_neighbors=10, eigen_tol=0.0, as-
sign_labels=’kmeans’, degree=3, coef0=1, ker-
nel_params=None, n_jobs=None)
Apply clustering to a projection to the normalized laplacian.

In practice Spectral Clustering is very useful when the structure of the individual clusters is highly non-convex
or more generally when a measure of the center and spread of the cluster is not a suitable description of the
complete cluster. For instance when clusters are nested circles on the 2D plan.
If affinity is the adjacency matrix of a graph, this method can be used to find normalized graph cuts.
When calling fit, an affinity matrix is constructed using either kernel function such the Gaussian (aka RBF)
kernel of the euclidean distanced d(X, X):
np.exp(-gamma * d(X,X) ** 2)
or a k-nearest neighbors connectivity matrix.

Alternatively, using precomputed, a user-provided affinity matrix can be used.
Parameters
n_clusters [integer, optional] The dimension of the projection subspace.
eigen_solver [{None, ‘arpack’, ‘lobpcg’, or ‘amg’}] The eigenvalue decomposition strategy to
use. AMG requires pyamg to be installed. It can be faster on very large, sparse problems,
but may also lead to instabilities
random_state [int, RandomState instance or None (default)] A pseudo random number gener-
ator used for the initialization of the lobpcg eigen vectors decomposition when eigen_solver
== ‘amg’ and by the K-Means initialization. Use an int to make the randomness determin-
istic. See Glossary.
n_init [int, optional, default: 10] Number of time the k-means algorithm will be run with dif-
ferent centroid seeds. The final results will be the best output of n_init consecutive runs in
terms of inertia.
gamma [float, default=1.0] Kernel coefficient for rbf, poly, sigmoid, laplacian and chi2 kernels.
Ignored for affinity='nearest_neighbors'.
affinity [string, array-like or callable, default ‘rbf’] If a string, this may be one
of ‘nearest_neighbors’, ‘precomputed’, ‘rbf’ or one of the kernels supported by
sklearn.metrics.pairwise_kernels.
Only kernels that produce similarity scores (non-negative values that increase with similar-
ity) should be used. This property is not checked by the clustering algorithm.
n_neighbors [integer] Number of neighbors to use when constructing the affinity matrix using
the nearest neighbors method. Ignored for affinity='rbf'.
eigen_tol [float, optional, default: 0.0] Stopping criterion for eigendecomposition of the Lapla-
cian matrix when using arpack eigen_solver.
assign_labels [{‘kmeans’, ‘discretize’}, default: ‘kmeans’] The strategy to use to assign labels
in the embedding space. There are two ways to assign labels after the laplacian embedding.
k-means can be applied and is a popular choice. But it can also be sensitive to initialization.
Discretization is another approach which is less sensitive to random initialization.
degree [float, default=3] Degree of the polynomial kernel. Ignored by other kernels.
coef0 [float, default=1] Zero coefficient for polynomial and sigmoid kernels. Ignored by other
kernels.
kernel_params [dictionary of string to any, optional] Parameters (keyword arguments) and
values for kernel passed as callable object. Ignored by other kernels.

Attributes
affinity_matrix_ [array-like, shape (n_samples, n_samples)] Affinity matrix used for cluster-
ing. Available only if after calling fit.
labels_ : Labels of each point
Notes
If you have an affinity matrix, such as a distance matrix, for which 0 means identical elements, and high values
means very dissimilar elements, it can be transformed in a similarity matrix that is well suited for the algorithm
by applying the Gaussian (RBF, heat) kernel:
np.exp(- dist_matrix ** 2 / (2. * delta ** 2))
Where delta is a free parameter representing the width of the Gaussian kernel.
Another alternative is to take a symmetric version of the k nearest neighbors connectivity matrix of the points.
If the pyamg package is installed, it is used: this greatly speeds up computation.
References
• Normalized cuts and image segmentation, 2000 Jianbo Shi, Jitendra Malik http://citeseer.ist.psu.edu/
viewdoc/summary?doi=10.1.1.160.2324
• A Tutorial on Spectral Clustering, 2007 Ulrike von Luxburg http://citeseerx.ist.psu.edu/viewdoc/
summary?doi=10.1.1.165.9323
• Multiclass spectral clustering, 2003 Stella X. Yu, Jianbo Shi http://www1.icsi.berkeley.edu/~stellayu/
publication/doc/2003kwayICCV.pdf
Examples
>>> from sklearn.cluster import SpectralClustering

>>> X = np.array([[1, 1], [2, 1], [1, 0],
... [4, 7], [3, 5], [3, 6]])
>>> clustering = SpectralClustering(n_clusters=2,
... assign_labels="discretize",
... random_state=0).fit(X)
array([1, 1, 1, 0, 0, 0])
>>> clustering
SpectralClustering(affinity='rbf', assign_labels='discretize', coef0=1,
degree=3, eigen_solver=None, eigen_tol=0.0, gamma=1.0,
kernel_params=None, n_clusters=2, n_init=10, n_jobs=None,
n_neighbors=10, random_state=0)

Methods
fit(X[, y]) Creates an affinity matrix for X using the selected affin-
ity, then applies spectral clustering to this affinity ma-
trix.
__init__(n_clusters=8, eigen_solver=None, random_state=None, n_init=10, gamma=1.0, affin-

ity=’rbf’, n_neighbors=10, eigen_tol=0.0, assign_labels=’kmeans’, degree=3, coef0=1,
kernel_params=None, n_jobs=None)
fit(X, y=None)
Creates an affinity matrix for X using the selected affinity, then applies spectral clustering to this affinity
matrix.
Parameters
X [array-like or sparse matrix, shape (n_samples, n_features)] OR, if affin-
ity==‘precomputed‘, a precomputed affinity matrix of shape (n_samples, n_samples)
y [Ignored]
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self

Examples using sklearn.cluster.SpectralClustering
6.3.2 Functions
cluster.affinity_propagation(S[, . . . ]) Perform Affinity Propagation Clustering of data

cluster.dbscan(X[, eps, min_samples, . . . ]) Perform DBSCAN clustering from vector array or distance
matrix.
cluster.estimate_bandwidth(X[, quantile, . . . ]) Estimate the bandwidth to use with the mean-shift algo-
rithm.
cluster.k_means(X, n_clusters[, . . . ]) K-means clustering algorithm.
cluster.mean_shift(X[, bandwidth, seeds, . . . ]) Perform mean shift clustering of data using a flat kernel.
cluster.spectral_clustering(affinity[, . . . ]) Apply clustering to a projection to the normalized lapla-
cian.
cluster.ward_tree(X[, connectivity, . . . ]) Ward clustering based on a Feature matrix.
sklearn.cluster.affinity_propagation
sklearn.cluster.affinity_propagation(S, preference=None, convergence_iter=15,

max_iter=200, damping=0.5, copy=True, ver-
bose=False, return_n_iter=False)
Perform Affinity Propagation Clustering of data
Parameters
S [array-like, shape (n_samples, n_samples)] Matrix of similarities between points
preference [array-like, shape (n_samples,) or float, optional] Preferences for each point - points
with larger values of preferences are more likely to be chosen as exemplars. The number of
exemplars, i.e. of clusters, is influenced by the input preferences value. If the preferences
are not passed as arguments, they will be set to the median of the input similarities (resulting
in a moderate number of clusters). For a smaller amount of clusters, this can be set to the
minimum value of the similarities.
convergence_iter [int, optional, default: 15] Number of iterations with no change in the number
of estimated clusters that stops the convergence.
max_iter [int, optional, default: 200] Maximum number of iterations
damping [float, optional, default: 0.5] Damping factor between 0.5 and 1.
copy [boolean, optional, default: True] If copy is False, the affinity matrix is modified inplace
by the algorithm, for memory efficiency
verbose [boolean, optional, default: False] The verbosity level
return_n_iter [bool, default False] Whether or not to return the number of iterations.
Returns
cluster_centers_indices [array, shape (n_clusters,)] index of clusters centers
labels [array, shape (n_samples,)] cluster labels for each point
n_iter [int] number of iterations run. Returned only if return_n_iter is set to True.

Notes
For an example, see examples/cluster/plot_affinity_propagation.py.

When the algorithm does not converge, it returns an empty array as cluster_center_indices and -1 as
label for each training sample.
When all training samples have equal similarities and equal preferences, the assignment of cluster centers and
labels depends on the preference. If the preference is smaller than the similarities, a single cluster center and
label 0 for every sample will be returned. Otherwise, every training sample becomes its own cluster center and
is assigned a unique label.
References
Brendan J. Frey and Delbert Dueck, “Clustering by Passing Messages Between Data Points”, Science Feb. 2007
Examples using sklearn.cluster.affinity_propagation
• Visualizing the stock market structure
sklearn.cluster.dbscan
sklearn.cluster.dbscan(X, eps=0.5, min_samples=5, metric=’minkowski’, metric_params=None, al-

gorithm=’auto’, leaf_size=30, p=2, sample_weight=None, n_jobs=None)
Perform DBSCAN clustering from vector array or distance matrix.
Parameters
eps [float, optional] The maximum distance between two samples for them to be considered as
in the same neighborhood.
min_samples [int, optional] The number of samples (or total weight) in a neighborhood for a
point to be considered as a core point. This includes the point itself.
a feature array. If metric is a string or callable, it must be one of the options allowed
by sklearn.metrics.pairwise_distances for its metric parameter. If metric
is “precomputed”, X is assumed to be a distance matrix and must be square. X may be
a sparse matrix, in which case only “nonzero” elements may be considered neighbors for
DBSCAN.
metric_params [dict, optional] Additional keyword arguments for the metric function.
algorithm [{‘auto’, ‘ball_tree’, ‘kd_tree’, ‘brute’}, optional] The algorithm to be used by the
NearestNeighbors module to compute pointwise distances and find nearest neighbors. See
NearestNeighbors module documentation for details.

leaf_size [int, optional (default = 30)] Leaf size passed to BallTree or cKDTree. This can affect
p [float, optional] The power of the Minkowski metric to be used to calculate distance between
points.
sample_weight [array, shape (n_samples,), optional] Weight of each sample, such that a sample
with a weight of at least min_samples is by itself a core sample; a sample with negative
weight may inhibit its eps-neighbor from being core. Note that weights are absolute, and
default to 1.
n_jobs [int or None, optional (default=None)] The number of parallel jobs to run for neighbors
search. None means 1 unless in a joblib.parallel_backend context. -1 means
using all processors. See Glossary for more details.
Returns
core_samples [array [n_core_samples]] Indices of core samples.
labels [array [n_samples]] Cluster labels for each point. Noisy samples are given the label -1.
See also:
DBSCAN An estimator interface for this clustering algorithm.
Notes
For an example, see examples/cluster/plot_dbscan.py.

This implementation bulk-computes all neighborhood queries, which increases the memory complexity to
O(n.d) where d is the average number of neighbors, while original DBSCAN had memory complexity O(n).
It may attract a higher memory complexity when querying these nearest neighborhoods, depending on the
algorithm.
One way to avoid the query complexity is to pre-compute sparse neighborhoods in chunks
using NearestNeighbors.radius_neighbors_graph with mode='distance', then using
metric='precomputed' here.
Another way to reduce memory and computation time is to remove (near-)duplicate points and use
sample_weight instead.
References
Ester, M., H. P. Kriegel, J. Sander, and X. Xu, “A Density-Based Algorithm for Discovering Clusters in Large
Spatial Databases with Noise”. In: Proceedings of the 2nd International Conference on Knowledge Discovery
and Data Mining, Portland, OR, AAAI Press, pp. 226-231. 1996
sklearn.cluster.estimate_bandwidth
sklearn.cluster.estimate_bandwidth(X, quantile=0.3, n_samples=None, random_state=0,

n_jobs=None)
Estimate the bandwidth to use with the mean-shift algorithm.
That this function takes time at least quadratic in n_samples. For large datasets, it’s wise to set that parameter
to a small value.

Parameters
X [array-like, shape=[n_samples, n_features]] Input points.
quantile [float, default 0.3] should be between [0, 1] 0.5 means that the median of all pairwise
distances is used.
n_samples [int, optional] The number of samples to use. If not given, all samples are used.
random_state [int, RandomState instance or None (default)] The generator used to randomly
select the samples from input points for bandwidth estimation. Use an int to make the
randomness deterministic. See Glossary.
Returns
bandwidth [float] The bandwidth parameter.
Examples using sklearn.cluster.estimate_bandwidth

sklearn.cluster.k_means
sklearn.cluster.k_means(X, n_clusters, sample_weight=None, init=’k-means++’, precom-

pute_distances=’auto’, n_init=10, max_iter=300, verbose=False,
tol=0.0001, random_state=None, copy_x=True, n_jobs=None, algo-
rithm=’auto’, return_n_iter=False)
K-means clustering algorithm.
Parameters
X [array-like or sparse matrix, shape (n_samples, n_features)] The observations to cluster. It
must be noted that the data will be converted to C ordering, which will cause a memory
n_clusters [int] The number of clusters to form as well as the number of centroids to generate.
sample_weight [array-like, shape (n_samples,), optional] The weights for each observation in
X. If None, all observations are assigned equal weight (default: None)
init [{‘k-means++’, ‘random’, or ndarray, or a callable}, optional] Method for initialization,
default to ‘k-means++’:
centers.
If a callable is passed, it should take arguments X, k and and a random state and return an
initialization.

precompute_distances [{‘auto’, True, False}] Precompute distances (faster but takes more
memory).
‘auto’ : do not precompute distances if n_samples * n_clusters > 12 million. This corre-
sponds to about 100MB overhead per job using double precision.
True : always precompute distances
False : never precompute distances
terms of inertia.
max_iter [int, optional, default 300] Maximum number of iterations of the k-means algorithm
to run.
verbose [boolean, optional] Verbosity mode.
tol [float, optional] The relative increment in the results before declaring convergence.
eration for centroid initialization. Use an int to make the randomness deterministic. See
Glossary.
copy_x [boolean, optional] When pre-computing distances it is more numerically accurate to
center the data first. If copy_x is True (default), then the original data is not modified,
ensuring X is C-contiguous. If False, the original data is modified, and put back before the
function returns, but small numerical differences may be introduced by subtracting and then
adding the data mean, in this case it will also not ensure that data is C-contiguous which
may cause a significant slowdown.
algorithm [“auto”, “full” or “elkan”, default=”auto”] K-means algorithm to use. The classical
EM-style algorithm is “full”. The “elkan” variation is more efficient by using the triangle
inequality, but currently doesn’t support sparse data. “auto” chooses “elkan” for dense data
and “full” for sparse data.
return_n_iter [bool, optional] Whether or not to return the number of iterations.
Returns
centroid [float ndarray with shape (k, n_features)] Centroids found at the last iteration of k-
means.
label [integer ndarray with shape (n_samples,)] label[i] is the code or index of the centroid the
i’th observation is closest to.
inertia [float] The final value of the inertia criterion (sum of squared distances to the closest
centroid for all observations in the training set).
best_n_iter [int] Number of iterations corresponding to the best results. Returned only if re-
turn_n_iter is set to True.

sklearn.cluster.mean_shift
sklearn.cluster.mean_shift(X, bandwidth=None, seeds=None, bin_seeding=False,

min_bin_freq=1, cluster_all=True, max_iter=300, n_jobs=None)
Perform mean shift clustering of data using a flat kernel.
Parameters
X [array-like, shape=[n_samples, n_features]] Input data.
bandwidth [float, optional] Kernel bandwidth.
If bandwidth is not given, it is determined using a heuristic based on the median of
all pairwise distances. This will take quadratic time in the number of samples. The
sklearn.cluster.estimate_bandwidth function can be used to do this more efficiently.
seeds [array-like, shape=[n_seeds, n_features] or None] Point used as initial kernel loca-
tions. If None and bin_seeding=False, each data point is used as a seed. If None and
bin_seeding=True, see bin_seeding.
bin_seeding [boolean, default=False] If true, initial kernel locations are not locations of all
points, but rather the location of the discretized version of points, where points are binned
onto a grid whose coarseness corresponds to the bandwidth. Setting this option to True will
speed up the algorithm because fewer seeds will be initialized. Ignored if seeds argument is
not None.
min_bin_freq [int, default=1] To speed up the algorithm, accept only those bins with at least
min_bin_freq points as seeds.
cluster_all [boolean, default True] If true, then all points are clustered, even those orphans that
are not within any kernel. Orphans are assigned to the nearest kernel. If false, then orphans
are given cluster label -1.
max_iter [int, default 300] Maximum number of iterations, per seed point before the clustering
operation terminates (for that seed point), if has not converged yet.
New in version 0.17: Parallel Execution using n_jobs.
Returns
cluster_centers [array, shape=[n_clusters, n_features]] Coordinates of cluster centers.
labels [array, shape=[n_samples]] Cluster labels for each point.
Notes
For an example, see examples/cluster/plot_mean_shift.py.

sklearn.cluster.spectral_clustering
sklearn.cluster.spectral_clustering(affinity, n_clusters=8, n_components=None,

eigen_solver=None, random_state=None, n_init=10,
eigen_tol=0.0, assign_labels=’kmeans’)
Apply clustering to a projection to the normalized laplacian.
In practice Spectral Clustering is very useful when the structure of the individual clusters is highly non-convex
or more generally when a measure of the center and spread of the cluster is not a suitable description of the
complete cluster. For instance when clusters are nested circles on the 2D plan.
If affinity is the adjacency matrix of a graph, this method can be used to find normalized graph cuts.
Parameters
affinity [array-like or sparse matrix, shape: (n_samples, n_samples)] The affinity matrix de-
scribing the relationship of the samples to embed. Must be symmetric.
Possible examples:
• adjacency matrix of a graph,
• heat kernel of the pairwise distance matrix of the samples,
• symmetric k-nearest neighbours connectivity matrix of the samples.
n_clusters [integer, optional] Number of clusters to extract.
n_components [integer, optional, default is n_clusters] Number of eigen vectors to use for the
spectral embedding
but may also lead to instabilities
random_state [int, RandomState instance or None (default)] A pseudo random number gener-
ator used for the initialization of the lobpcg eigen vectors decomposition when eigen_solver
== ‘amg’ and by the K-Means initialization. Use an int to make the randomness determin-
istic. See Glossary.
terms of inertia.
eigen_tol [float, optional, default: 0.0] Stopping criterion for eigendecomposition of the Lapla-
assign_labels [{‘kmeans’, ‘discretize’}, default: ‘kmeans’] The strategy to use to assign labels
in the embedding space. There are two ways to assign labels after the laplacian embedding.
k-means can be applied and is a popular choice. But it can also be sensitive to initialization.
Discretization is another approach which is less sensitive to random initialization. See the
‘Multiclass spectral clustering’ paper referenced below for more details on the discretization
approach.
Returns
labels [array of integers, shape: n_samples] The labels of the clusters.

Notes
The graph should contain only one connect component, elsewhere the results make little sense.
This algorithm solves the normalized cut for k=2: it is a normalized spectral clustering.
References
summary?doi=10.1.1.165.9323
• Multiclass spectral clustering, 2003 Stella X. Yu, Jianbo Shi http://www1.icsi.berkeley.edu/~stellayu/
publication/doc/2003kwayICCV.pdf
Examples using sklearn.cluster.spectral_clustering
• Segmenting the picture of greek coins in regions

• Spectral clustering for image segmentation
sklearn.cluster.ward_tree
sklearn.cluster.ward_tree(X, connectivity=None, n_clusters=None, return_distance=False)

Ward clustering based on a Feature matrix.
Recursively merges the pair of clusters that minimally increases within-cluster variance.
The inertia matrix uses a Heapq-based representation.
This is the structured version, that takes into account some topological structure between samples.
Parameters
X [array, shape (n_samples, n_features)] feature matrix representing n_samples samples to be
clustered
connectivity [sparse matrix (optional).] connectivity matrix. Defines for each sample the neigh-
boring samples following a given structure of the data. The matrix is assumed to be sym-
metric and only the upper triangular half is used. Default is None, i.e, the Ward algorithm is
unstructured.
n_clusters [int (optional)] Stop early the construction of the tree at n_clusters. This is useful
to decrease computation time if the number of clusters is not small compared to the number
of samples. In this case, the complete tree is not computed, thus the ‘children’ output is of
limited use, and the ‘parents’ output should rather be used. This option is valid only when
specifying a connectivity matrix.
return_distance [bool (optional)] If True, return the distance between the clusters.
Returns

children [2D array, shape (n_nodes-1, 2)] The children of each non-leaf node. Values less than
n_samples correspond to leaves of the tree which are the original samples. A node i greater
than or equal to n_samples is a non-leaf node and has children children_[i - n_samples].
Alternatively at the i-th iteration, children[i][0] and children[i][1] are merged to form node
n_samples + i
n_components [int] The number of connected components in the graph.
n_leaves [int] The number of leaves in the tree
parents [1D array, shape (n_nodes, ) or None] The parent of each node. Only returned when a
connectivity matrix is specified, elsewhere ‘None’ is returned.
distances [1D array, shape (n_nodes-1, )] Only returned if return_distance is set to True (for
compatibility). The distances between the centers of the nodes. distances[i] corresponds
to a weighted euclidean distance between the nodes children[i, 1] and children[i, 2]. If the
nodes refer to leaves of the tree, then distances[i] is their unweighted euclidean distance.
Distances are updated in the following way (from scipy.hierarchy.linkage):
The new entry 𝑑(𝑢, 𝑣) is computed as follows,
√︂
|𝑣| + |𝑠| |𝑣| + |𝑡| |𝑣|
𝑑(𝑢, 𝑣) = 𝑑(𝑣, 𝑠)2 + 𝑑(𝑣, 𝑡)2 − 𝑑(𝑠, 𝑡)2
𝑇 𝑇 𝑇
where 𝑢 is the newly joined cluster consisting of clusters 𝑠 and 𝑡, 𝑣 is an unused cluster in
the forest, 𝑇 = |𝑣| + |𝑠| + |𝑡|, and | * | is the cardinality of its argument. This is also known
as the incremental algorithm.
6.4 sklearn.cluster.bicluster: Biclustering
Spectral biclustering algorithms.

Authors : Kemal Eren License: BSD 3 clause
User guide: See the Biclustering section for further details.
6.4.1 Classes
SpectralBiclustering([n_clusters, method, . . . ]) Spectral biclustering (Kluger, 2003).

SpectralCoclustering([n_clusters, . . . ]) Spectral Co-Clustering algorithm (Dhillon, 2001).
sklearn.cluster.bicluster.SpectralBiclustering
class sklearn.cluster.bicluster.SpectralBiclustering(n_clusters=3,
method=’bistochastic’,
n_components=6, n_best=3,
svd_method=’randomized’,
n_svd_vecs=None,
mini_batch=False, init=’k-
means++’, n_init=10,
n_jobs=None, ran-
dom_state=None)
Spectral biclustering (Kluger, 2003).
Partitions rows and columns under the assumption that the data has an underlying checkerboard structure. For
6.4. sklearn.cluster.bicluster: Biclustering 1443

instance, if there are two row partitions and three column partitions, each row will belong to three biclusters,
and each column will belong to two biclusters. The outer product of the corresponding row and column label
vectors gives this checkerboard structure.
Parameters
n_clusters [integer or tuple (n_row_clusters, n_column_clusters)] The number of row and col-
umn clusters in the checkerboard structure.
method [string, optional, default: ‘bistochastic’] Method of normalizing and converting sin-
gular vectors into biclusters. May be one of ‘scale’, ‘bistochastic’, or ‘log’. The authors
recommend using ‘log’. If the data is sparse, however, log normalization will not work,
which is why the default is ‘bistochastic’. CAUTION: if method=’log’, the data must not
be sparse.
n_components [integer, optional, default: 6] Number of singular vectors to check.
n_best [integer, optional, default: 3] Number of best singular vectors to which to project the
data for clustering.
svd_method [string, optional, default: ‘randomized’] Selects the algorithm for find-
ing singular vectors. May be ‘randomized’ or ‘arpack’. If ‘randomized’, uses
sklearn.utils.extmath.randomized_svd, which may be faster for large matrices. If ‘arpack’,
uses scipy.sparse.linalg.svds, which is more accurate, but possibly slower in some cases.
n_svd_vecs [int, optional, default: None] Number of vectors to use in calculating the SVD. Cor-
responds to ncv when svd_method=arpack and n_oversamples when svd_method is ‘ran-
domized‘.
mini_batch [bool, optional, default: False] Whether to use mini-batch k-means, which is faster
but may get different results.
init [{‘k-means++’, ‘random’ or an ndarray}] Method for initialization of k-means algorithm;
defaults to ‘k-means++’.
n_init [int, optional, default: 10] Number of random initializations that are tried with the k-
means algorithm.
If mini-batch k-means is used, the best initialization is chosen and the algorithm runs once.
Otherwise, the algorithm is run for each initialization and the best solution chosen.
This works by breaking down the pairwise matrix into n_jobs even slices and computing
them in parallel.
random_state [int, RandomState instance or None (default)] Used for randomizing the singular
value decomposition and the k-means initialization. Use an int to make the randomness
deterministic. See Glossary.
Attributes
rows_ [array-like, shape (n_row_clusters, n_rows)] Results of the clustering. rows[i, r] is True
if cluster i contains row r. Available only after calling fit.
columns_ [array-like, shape (n_column_clusters, n_columns)] Results of the clustering, like
rows.
row_labels_ [array-like, shape (n_rows,)] Row partition labels.

column_labels_ [array-like, shape (n_cols,)] Column partition labels.
References
• Kluger, Yuval, et. al., 2003. Spectral biclustering of microarray data: coclustering genes and conditions.
Examples
>>> from sklearn.cluster import SpectralBiclustering

>>> X = np.array([[1, 1], [2, 1], [1, 0],
... [4, 7], [3, 5], [3, 6]])
>>> clustering = SpectralBiclustering(n_clusters=2, random_state=0).fit(X)
>>> clustering.row_labels_
array([1, 1, 1, 0, 0, 0], dtype=int32)
>>> clustering.column_labels_
>>> clustering
SpectralBiclustering(init='k-means++', method='bistochastic',
mini_batch=False, n_best=3, n_clusters=2, n_components=6,
n_init=10, n_jobs=None, n_svd_vecs=None, random_state=0,
svd_method='randomized')
Methods
fit(X[, y]) Creates a biclustering for X.

__init__(n_clusters=3, method=’bistochastic’, n_components=6, n_best=3,

svd_method=’randomized’, n_svd_vecs=None, mini_batch=False, init=’k-means++’,
n_init=10, n_jobs=None, random_state=None)
biclusters_
fit(X, y=None)
Creates a biclustering for X.
Parameters
X [array-like, shape (n_samples, n_features)]
y [Ignored]
get_indices(i)

Parameters
Returns
ter.
Parameters
Returns
get_shape(i)
Parameters
Returns
Parameters
Returns
Notes
of a nested object.
Returns
self

sklearn.cluster.bicluster.SpectralCoclustering
class sklearn.cluster.bicluster.SpectralCoclustering(n_clusters=3,
svd_method=’randomized’,
n_svd_vecs=None,
mini_batch=False, init=’k-
means++’, n_init=10,
n_jobs=None, ran-
dom_state=None)
Spectral Co-Clustering algorithm (Dhillon, 2001).
Clusters rows and columns of an array X to solve the relaxed normalized cut of the bipartite graph created from
X as follows: the edge between row vertex i and column vertex j has weight X[i, j].
The resulting bicluster structure is block-diagonal, since each row and each column belongs to exactly one
bicluster.
Supports sparse matrices, as long as they are nonnegative.
Parameters
n_clusters [integer, optional, default: 3] The number of biclusters to find.
svd_method [string, optional, default: ‘randomized’] Selects the algorithm for finding singu-
lar vectors. May be ‘randomized’ or ‘arpack’. If ‘randomized’, use sklearn.utils.
extmath.randomized_svd, which may be faster for large matrices. If ‘arpack’, use
scipy.sparse.linalg.svds, which is more accurate, but possibly slower in some
cases.
n_svd_vecs [int, optional, default: None] Number of vectors to use in calculating the SVD. Cor-
responds to ncv when svd_method=arpack and n_oversamples when svd_method is ‘ran-
domized‘.
mini_batch [bool, optional, default: False] Whether to use mini-batch k-means, which is faster
but may get different results.
init [{‘k-means++’, ‘random’ or an ndarray}] Method for initialization of k-means algorithm;
defaults to ‘k-means++’.
n_init [int, optional, default: 10] Number of random initializations that are tried with the k-
means algorithm.
If mini-batch k-means is used, the best initialization is chosen and the algorithm runs once.
Otherwise, the algorithm is run for each initialization and the best solution chosen.
them in parallel.
random_state [int, RandomState instance or None (default)] Used for randomizing the singular
value decomposition and the k-means initialization. Use an int to make the randomness
deterministic. See Glossary.
Attributes
rows_ [array-like, shape (n_row_clusters, n_rows)] Results of the clustering. rows[i, r] is True
if cluster i contains row r. Available only after calling fit.

columns_ [array-like, shape (n_column_clusters, n_columns)] Results of the clustering, like

rows.
row_labels_ [array-like, shape (n_rows,)] The bicluster label of each row.
column_labels_ [array-like, shape (n_cols,)] The bicluster label of each column.
References
• Dhillon, Inderjit S, 2001. Co-clustering documents and words using bipartite spectral graph partitioning.
Examples
>>> from sklearn.cluster import SpectralCoclustering

>>> X = np.array([[1, 1], [2, 1], [1, 0],
... [4, 7], [3, 5], [3, 6]])
>>> clustering = SpectralCoclustering(n_clusters=2, random_state=0).fit(X)
>>> clustering.row_labels_
array([0, 1, 1, 0, 0, 0], dtype=int32)
>>> clustering.column_labels_
>>> clustering
SpectralCoclustering(init='k-means++', mini_batch=False, n_clusters=2,
n_init=10, n_jobs=None, n_svd_vecs=None, random_state=0,
svd_method='randomized')
Methods
fit(X[, y]) Creates a biclustering for X.

__init__(n_clusters=3, svd_method=’randomized’, n_svd_vecs=None, mini_batch=False, init=’k-

means++’, n_init=10, n_jobs=None, random_state=None)
biclusters_
fit(X, y=None)
Creates a biclustering for X.
Parameters
y [Ignored]
get_indices(i)


Parameters
Returns
ter.
Parameters
Returns
get_shape(i)
Parameters
Returns
Parameters
Returns
Notes
of a nested object.
Returns
self

6.5 sklearn.compose: Composite Estimators
Meta-estimators for building composite models with transformers

In addition to its current contents, this module will eventually be home to refurbished versions of Pipeline and Fea-
tureUnion.
User guide: See the Pipelines and composite estimators section for further details.
compose.ColumnTransformer(transformers[, . . . ]) Applies transformers to columns of an array or pandas

DataFrame.
compose.TransformedTargetRegressor([. . . ]) Meta-estimator to regress on a transformed target.
6.5.1 sklearn.compose.ColumnTransformer
class sklearn.compose.ColumnTransformer(transformers, remainder=’drop’,

sparse_threshold=0.3, n_jobs=None, trans-
former_weights=None)
Applies transformers to columns of an array or pandas DataFrame.
EXPERIMENTAL: some behaviors may change between releases without deprecation.
This estimator allows different columns or column subsets of the input to be transformed separately and the
features generated by each transformer will be concatenated to form a single feature space. This is useful for
heterogeneous or columnar data, to combine several feature extraction mechanisms or transformations into a
single transformer.
Parameters
transformers [list of tuples] List of (name, transformer, column(s)) tuples specifying the trans-
former objects to be applied to subsets of the data.
name [string] Like in Pipeline and FeatureUnion, this allows the transformer and its pa-
rameters to be set using set_params and searched in grid search.
transformer [estimator or {‘passthrough’, ‘drop’}] Estimator must support fit and trans-
form. Special-cased strings ‘drop’ and ‘passthrough’ are accepted as well, to indicate to
drop the columns or to pass them through untransformed, respectively.
column(s) [string or int, array-like of string or int, slice, boolean mask array or callable]
Indexes the data on its second axis. Integers are interpreted as positional columns, while
strings can reference DataFrame columns by name. A scalar string or int should be used
where transformer expects X to be a 1d array-like (vector), otherwise a 2d array will
be passed to the transformer. A callable is passed the input data X and can return any of
the above.
remainder [{‘drop’, ‘passthrough’} or estimator, default ‘drop’] By default, only the
specified columns in transformers are transformed and combined in the output, and
the non-specified columns are dropped. (default of 'drop'). By specifying
remainder='passthrough', all remaining columns that were not specified in trans-
formers will be automatically passed through. This subset of columns is concatenated with
the output of the transformers. By setting remainder to be an estimator, the remaining
non-specified columns will use the remainder estimator. The estimator must support fit
and transform.

sparse_threshold [float, default = 0.3] If the output of the different transfromers contains sparse
matrices, these will be stacked as a sparse matrix if the overall density is lower than this
value. Use sparse_threshold=0 to always return dense. When the transformed output
consists of all dense data, the stacked result will be dense, and this keyword will be ignored.
n_jobs [int or None, optional (default=None)] Number of jobs to run in parallel. None means 1
transformer_weights [dict, optional] Multiplicative weights for features per transformer. The
output of the transformer is multiplied by these weights. Keys are transformer names, values
the weights.
Attributes
transformers_ [list] The collection of fitted transformers as tuples of (name, fitted_transformer,
column). fitted_transformer can be an estimator, ‘drop’, or ‘passthrough’. In case there
were no columns selected, this will be the unfitted transformer. If there are remain-
ing columns, the final element is a tuple of the form: (‘remainder’, transformer, re-
maining_columns) corresponding to the remainder parameter. If there are remain-
ing columns, then len(transformers_)==len(transformers)+1, otherwise
len(transformers_)==len(transformers).
named_transformers_ [Bunch object, a dictionary with attribute access] Access the fitted
transformer by name.
sparse_output_ [boolean] Boolean flag indicating wether the output of transform is a sparse
matrix or a dense numpy array, which depends on the output of the individual transformers
and the sparse_threshold keyword.
See also:
sklearn.compose.make_column_transformer convenience function for combining the outputs of

multiple transformer objects applied to column subsets of the original feature space.
Notes
The order of the columns in the transformed feature matrix follows the order of how the columns are specified
in the transformers list. Columns of the original feature matrix that are not specified are dropped from the
resulting transformed feature matrix, unless specified in the passthrough keyword. Those columns specified
with passthrough are added at the right to the output of the transformers.
Examples
>>> from sklearn.compose import ColumnTransformer

>>> from sklearn.preprocessing import Normalizer
>>> ct = ColumnTransformer(
... [("norm1", Normalizer(norm='l1'), [0, 1]),
... ("norm2", Normalizer(norm='l1'), slice(2, 4))])
>>> X = np.array([[0., 1., 2., 2.],
... [1., 1., 0., 1.]])
>>> # Normalizer scales each row of X to unit norm. A separate scaling
>>> # is applied for the two first and two last elements of each
>>> # row independently.
>>> ct.fit_transform(X)
6.5. sklearn.compose: Composite Estimators 1451

array([[0. , 1. , 0.5, 0.5],

[0.5, 0.5, 0. , 1. ]])
Methods
fit(X[, y]) Fit all transformers using X.

fit_transform(X[, y]) Fit all transformers, transform the data and concatenate
results.
get_feature_names() Get feature names from all transformers.
set_params(**kwargs) Set the parameters of this estimator.
transform(X) Transform X separately by each transformer, concate-
nate results.
__init__(transformers, remainder=’drop’, sparse_threshold=0.3, n_jobs=None, trans-

fit(X, y=None)
Fit all transformers using X.
Parameters
X [array-like or DataFrame of shape [n_samples, n_features]] Input data, of which specified
subsets are used to fit the transformers.
y [array-like, shape (n_samples, . . . ), optional] Targets for supervised learning.
Returns
self [ColumnTransformer] This estimator
fit_transform(X, y=None)
Fit all transformers, transform the data and concatenate results.
Parameters
X [array-like or DataFrame of shape [n_samples, n_features]] Input data, of which specified
subsets are used to fit the transformers.
Returns
X_t [array-like or sparse matrix, shape (n_samples, sum_n_components)] hstack of results
of transformers. sum_n_components is the sum of n_components (output dimension) over
transformers. If any result is a sparse matrix, everything will be converted to sparse matri-
ces.
get_feature_names()
Get feature names from all transformers.
Returns
feature_names [list of strings] Names of the features produced by transform.
Parameters

Returns
named_transformers_
Access the fitted transformer by name.
Read-only attribute to access any transformer by given name. Keys are transformer names and values are
the fitted transformer objects.
set_params(**kwargs)
Valid parameter keys can be listed with get_params().
Returns
self
transform(X)
Transform X separately by each transformer, concatenate results.
Parameters
X [array-like or DataFrame of shape [n_samples, n_features]] The data to be transformed
by subset.
Returns
transformers. If any result is a sparse matrix, everything will be converted to sparse matri-
ces.
Examples using sklearn.compose.ColumnTransformer

6.5.2 sklearn.compose.TransformedTargetRegressor
class sklearn.compose.TransformedTargetRegressor(regressor=None, transformer=None,

func=None, inverse_func=None,
check_inverse=True)
Meta-estimator to regress on a transformed target.
Useful for applying a non-linear transformation in regression problems. This transformation can be given as a
Transformer such as the QuantileTransformer or as a function and its inverse such as log and exp.
The computation during fit is:
regressor.fit(X, func(y))
or:
regressor.fit(X, transformer.transform(y))

The computation during predict is:

inverse_func(regressor.predict(X))
or:
transformer.inverse_transform(regressor.predict(X))

Parameters
regressor [object, default=LinearRegression()] Regressor object such as derived from
RegressorMixin. This regressor will automatically be cloned each time prior to fitting.
transformer [object, default=None] Estimator object such as derived from
TransformerMixin. Cannot be set at the same time as func and inverse_func. If
transformer is None as well as func and inverse_func, the transformer will be
an identity transformer. Note that the transformer will be cloned during fitting. Also, the
transformer is restricting y to be a numpy array.
func [function, optional] Function to apply to y before passing to fit. Cannot be set at the
same time as transformer. The function needs to return a 2-dimensional array. If func
is None, the function used will be the identity function.
inverse_func [function, optional] Function to apply to the prediction of the regressor. Can-
not be set at the same time as transformer as well. The function needs to return a
2-dimensional array. The inverse function is used to return predictions to the same space of
the original training labels.
check_inverse [bool, default=True] Whether to check that transform followed by
inverse_transform or func followed by inverse_func leads to the original tar-
gets.
Attributes
regressor_ [object] Fitted regressor.
transformer_ [object] Transformer used in fit and predict.
Notes
Internally, the target y is always converted into a 2-dimensional array to be used by scikit-learn transformers.
At the time of prediction, the output will be reshaped to a have the same number of dimensions as y.
See examples/compose/plot_transformed_target.py.
Examples

>>> from sklearn.compose import TransformedTargetRegressor
>>> tt = TransformedTargetRegressor(regressor=LinearRegression(),
... func=np.log, inverse_func=np.exp)
>>> X = np.arange(4).reshape(-1, 1)
>>> y = np.exp(2 * X).ravel()
>>> tt.fit(X, y)

>>> tt.score(X, y)
1.0
>>> tt.regressor_.coef_
array([2.])
Methods
predict(X) Predict using the base regressor, applying inverse.
diction.
__init__(regressor=None, transformer=None, func=None, inverse_func=None,

check_inverse=True)
Parameters
Returns
self [object]
Parameters
Returns
predict(X)
Predict using the base regressor, applying inverse.
The regressor is used to predict and the inverse_func or inverse_transform is applied before
returning the prediction.
Parameters
X [{array-like, sparse matrix}, shape = (n_samples, n_features)] Samples.
Returns
y_hat [array, shape = (n_samples,)] Predicted values.

Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.compose.TransformedTargetRegressor
compose.make_column_transformer(. . . ) Construct a ColumnTransformer from the given transform-

ers.
6.5.3 sklearn.compose.make_column_transformer
sklearn.compose.make_column_transformer(*transformers, **kwargs)
Construct a ColumnTransformer from the given transformers.
This is a shorthand for the ColumnTransformer constructor; it does not require, and does not permit, naming
the transformers. Instead, they will be given names automatically based on their types. It also does not allow
weighting with transformer_weights.
Parameters
*transformers [tuples of transformers and column selections]
remainder [{‘drop’, ‘passthrough’} or estimator, default ‘drop’] By default, only the
specified columns in transformers are transformed and combined in the output, and
the non-specified columns are dropped. (default of 'drop'). By specifying
remainder='passthrough', all remaining columns that were not specified in trans-
formers will be automatically passed through. This subset of columns is concatenated with
the output of the transformers. By setting remainder to be an estimator, the remaining

non-specified columns will use the remainder estimator. The estimator must support fit
and transform.
sparse_threshold [float, default = 0.3] If the transformed output consists of a mix of sparse and
dense data, it will be stacked as a sparse matrix if the density is lower than this value. Use
sparse_threshold=0 to always return dense. When the transformed output consists
of all sparse or all dense data, the stacked result will be sparse or dense, respectively, and
this keyword will be ignored.
Returns
ct [ColumnTransformer]
See also:
sklearn.compose.ColumnTransformer Class that allows combining the outputs of multiple trans-

former objects used on column subsets of the data into a single feature space.
Examples
>>> from sklearn.preprocessing import StandardScaler, OneHotEncoder

>>> from sklearn.compose import make_column_transformer
>>> make_column_transformer(
... (StandardScaler(), ['numerical_column']),
... (OneHotEncoder(), ['categorical_column']))
...
ColumnTransformer(n_jobs=None, remainder='drop', sparse_threshold=0.3,
transformers=[('standardscaler',
StandardScaler(...),
['numerical_column']),
('onehotencoder',
OneHotEncoder(...),
['categorical_column'])])
6.6 sklearn.covariance: Covariance Estimators
The sklearn.covariance module includes methods and algorithms to robustly estimate the covariance of fea-
tures given a set of points. The precision matrix defined as the inverse of the covariance is also estimated. Covariance
estimation is closely related to the theory of Gaussian Graphical Models.
User guide: See the Covariance estimation section for further details.
covariance.EmpiricalCovariance([. . . ]) Maximum likelihood covariance estimator

covariance.EllipticEnvelope([. . . ]) An object for detecting outliers in a Gaussian distributed
dataset.
covariance.GraphicalLasso([alpha, mode, . . . ]) Sparse inverse covariance estimation with an l1-penalized
estimator.
6.6. sklearn.covariance: Covariance Estimators 1457


covariance.GraphicalLassoCV ([alphas, . . . ]) Sparse inverse covariance w/ cross-validated choice of the
l1 penalty.
covariance.LedoitWolf([store_precision, . . . ]) LedoitWolf Estimator
covariance.MinCovDet([store_precision, . . . ]) Minimum Covariance Determinant (MCD): robust estima-
tor of covariance.
covariance.OAS([store_precision, . . . ]) Oracle Approximating Shrinkage Estimator
covariance.ShrunkCovariance([. . . ]) Covariance estimator with shrinkage
6.6.1 sklearn.covariance.EmpiricalCovariance
class sklearn.covariance.EmpiricalCovariance(store_precision=True, as-

sume_centered=False)
Maximum likelihood covariance estimator
Parameters
store_precision [bool] Specifies if the estimated precision is stored.
assume_centered [bool] If True, data are not centered before computation. Useful when work-
ing with data whose mean is almost, but not exactly zero. If False (default), data are centered
before computation.
Attributes
location_ [array-like, shape (n_features,)] Estimated location, i.e. the estimated mean.
covariance_ [2D ndarray, shape (n_features, n_features)] Estimated covariance matrix
precision_ [2D ndarray, shape (n_features, n_features)] Estimated pseudo-inverse matrix.
(stored only if store_precision is True)
Examples

>>> from sklearn.covariance import EmpiricalCovariance
>>> from sklearn.datasets import make_gaussian_quantiles
>>> real_cov = np.array([[.8, .3],
... [.3, .4]])
>>> X = np.random.multivariate_normal(mean=[0, 0],
... cov=real_cov,
... size=500)
>>> cov = EmpiricalCovariance().fit(X)
>>> cov.covariance_
array([[0.7569..., 0.2818...],
[0.2818..., 0.3928...]])
>>> cov.location_
array([0.0622..., 0.0193...])
Methods

error_norm(comp_cov[, norm, scaling, squared]) Computes the Mean Squared Error between two covari-
ance estimators.
fit(X[, y]) Fits the Maximum Likelihood Estimator covariance
model according to the given training data and parame-
ters.
get_precision() Getter for the precision matrix.
mahalanobis(X) Computes the squared Mahalanobis distances of given
observations.
score(X_test[, y]) Computes the log-likelihood of a Gaussian data set with
self.covariance_ as an estimator of its covariance ma-
trix.
__init__(store_precision=True, assume_centered=False)
error_norm(comp_cov, norm=’frobenius’, scaling=True, squared=True)
Computes the Mean Squared Error between two covariance estimators. (In the sense of the Frobenius
norm).
Parameters
comp_cov [array-like, shape = [n_features, n_features]] The covariance to compare with.
norm [str] The type of norm used to compute the error. Available error types: - ‘frobenius’
(default): sqrt(tr(A^t.A)) - ‘spectral’: sqrt(max(eigenvalues(A^t.A)) where A is the error
(comp_cov - self.covariance_).
scaling [bool] If True (default), the squared error norm is divided by n_features. If False,
the squared error norm is not rescaled.
squared [bool] Whether to compute the squared error norm or the error norm. If True
(default), the squared error norm is returned. If False, the error norm is returned.
Returns
The Mean Squared Error (in the sense of the Frobenius norm) between
‘self‘ and ‘comp_cov‘ covariance estimators.
fit(X, y=None)
Fits the Maximum Likelihood Estimator covariance model according to the given training data and param-
eters.
Parameters
X [array-like, shape = [n_samples, n_features]] Training data, where n_samples is the num-
y not used, present for API consistence purpose.
Returns
self [object]
Parameters

Returns
get_precision()
Getter for the precision matrix.
Returns
precision_ [array-like] The precision matrix associated to the current covariance object.
mahalanobis(X)
Computes the squared Mahalanobis distances of given observations.
Parameters
X [array-like, shape = [n_samples, n_features]] The observations, the Mahalanobis distances
of the which we compute. Observations are assumed to be drawn from the same distribu-
tion than the data used in fit.
Returns
dist [array, shape = [n_samples,]] Squared Mahalanobis distances of the observations.
score(X_test, y=None)
Computes the log-likelihood of a Gaussian data set with self.covariance_ as an estimator of its covariance
matrix.
Parameters
X_test [array-like, shape = [n_samples, n_features]] Test data of which we compute the
likelihood, where n_samples is the number of samples and n_features is the number of
features. X_test is assumed to be drawn from the same distribution than the data used in
fit (including centering).
Returns
res [float] The likelihood of the data set with self.covariance_ as an estimator of its covari-
ance matrix.
of a nested object.
Returns
self
Examples using sklearn.covariance.EmpiricalCovariance
• Robust covariance estimation and Mahalanobis distances relevance

• Robust vs Empirical covariance estimate

6.6.2 sklearn.covariance.EllipticEnvelope
class sklearn.covariance.EllipticEnvelope(store_precision=True, assume_centered=False,

support_fraction=None, contamination=0.1,
random_state=None)
An object for detecting outliers in a Gaussian distributed dataset.
Parameters
store_precision [boolean, optional (default=True)] Specify if the estimated precision is stored.
assume_centered [boolean, optional (default=False)] If True, the support of robust location and
covariance estimates is computed, and a covariance estimate is recomputed from it, without
centering the data. Useful to work with data whose mean is significantly equal to zero but is
not exactly zero. If False, the robust location and covariance are directly computed with the
FastMCD algorithm without additional treatment.
support_fraction [float in (0., 1.), optional (default=None)] The proportion of points to be
included in the support of the raw MCD estimate. If None, the minimum value of sup-
port_fraction will be used within the algorithm: [n_sample + n_features + 1] / 2.
contamination [float in (0., 0.5), optional (default=0.1)] The amount of contamination of the
data set, i.e. the proportion of outliers in the data set.
random_state [int, RandomState instance or None, optional (default=None)] The seed of the
pseudo random number generator to use when shuffling the data. If int, random_state is
the seed used by the random number generator; If RandomState instance, random_state is
Attributes
location_ [array-like, shape (n_features,)] Estimated robust location
covariance_ [array-like, shape (n_features, n_features)] Estimated robust covariance matrix
precision_ [array-like, shape (n_features, n_features)] Estimated pseudo inverse matrix. (stored
only if store_precision is True)
support_ [array-like, shape (n_samples,)] A mask of the observations that have been used to
compute the robust estimates of location and shape.
offset_ [float] Offset used to define the decision function from the raw scores. We have the rela-
tion: decision_function = score_samples - offset_. The offset depends
on the contamination parameter and is defined in such a way we obtain the expected number
of outliers (samples with decision function < 0) in training.
See also:
EmpiricalCovariance, MinCovDet
Notes
Outlier detection from covariance estimation may break or not perform well in high-dimensional settings. In
particular, one will always take care to work with n_samples > n_features ** 2.

References
[1]
Methods
correct_covariance(data) Apply a correction to raw Minimum Covariance Deter-

minant estimates.
decision_function(X[, raw_values]) Compute the decision function of the given observa-
tions.
ance estimators.
fit(X[, y]) Fit the EllipticEnvelope model.
fit_predict(X[, y]) Performs fit on X and returns labels for X.
observations.
predict(X) Predict the labels (1 inlier, -1 outlier) of X according to
the fitted model.
reweight_covariance(data) Re-weight raw Minimum Covariance Determinant esti-
mates.
labels.
score_samples(X) Compute the negative Mahalanobis distances.
__init__(store_precision=True, assume_centered=False, support_fraction=None, contamina-

tion=0.1, random_state=None)
correct_covariance(data)
Apply a correction to raw Minimum Covariance Determinant estimates.
Correction using the empirical correction factor suggested by Rousseeuw and Van Driessen in [RVD].
Parameters
data [array-like, shape (n_samples, n_features)] The data matrix, with p features and n sam-
ples. The data set must be the one which was used to compute the raw estimates.
Returns
covariance_corrected [array-like, shape (n_features, n_features)] Corrected robust covari-
ance estimate.
References
[RVD]
decision_function(X, raw_values=None)
Compute the decision function of the given observations.
Parameters

raw_values [bool, optional] Whether or not to consider raw Mahalanobis distances as the
decision function. Must be False (default) for compatibility with the others outlier detec-
tion tools.
Deprecated since version 0.20: raw_values has been deprecated in 0.20 and will be
removed in 0.22.
Returns
decision [array-like, shape (n_samples, )] Decision function of the samples. It is equal to
the shifted Mahalanobis distances. The threshold for being an outlier is 0, which ensures
a compatibility with other outlier detection algorithms.
norm).
Parameters
Returns
fit(X, y=None)
Fit the EllipticEnvelope model.
Parameters
X [numpy array or sparse matrix, shape (n_samples, n_features).] Training data
Performs fit on X and returns labels for X.
Returns -1 for outliers and 1 for inliers.
Parameters
Returns
y [ndarray, shape (n_samples,)] 1 for inliers, -1 for outliers.
Parameters

Returns
get_precision()
Returns
mahalanobis(X)
Parameters
Returns
predict(X)
Predict the labels (1 inlier, -1 outlier) of X according to the fitted model.
Parameters
Returns
is_inlier [array, shape (n_samples,)] Returns -1 for anomalies/outliers and +1 for inliers.
reweight_covariance(data)
Re-weight raw Minimum Covariance Determinant estimates.
Re-weight observations using Rousseeuw’s method (equivalent to deleting outlying observations from the
data set before computing location and covariance estimates) described in [RVDriessen].
Parameters
Returns
location_reweighted [array-like, shape (n_features, )] Re-weighted robust location esti-
mate.
covariance_reweighted [array-like, shape (n_features, n_features)] Re-weighted robust co-
variance estimate.
support_reweighted [array-like, type boolean, shape (n_samples,)] A mask of the obser-
vations that have been used to compute the re-weighted robust location and covariance
estimates.
References
[RVDriessen]

Parameters
X [array-like, shape (n_samples, n_features)] Test samples.
y [array-like, shape (n_samples,) or (n_samples, n_outputs)] True labels for X.
sample_weight [array-like, shape (n_samples,), optional] Sample weights.
Returns
score_samples(X)
Compute the negative Mahalanobis distances.
Parameters
Returns
negative_mahal_distances [array-like, shape (n_samples, )] Opposite of the Mahalanobis
distances.
of a nested object.
Returns
self
Examples using sklearn.covariance.EllipticEnvelope
• Comparing anomaly detection algorithms for outlier detection on toy datasets

• Outlier detection on a real data set
6.6.3 sklearn.covariance.GraphicalLasso
class sklearn.covariance.GraphicalLasso(alpha=0.01, mode=’cd’, tol=0.0001,

enet_tol=0.0001, max_iter=100, verbose=False,
assume_centered=False)
Sparse inverse covariance estimation with an l1-penalized estimator.
Parameters
alpha [positive float, default 0.01] The regularization parameter: the higher alpha, the more
regularization, the sparser the inverse covariance.

mode [{‘cd’, ‘lars’}, default ‘cd’] The Lasso solver to use: coordinate descent or LARS. Use
LARS for very sparse underlying graphs, where p > n. Elsewhere prefer cd which is more
numerically stable.
tol [positive float, default 1e-4] The tolerance to declare convergence: if the dual gap goes below
this value, iterations are stopped.
enet_tol [positive float, optional] The tolerance for the elastic net solver used to calculate the
descent direction. This parameter controls the accuracy of the search direction for a given
column update, not of the overall parameter estimate. Only used for mode=’cd’.
max_iter [integer, default 100] The maximum number of iterations.
verbose [boolean, default False] If verbose is True, the objective function and dual gap are
plotted at each iteration.
assume_centered [boolean, default False] If True, data are not centered before computation.
Useful when working with data whose mean is almost, but not exactly zero. If False, data
are centered before computation.
Attributes
covariance_ [array-like, shape (n_features, n_features)] Estimated covariance matrix
precision_ [array-like, shape (n_features, n_features)] Estimated pseudo inverse matrix.
See also:
graphical_lasso, GraphicalLassoCV
Methods
ance estimators.
fit(X[, y]) Fits the GraphicalLasso model to X.
observations.
trix.
__init__(alpha=0.01, mode=’cd’, tol=0.0001, enet_tol=0.0001, max_iter=100, verbose=False, as-

norm).
Parameters

Returns
fit(X, y=None)
Fits the GraphicalLasso model to X.
Parameters
X [ndarray, shape (n_samples, n_features)] Data from which to compute the covariance es-
timate
y [(ignored)]
Parameters
Returns
get_precision()
Returns
mahalanobis(X)
Parameters
Returns
matrix.
Parameters


Returns
ance matrix.
of a nested object.
Returns
self
6.6.4 sklearn.covariance.GraphicalLassoCV
class sklearn.covariance.GraphicalLassoCV(alphas=4, n_refinements=4, cv=’warn’,

tol=0.0001, enet_tol=0.0001, max_iter=100,
mode=’cd’, n_jobs=None, verbose=False, as-
Sparse inverse covariance w/ cross-validated choice of the l1 penalty.
Parameters
alphas [integer, or list positive float, optional] If an integer is given, it fixes the number of points
on the grids of alpha to be used. If a list is given, it gives the grid to be used. See the notes
in the class docstring for more details.
n_refinements [strictly positive integer] The number of times the grid is refined. Not used if
explicit values of alphas are passed.
• CV splitter,
For integer/None inputs KFold is used.
v0.22.
tol [positive float, optional] The tolerance to declare convergence: if the dual gap goes below

max_iter [integer, optional] Maximum number of iterations.

mode [{‘cd’, ‘lars’}] The Lasso solver to use: coordinate descent or LARS. Use LARS for
very sparse underlying graphs, where number of features is greater than number of samples.
Elsewhere prefer cd which is more numerically stable.
n_jobs [int or None, optional (default=None)] number of jobs to run in parallel. None means 1
verbose [boolean, optional] If verbose is True, the objective function and duality gap are printed
at each iteration.
assume_centered [boolean] If True, data are not centered before computation. Useful when
working with data whose mean is almost, but not exactly zero. If False, data are centered
before computation.
Attributes
covariance_ [numpy.ndarray, shape (n_features, n_features)] Estimated covariance matrix.
precision_ [numpy.ndarray, shape (n_features, n_features)] Estimated precision matrix (inverse
covariance).
alpha_ [float] Penalization parameter selected.
cv_alphas_ [list of float] All penalization parameters explored.
grid_scores_ [2D numpy.ndarray (n_alphas, n_folds)] Log-likelihood score on left-out data
across folds.
n_iter_ [int] Number of iterations run for the optimal alpha.
See also:
graphical_lasso, GraphicalLasso
Notes
The search for the optimal penalization parameter (alpha) is done on an iteratively refined grid: first the cross-
validated scores on a grid are computed, then a new refined grid is centered around the maximum, and so on.
One of the challenges which is faced here is that the solvers can fail to converge to a well-conditioned estimate.
The corresponding values of alpha then come out as missing values, but the optimum may be close to these
missing values.
Methods
ance estimators.
fit(X[, y]) Fits the GraphicalLasso covariance model to X.
observations.


trix.
__init__(alphas=4, n_refinements=4, cv=’warn’, tol=0.0001, enet_tol=0.0001, max_iter=100,

mode=’cd’, n_jobs=None, verbose=False, assume_centered=False)
norm).
Parameters
Returns
fit(X, y=None)
Fits the GraphicalLasso covariance model to X.
Parameters
X [ndarray, shape (n_samples, n_features)] Data from which to compute the covariance es-
timate
y [(ignored)]
Parameters
Returns
get_precision()
Returns
grid_scores
DEPRECATED: Attribute grid_scores was deprecated in version 0.19 and will be removed in 0.21. Use
grid_scores_ instead

mahalanobis(X)
Parameters
Returns
matrix.
Parameters
Returns
ance matrix.
of a nested object.
Returns
self
Examples using sklearn.covariance.GraphicalLassoCV

• Sparse inverse covariance estimation
6.6.5 sklearn.covariance.LedoitWolf
class sklearn.covariance.LedoitWolf(store_precision=True, assume_centered=False,

block_size=1000)
LedoitWolf Estimator
Ledoit-Wolf is a particular form of shrinkage, where the shrinkage coefficient is computed using O. Ledoit and
M. Wolf’s formula as described in “A Well-Conditioned Estimator for Large-Dimensional Covariance Matrices”,
Ledoit and Wolf, Journal of Multivariate Analysis, Volume 88, Issue 2, February 2004, pages 365-411.
Parameters

store_precision [bool, default=True] Specify if the estimated precision is stored.

assume_centered [bool, default=False] If True, data are not centered before computation. Use-
ful when working with data whose mean is almost, but not exactly zero. If False (default),
data are centered before computation.
block_size [int, default=1000] Size of the blocks into which the covariance matrix will be split
during its Ledoit-Wolf estimation. This is purely a memory optimization and does not affect
results.
Attributes
shrinkage_ [float, 0 <= shrinkage <= 1] Coefficient in the convex combination used for the
computation of the shrunk estimate.
Notes
The regularised covariance is:

(1 - shrinkage) * cov + shrinkage * mu * np.identity(n_features)
where mu = trace(cov) / n_features and shrinkage is given by the Ledoit and Wolf formula (see References)
References
“A Well-Conditioned Estimator for Large-Dimensional Covariance Matrices”, Ledoit and Wolf, Journal of Mul-
tivariate Analysis, Volume 88, Issue 2, February 2004, pages 365-411.
Examples

>>> from sklearn.covariance import LedoitWolf
... [.2, .8]])
... cov=real_cov,
... size=50)
>>> cov = LedoitWolf().fit(X)
>>> cov.covariance_
array([[0.4406..., 0.1616...],
[0.1616..., 0.8022...]])
>>> cov.location_
array([ 0.0595... , -0.0075...])
Methods

ance estimators.
fit(X[, y]) Fits the Ledoit-Wolf shrunk covariance model accord-
ing to the given training data and parameters.
observations.
trix.
__init__(store_precision=True, assume_centered=False, block_size=1000)

norm).
Parameters
Returns
fit(X, y=None)
Fits the Ledoit-Wolf shrunk covariance model according to the given training data and parameters.
Parameters
Returns
self [object]
Parameters
Returns


get_precision()
Returns
mahalanobis(X)
Parameters
Returns
matrix.
Parameters
Returns
ance matrix.
of a nested object.
Returns
self
Examples using sklearn.covariance.LedoitWolf
• Ledoit-Wolf vs OAS estimation

• Shrinkage covariance estimation: LedoitWolf vs OAS and max-likelihood

6.6.6 sklearn.covariance.MinCovDet
class sklearn.covariance.MinCovDet(store_precision=True, assume_centered=False, sup-

port_fraction=None, random_state=None)
Minimum Covariance Determinant (MCD): robust estimator of covariance.
The Minimum Covariance Determinant covariance estimator is to be applied on Gaussian-distributed data, but
could still be relevant on data drawn from a unimodal, symmetric distribution. It is not meant to be used with
multi-modal data (the algorithm used to fit a MinCovDet object is likely to fail in such a case). One should
consider projection pursuit methods to deal with multi-modal datasets.
Parameters
store_precision [bool] Specify if the estimated precision is stored.
assume_centered [bool] If True, the support of the robust location and the covariance estimates
is computed, and a covariance estimate is recomputed from it, without centering the data.
Useful to work with data whose mean is significantly equal to zero but is not exactly zero. If
False, the robust location and covariance are directly computed with the FastMCD algorithm
without additional treatment.
support_fraction [float, 0 < support_fraction < 1] The proportion of points to be included in
the support of the raw MCD estimate. Default is None, which implies that the minimum
value of support_fraction will be used within the algorithm: [n_sample + n_features + 1] / 2
Attributes
raw_location_ [array-like, shape (n_features,)] The raw robust estimated location before cor-
rection and re-weighting.
raw_covariance_ [array-like, shape (n_features, n_features)] The raw robust estimated covari-
ance before correction and re-weighting.
raw_support_ [array-like, shape (n_samples,)] A mask of the observations that have been
used to compute the raw robust estimates of location and shape, before correction and re-
weighting.
location_ [array-like, shape (n_features,)] Estimated robust location
covariance_ [array-like, shape (n_features, n_features)] Estimated robust covariance matrix
support_ [array-like, shape (n_samples,)] A mask of the observations that have been used to
compute the robust estimates of location and shape.
dist_ [array-like, shape (n_samples,)] Mahalanobis distances of the training set (on which fit is
called) observations.
References
[Rouseeuw1984], [Rousseeuw], [ButlerDavies]

Examples

>>> from sklearn.covariance import MinCovDet
... [.3, .4]])
... cov=real_cov,
... size=500)
>>> cov = MinCovDet(random_state=0).fit(X)
>>> cov.covariance_
array([[0.7411..., 0.2535...],
[0.2535..., 0.3053...]])
>>> cov.location_
array([0.0813... , 0.0427...])
Methods
correct_covariance(data) Apply a correction to raw Minimum Covariance Deter-

minant estimates.
ance estimators.
fit(X[, y]) Fits a Minimum Covariance Determinant with the
FastMCD algorithm.
observations.
reweight_covariance(data) Re-weight raw Minimum Covariance Determinant esti-
mates.
trix.
__init__(store_precision=True, assume_centered=False, support_fraction=None, ran-

dom_state=None)
correct_covariance(data)
Apply a correction to raw Minimum Covariance Determinant estimates.
Correction using the empirical correction factor suggested by Rousseeuw and Van Driessen in [RVD].
Parameters
Returns
covariance_corrected [array-like, shape (n_features, n_features)] Corrected robust covari-
ance estimate.

References
[RVD]
norm).
Parameters
Returns
fit(X, y=None)
Fits a Minimum Covariance Determinant with the FastMCD algorithm.
Parameters
Returns
self [object]
Parameters
Returns
get_precision()
Returns
mahalanobis(X)
Parameters


Returns
reweight_covariance(data)
Re-weight raw Minimum Covariance Determinant estimates.
Re-weight observations using Rousseeuw’s method (equivalent to deleting outlying observations from the
data set before computing location and covariance estimates) described in [RVDriessen].
Parameters
Returns
location_reweighted [array-like, shape (n_features, )] Re-weighted robust location esti-
mate.
covariance_reweighted [array-like, shape (n_features, n_features)] Re-weighted robust co-
variance estimate.
support_reweighted [array-like, type boolean, shape (n_samples,)] A mask of the obser-
vations that have been used to compute the re-weighted robust location and covariance
estimates.
References
[RVDriessen]
matrix.
Parameters
Returns
ance matrix.
of a nested object.
Returns
self

Examples using sklearn.covariance.MinCovDet
• Robust covariance estimation and Mahalanobis distances relevance

• Robust vs Empirical covariance estimate
6.6.7 sklearn.covariance.OAS
class sklearn.covariance.OAS(store_precision=True, assume_centered=False)

Oracle Approximating Shrinkage Estimator
OAS is a particular form of shrinkage described in “Shrinkage Algorithms for MMSE Covariance Estimation”
Chen et al., IEEE Trans. on Sign. Proc., Volume 58, Issue 10, October 2010.
The formula used here does not correspond to the one given in the article. In the original article, formula (23)
states that 2/p is multiplied by Trace(cov*cov) in both the numerator and denominator, but this operation is
omitted because for a large p, the value of 2/p is so small that it doesn’t affect the value of the estimator.
Parameters
store_precision [bool, default=True] Specify if the estimated precision is stored.
assume_centered [bool, default=False] If True, data are not centered before computation. Use-
ful when working with data whose mean is almost, but not exactly zero. If False (default),
data are centered before computation.
Attributes
covariance_ [array-like, shape (n_features, n_features)] Estimated covariance matrix.
shrinkage_ [float, 0 <= shrinkage <= 1] coefficient in the convex combination used for the
Notes
The regularised covariance is:

where mu = trace(cov) / n_features and shrinkage is given by the OAS formula (see References)
References
“Shrinkage Algorithms for MMSE Covariance Estimation” Chen et al., IEEE Trans. on Sign. Proc., Volume 58,
Issue 10, October 2010.
Methods

ance estimators.
fit(X[, y]) Fits the Oracle Approximating Shrinkage covariance
model according to the given training data and parame-
ters.
observations.
trix.
__init__(store_precision=True, assume_centered=False)
norm).
Parameters
Returns
fit(X, y=None)
Fits the Oracle Approximating Shrinkage covariance model according to the given training data and pa-
rameters.
Parameters
Returns
self [object]
Parameters

Returns
get_precision()
Returns
mahalanobis(X)
Parameters
Returns
matrix.
Parameters
Returns
ance matrix.
of a nested object.
Returns
self
Examples using sklearn.covariance.OAS
• Ledoit-Wolf vs OAS estimation


6.6.8 sklearn.covariance.ShrunkCovariance
class sklearn.covariance.ShrunkCovariance(store_precision=True, assume_centered=False,

shrinkage=0.1)
Covariance estimator with shrinkage
Parameters
store_precision [boolean, default True] Specify if the estimated precision is stored
assume_centered [boolean, default False] If True, data are not centered before computation.
Useful when working with data whose mean is almost, but not exactly zero. If False, data
are centered before computation.
shrinkage [float, 0 <= shrinkage <= 1, default 0.1] Coefficient in the convex combination used
for the computation of the shrunk estimate.
Attributes
shrinkage [float, 0 <= shrinkage <= 1] Coefficient in the convex combination used for the
Notes
The regularized covariance is given by:

where mu = trace(cov) / n_features
Examples

>>> from sklearn.covariance import ShrunkCovariance
... [.3, .4]])
... cov=real_cov,
... size=500)
>>> cov = ShrunkCovariance().fit(X)
>>> cov.covariance_
array([[0.7387..., 0.2536...],
[0.2536..., 0.4110...]])
>>> cov.location_
array([0.0622..., 0.0193...])

Methods
ance estimators.
fit(X[, y]) Fits the shrunk covariance model according to the given
training data and parameters.
observations.
trix.
__init__(store_precision=True, assume_centered=False, shrinkage=0.1)

norm).
Parameters
Returns
fit(X, y=None)
Fits the shrunk covariance model according to the given training data and parameters.
Parameters
Returns
self [object]
Parameters

Returns
get_precision()
Returns
mahalanobis(X)
Parameters
Returns
matrix.
Parameters
Returns
ance matrix.
of a nested object.
Returns
self
Examples using sklearn.covariance.ShrunkCovariance


covariance.empirical_covariance(X[, . . . ]) Computes the Maximum likelihood covariance estimator

covariance.graphical_lasso(emp_cov, alpha[, l1-penalized covariance estimator
. . . ])
covariance.ledoit_wolf(X[, assume_centered, Estimates the shrunk Ledoit-Wolf covariance matrix.
. . . ])
covariance.oas(X[, assume_centered]) Estimate covariance with the Oracle Approximating
Shrinkage algorithm.
covariance.shrunk_covariance(emp_cov[, . . . ]) Calculates a covariance matrix shrunk on the diagonal
6.6.9 sklearn.covariance.empirical_covariance
sklearn.covariance.empirical_covariance(X, assume_centered=False)
Computes the Maximum likelihood covariance estimator
Parameters
X [ndarray, shape (n_samples, n_features)] Data from which to compute the covariance estimate
assume_centered [boolean] If True, data are not centered before computation. Useful when
working with data whose mean is almost, but not exactly zero. If False, data are centered
before computation.
Returns
covariance [2D ndarray, shape (n_features, n_features)] Empirical covariance (Maximum Like-
lihood Estimator).
Examples using sklearn.covariance.empirical_covariance
6.6.10 sklearn.covariance.graphical_lasso
sklearn.covariance.graphical_lasso(emp_cov, alpha, cov_init=None, mode=’cd’, tol=0.0001,

enet_tol=0.0001, max_iter=100, verbose=False, re-
turn_costs=False, eps=2.220446049250313e-16, re-
turn_n_iter=False)
l1-penalized covariance estimator
Parameters
emp_cov [2D ndarray, shape (n_features, n_features)] Empirical covariance from which to
compute the covariance estimate.
alpha [positive float] The regularization parameter: the higher alpha, the more regularization,
the sparser the inverse covariance.
cov_init [2D array (n_features, n_features), optional] The initial guess for the covariance.
mode [{‘cd’, ‘lars’}] The Lasso solver to use: coordinate descent or LARS. Use LARS for
very sparse underlying graphs, where p > n. Elsewhere prefer cd which is more numerically
stable.

tol [positive float, optional] The tolerance to declare convergence: if the dual gap goes below
max_iter [integer, optional] The maximum number of iterations.
verbose [boolean, optional] If verbose is True, the objective function and dual gap are printed
at each iteration.
return_costs [boolean, optional] If return_costs is True, the objective function and dual gap at
each iteration are returned.
Returns
covariance [2D ndarray, shape (n_features, n_features)] The estimated covariance matrix.
precision [2D ndarray, shape (n_features, n_features)] The estimated (sparse) precision matrix.
costs [list of (objective, dual_gap) pairs] The list of values of the objective function and the dual
gap at each iteration. Returned only if return_costs is True.
n_iter [int] Number of iterations. Returned only if return_n_iter is set to True.
See also:
GraphicalLasso, GraphicalLassoCV
Notes
The algorithm employed to solve this problem is the GLasso algorithm, from the Friedman 2008 Biostatistics
paper. It is the same algorithm as in the R glasso package.
One possible difference with the glasso R package is that the diagonal coefficients are not penalized.
6.6.11 sklearn.covariance.ledoit_wolf
sklearn.covariance.ledoit_wolf(X, assume_centered=False, block_size=1000)

Estimates the shrunk Ledoit-Wolf covariance matrix.
Parameters
X [array-like, shape (n_samples, n_features)] Data from which to compute the covariance esti-
mate
assume_centered [boolean, default=False] If True, data are not centered before computation.
Useful to work with data whose mean is significantly equal to zero but is not exactly zero.
If False, data are centered before computation.
block_size [int, default=1000] Size of the blocks into which the covariance matrix will be split.
This is purely a memory optimization and does not affect results.
Returns

shrunk_cov [array-like, shape (n_features, n_features)] Shrunk covariance.

shrinkage [float] Coefficient in the convex combination used for the computation of the shrunk
estimate.
Notes
The regularized (shrunk) covariance is:

Examples using sklearn.covariance.ledoit_wolf
6.6.12 sklearn.covariance.oas
sklearn.covariance.oas(X, assume_centered=False)
Estimate covariance with the Oracle Approximating Shrinkage algorithm.
Parameters
X [array-like, shape (n_samples, n_features)] Data from which to compute the covariance esti-
mate.
assume_centered [boolean] If True, data are not centered before computation. Useful to work
with data whose mean is significantly equal to zero but is not exactly zero. If False, data are
centered before computation.
Returns
shrunk_cov [array-like, shape (n_features, n_features)] Shrunk covariance.
shrinkage [float] Coefficient in the convex combination used for the computation of the shrunk
estimate.
Notes
The regularised (shrunk) covariance is:

The formula we used to implement the OAS is slightly modified compared to the one given in the article. See
OAS for more details.
6.6.13 sklearn.covariance.shrunk_covariance
sklearn.covariance.shrunk_covariance(emp_cov, shrinkage=0.1)
Calculates a covariance matrix shrunk on the diagonal
Parameters

emp_cov [array-like, shape (n_features, n_features)] Covariance matrix to be shrunk

shrinkage [float, 0 <= shrinkage <= 1] Coefficient in the convex combination used for the
Returns
shrunk_cov [array-like] Shrunk covariance.
Notes
The regularized (shrunk) covariance is given by:

6.7 sklearn.cross_decomposition: Cross decomposition
User guide: See the Cross decomposition section for further details.
cross_decomposition.CCA([n_components, . . . ]) CCA Canonical Correlation Analysis.

cross_decomposition.PLSCanonical([. . . ]) PLSCanonical implements the 2 blocks canonical PLS of
the original Wold algorithm [Tenenhaus 1998] p.204, re-
ferred as PLS-C2A in [Wegelin 2000].
cross_decomposition.PLSRegression([. . . ]) PLS regression
cross_decomposition.PLSSVD([n_components, Partial Least Square SVD
. . . ])
6.7.1 sklearn.cross_decomposition.CCA
class sklearn.cross_decomposition.CCA(n_components=2, scale=True, max_iter=500, tol=1e-06,

copy=True)
CCA Canonical Correlation Analysis.
CCA inherits from PLS with mode=”B” and deflation_mode=”canonical”.
Parameters
n_components [int, (default 2).] number of components to keep.
scale [boolean, (default True)] whether to scale the data?
max_iter [an integer, (default 500)] the maximum number of iterations of the NIPALS inner
loop
tol [non-negative real, default 1e-06.] the tolerance used in the iterative algorithm
copy [boolean] Whether the deflation be done on a copy. Let the default value to True unless
you don’t care about side effects
Attributes
x_weights_ [array, [p, n_components]] X block weights vectors.
y_weights_ [array, [q, n_components]] Y block weights vectors.

x_loadings_ [array, [p, n_components]] X block loadings vectors.

y_loadings_ [array, [q, n_components]] Y block loadings vectors.
x_scores_ [array, [n_samples, n_components]] X scores.
y_scores_ [array, [n_samples, n_components]] Y scores.
x_rotations_ [array, [p, n_components]] X block to latents rotations.
y_rotations_ [array, [q, n_components]] Y block to latents rotations.
n_iter_ [array-like] Number of iterations of the NIPALS inner loop for each component.
See also:
PLSCanonical, PLSSVD
Notes
For each component k, find the weights u, v that maximizes max corr(Xk u, Yk v), such that |u| = |v| =
1
Note that it maximizes only the correlations between the scores.
The residual matrix of X (Xk+1) block is obtained by the deflation on the current X score: x_score.
The residual matrix of Y (Yk+1) block is obtained by deflation on the current Y score.
References
Jacob A. Wegelin. A survey of Partial Least Squares (PLS) methods, with emphasis on the two-block case.
Technical Report 371, Department of Statistics, University of Washington, Seattle, 2000.
In french but still a reference: Tenenhaus, M. (1998). La regression PLS: theorie et pratique. Paris: Editions
Technic.
Examples
>>> from sklearn.cross_decomposition import CCA

>>> X = [[0., 0., 1.], [1.,0.,0.], [2.,2.,2.], [3.,5.,4.]]
>>> Y = [[0.1, -0.2], [0.9, 1.1], [6.2, 5.9], [11.9, 12.3]]
>>> cca = CCA(n_components=1)
>>> cca.fit(X, Y)
...
CCA(copy=True, max_iter=500, n_components=1, scale=True, tol=1e-06)
>>> X_c, Y_c = cca.transform(X, Y)
Methods
fit(X, Y) Fit model to data.

fit_transform(X[, y]) Learn and apply the dimension reduction on the train
data.
6.7. sklearn.cross_decomposition: Cross decomposition 1489


predict(X[, copy]) Apply the dimension reduction learned on the train data.
diction.
transform(X[, Y, copy]) Apply the dimension reduction learned on the train data.
__init__(n_components=2, scale=True, max_iter=500, tol=1e-06, copy=True)

fit(X, Y)
Fit model to data.
Parameters
X [array-like, shape = [n_samples, n_features]] Training vectors, where n_samples is the
number of samples and n_features is the number of predictors.
Y [array-like, shape = [n_samples, n_targets]] Target vectors, where n_samples is the num-
ber of samples and n_targets is the number of response variables.
Learn and apply the dimension reduction on the train data.
Parameters
y [array-like, shape = [n_samples, n_targets]] Target vectors, where n_samples is the number
of samples and n_targets is the number of response variables.
Returns
x_scores if Y is not given, (x_scores, y_scores) otherwise.
Parameters
Returns
predict(X, copy=True)
Apply the dimension reduction learned on the train data.
Parameters
copy [boolean, default True] Whether to copy X and Y, or perform in-place normalization.
Notes
This call requires the estimation of a p x q matrix, which may be an issue in high dimensional space.

Parameters
Returns
of a nested object.
Returns
self
transform(X, Y=None, copy=True)
Parameters
Returns
Examples using sklearn.cross_decomposition.CCA
6.7.2 sklearn.cross_decomposition.PLSCanonical
class sklearn.cross_decomposition.PLSCanonical(n_components=2, scale=True, algo-

rithm=’nipals’, max_iter=500, tol=1e-06,
copy=True)
PLSCanonical implements the 2 blocks canonical PLS of the original Wold algorithm [Tenenhaus 1998] p.204,

referred as PLS-C2A in [Wegelin 2000].

This class inherits from PLS with mode=”A” and deflation_mode=”canonical”, norm_y_weights=True and al-
gorithm=”nipals”, but svd should provide similar results up to numerical errors.
Parameters
n_components [int, (default 2).] Number of components to keep
scale [boolean, (default True)] Option to scale data
algorithm [string, “nipals” or “svd”] The algorithm used to estimate the weights. It will be
called n_components times, i.e. once for each iteration of the outer loop.
loop (used only if algorithm=”nipals”)
tol [non-negative real, default 1e-06] the tolerance used in the iterative algorithm
copy [boolean, default True] Whether the deflation should be done on a copy. Let the default
value to True unless you don’t care about side effect
Attributes
x_weights_ [array, shape = [p, n_components]] X block weights vectors.
y_weights_ [array, shape = [q, n_components]] Y block weights vectors.
x_loadings_ [array, shape = [p, n_components]] X block loadings vectors.
y_loadings_ [array, shape = [q, n_components]] Y block loadings vectors.
x_scores_ [array, shape = [n_samples, n_components]] X scores.
y_scores_ [array, shape = [n_samples, n_components]] Y scores.
x_rotations_ [array, shape = [p, n_components]] X block to latents rotations.
y_rotations_ [array, shape = [q, n_components]] Y block to latents rotations.
n_iter_ [array-like] Number of iterations of the NIPALS inner loop for each component. Not
useful if the algorithm provided is “svd”.
See also:
CCA, PLSSVD
Notes
Matrices:
T: x_scores_
U: y_scores_
W: x_weights_
C: y_weights_
P: x_loadings_
Q: y_loadings__
Are computed such that:

X = T P.T + Err and Y = U Q.T + Err

T[:, k] = Xk W[:, k] for k in range(n_components)
U[:, k] = Yk C[:, k] for k in range(n_components)
x_rotations_ = W (P.T W)^(-1)
y_rotations_ = C (Q.T C)^(-1)
where Xk and Yk are residual matrices at iteration k.

Slides explaining PLS
For each component k, find weights u, v that optimize:
max corr(Xk u, Yk v) * std(Xk u) std(Yk u), such that ``|u| = |v| = 1``
Note that it maximizes both the correlations between the scores and the intra-block variances.
The residual matrix of Y (Yk+1) block is obtained by deflation on the current Y score. This performs a canonical
symmetric version of the PLS regression. But slightly different than the CCA. This is mostly used for modeling.
This implementation provides the same results that the “plspm” package provided in the R language (R-
project), using the function plsca(X, Y). Results are equal or collinear with the function pls(..., mode
= "canonical") of the “mixOmics” package. The difference relies in the fact that mixOmics implementa-
tion does not exactly implement the Wold algorithm since it does not normalize y_weights to one.
References
Tenenhaus, M. (1998). La regression PLS: theorie et pratique. Paris: Editions Technic.
Examples
>>> from sklearn.cross_decomposition import PLSCanonical

>>> X = [[0., 0., 1.], [1.,0.,0.], [2.,2.,2.], [2.,5.,4.]]
>>> Y = [[0.1, -0.2], [0.9, 1.1], [6.2, 5.9], [11.9, 12.3]]
>>> plsca = PLSCanonical(n_components=2)
>>> plsca.fit(X, Y)
...
PLSCanonical(algorithm='nipals', copy=True, max_iter=500, n_components=2,
scale=True, tol=1e-06)
>>> X_c, Y_c = plsca.transform(X, Y)
Methods

data.


diction.
__init__(n_components=2, scale=True, algorithm=’nipals’, max_iter=500, tol=1e-06, copy=True)

fit(X, Y)
Fit model to data.
Parameters
Parameters
Returns
Parameters
Returns
Parameters
Notes

Parameters
Returns
of a nested object.
Returns
self
Parameters
Returns
Examples using sklearn.cross_decomposition.PLSCanonical
6.7.3 sklearn.cross_decomposition.PLSRegression
class sklearn.cross_decomposition.PLSRegression(n_components=2, scale=True,

max_iter=500, tol=1e-06, copy=True)
PLS regression
PLSRegression implements the PLS 2 blocks regression known as PLS2 or PLS1 in case of one dimensional
response. This class inherits from _PLS with mode=”A”, deflation_mode=”regression”, norm_y_weights=False
and algorithm=”nipals”.

Parameters
n_components [int, (default 2)] Number of components to keep.
scale [boolean, (default True)] whether to scale the data
loop (used only if algorithm=”nipals”)
tol [non-negative real] Tolerance used in the iterative algorithm default 1e-06.
copy [boolean, default True] Whether the deflation should be done on a copy. Let the default
value to True unless you don’t care about side effect
Attributes
x_loadings_ [array, [p, n_components]] X block loadings vectors.
y_loadings_ [array, [q, n_components]] Y block loadings vectors.
x_rotations_ [array, [p, n_components]] X block to latents rotations.
y_rotations_ [array, [q, n_components]] Y block to latents rotations.
coef_ [array, [p, q]] The coefficients of the linear model: Y = X coef_ + Err
n_iter_ [array-like] Number of iterations of the NIPALS inner loop for each component.
Notes
Matrices:
T: x_scores_
U: y_scores_
W: x_weights_
C: y_weights_
P: x_loadings_
Q: y_loadings__
Are computed such that:
X = T P.T + Err and Y = U Q.T + Err

T[:, k] = Xk W[:, k] for k in range(n_components)
U[:, k] = Yk C[:, k] for k in range(n_components)
x_rotations_ = W (P.T W)^(-1)
y_rotations_ = C (Q.T C)^(-1)
where Xk and Yk are residual matrices at iteration k.

Slides explaining PLS
For each component k, find weights u, v that optimizes: max corr(Xk u, Yk v) * std(Xk u)
std(Yk u), such that |u| = 1
Note that it maximizes both the correlations between the scores and the intra-block variances.

The residual matrix of Y (Yk+1) block is obtained by deflation on the current X score. This performs the PLS
regression known as PLS2. This mode is prediction oriented.
This implementation provides the same results that 3 PLS packages provided in the R language (R-project):
• “mixOmics” with function pls(X, Y, mode = “regression”)
• “plspm ” with function plsreg2(X, Y)
• “pls” with function oscorespls.fit(X, Y)
References
In french but still a reference: Tenenhaus, M. (1998). La regression PLS: theorie et pratique. Paris: Editions
Technic.
Examples
>>> from sklearn.cross_decomposition import PLSRegression

>>> X = [[0., 0., 1.], [1.,0.,0.], [2.,2.,2.], [2.,5.,4.]]
>>> Y = [[0.1, -0.2], [0.9, 1.1], [6.2, 5.9], [11.9, 12.3]]
>>> pls2 = PLSRegression(n_components=2)
>>> pls2.fit(X, Y)
...
PLSRegression(copy=True, max_iter=500, n_components=2, scale=True,
tol=1e-06)
>>> Y_pred = pls2.predict(X)
Methods

data.
diction.
__init__(n_components=2, scale=True, max_iter=500, tol=1e-06, copy=True)

fit(X, Y)
Fit model to data.
Parameters

Parameters
Returns
Parameters
Returns
Parameters
Notes
Parameters
Returns


of a nested object.
Returns
self
Parameters
Returns
Examples using sklearn.cross_decomposition.PLSRegression
6.7.4 sklearn.cross_decomposition.PLSSVD
class sklearn.cross_decomposition.PLSSVD(n_components=2, scale=True, copy=True)

Partial Least Square SVD
Simply perform a svd on the crosscovariance matrix: X’Y There are no iterative deflation here.
Parameters
n_components [int, default 2] Number of components to keep.
scale [boolean, default True] Whether to scale X and Y.
copy [boolean, default True] Whether to copy X and Y, or perform in-place computations.
Attributes
See also:
PLSCanonical, CCA

Examples

>>> from sklearn.cross_decomposition import PLSSVD
>>> X = np.array([[0., 0., 1.],
... [1.,0.,0.],
... [2.,2.,2.],
... [2.,5.,4.]])
>>> Y = np.array([[0.1, -0.2],
... [0.9, 1.1],
... [6.2, 5.9],
... [11.9, 12.3]])
>>> plsca = PLSSVD(n_components=2)
>>> plsca.fit(X, Y)
PLSSVD(copy=True, n_components=2, scale=True)
>>> X_c, Y_c = plsca.transform(X, Y)
>>> X_c.shape, Y_c.shape
((4, 2), (4, 2))
Methods

data.
transform(X[, Y]) Apply the dimension reduction learned on the train data.
__init__(n_components=2, scale=True, copy=True)

fit(X, Y)
Fit model to data.
Parameters
Parameters
Returns

Parameters
Returns
of a nested object.
Returns
self
transform(X, Y=None)
Parameters
6.8 sklearn.datasets: Datasets
The sklearn.datasets module includes utilities to load datasets, including methods to load and fetch popular
reference datasets. It also features some artificial data generators.
User guide: See the Dataset loading utilities section for further details.
6.8.1 Loaders
datasets.clear_data_home([data_home]) Delete all the content of the data home cache.

datasets.dump_svmlight_file(X, y, f[, . . . ]) Dump the dataset in svmlight / libsvm file format.
datasets.fetch_20newsgroups([data_home, Load the filenames and data from the 20 newsgroups
. . . ]) dataset (classification).
datasets.fetch_20newsgroups_vectorized([. . . Load
]) the 20 newsgroups dataset and vectorize it into token
counts (classification).
datasets.fetch_california_housing([. . . ]) Load the California housing dataset (regression).
datasets.fetch_covtype([data_home, . . . ]) Load the covertype dataset (classification).
datasets.fetch_kddcup99([subset, data_home, Load the kddcup99 dataset (classification).
. . . ])
datasets.fetch_lfw_pairs([subset, . . . ]) Load the Labeled Faces in the Wild (LFW) pairs dataset
(classification).
6.8. sklearn.datasets: Datasets 1501


datasets.fetch_lfw_people([data_home, . . . ]) Load the Labeled Faces in the Wild (LFW) people dataset
(classification).
datasets.fetch_olivetti_faces([data_home, Load the Olivetti faces data-set from AT&T (classification).
. . . ])
datasets.fetch_openml([name, version, . . . ]) Fetch dataset from openml by name or dataset id.
datasets.fetch_rcv1([data_home, subset, . . . ]) Load the RCV1 multilabel dataset (classification).
datasets.fetch_species_distributions([. . . ]) Loader for species distribution dataset from Phillips et.
datasets.get_data_home([data_home]) Return the path of the scikit-learn data dir.
datasets.load_boston([return_X_y]) Load and return the boston house-prices dataset (regres-
sion).
datasets.load_breast_cancer([return_X_y]) Load and return the breast cancer wisconsin dataset (clas-
sification).
datasets.load_diabetes([return_X_y]) Load and return the diabetes dataset (regression).
datasets.load_digits([n_class, return_X_y]) Load and return the digits dataset (classification).
datasets.load_files(container_path[, . . . ]) Load text files with categories as subfolder names.
datasets.load_iris([return_X_y]) Load and return the iris dataset (classification).
datasets.load_linnerud([return_X_y]) Load and return the linnerud dataset (multivariate regres-
sion).
datasets.load_sample_image(image_name) Load the numpy array of a single sample image
datasets.load_sample_images() Load sample images for image manipulation.
datasets.load_svmlight_file(f[, n_features, Load datasets in the svmlight / libsvm format into sparse
. . . ]) CSR matrix
datasets.load_svmlight_files(files[, . . . ]) Load dataset from multiple files in SVMlight format
datasets.load_wine([return_X_y]) Load and return the wine dataset (classification).
sklearn.datasets.clear_data_home
sklearn.datasets.clear_data_home(data_home=None)
Delete all the content of the data home cache.
Parameters
data_home [str | None] The path to scikit-learn data dir.
sklearn.datasets.dump_svmlight_file
sklearn.datasets.dump_svmlight_file(X, y, f, zero_based=True, comment=None,

query_id=None, multilabel=False)
Dump the dataset in svmlight / libsvm file format.
This format is a text-based format, with one sample per line. It does not store zero valued features hence is
suitable for sparse dataset.
The first element of each line can be used to store a target variable to predict.
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] Training vectors, where
y [{array-like, sparse matrix}, shape = [n_samples (, n_labels)]] Target values. Class labels
must be an integer or float, or array-like objects of integer or float for multilabel classifica-
tions.
f [string or file-like in binary mode] If string, specifies the path that will contain the data. If
file-like, data will be written to f. f should be opened in binary mode.

zero_based [boolean, optional] Whether column indices should be written zero-based (True) or
one-based (False).
comment [string, optional] Comment to insert at the top of the file. This should be either a
Unicode string, which will be encoded as UTF-8, or an ASCII byte string. If a comment
is given, then it will be preceded by one that identifies the file as having been dumped by
scikit-learn. Note that not all tools grok comments in SVMlight files.
query_id [array-like, shape = [n_samples]] Array containing pairwise preference constraints
(qid in svmlight format).
multilabel [boolean, optional] Samples may have several labels each (see https://www.csie.ntu.
edu.tw/~cjlin/libsvmtools/datasets/multilabel.html)
New in version 0.17: parameter multilabel to support multilabel datasets.
Examples using sklearn.datasets.dump_svmlight_file
• Libsvm GUI
sklearn.datasets.fetch_20newsgroups
sklearn.datasets.fetch_20newsgroups(data_home=None, subset=’train’, categories=None,

shuffle=True, random_state=42, remove=(), down-
load_if_missing=True)
Load the filenames and data from the 20 newsgroups dataset (classification).
Download it if necessary.
Classes 20
Samples total 18846
Dimensionality 1
Features text

Parameters
data_home [optional, default: None] Specify a download and cache folder for the datasets. If
None, all scikit-learn data is stored in ‘~/scikit_learn_data’ subfolders.
subset [‘train’ or ‘test’, ‘all’, optional] Select the dataset to load: ‘train’ for the training set,
‘test’ for the test set, ‘all’ for both, with shuffled ordering.
categories [None or collection of string or unicode] If None (default), load all the categories. If
not None, list of category names to load (other categories ignored).
shuffle [bool, optional] Whether or not to shuffle the data: might be important for models that
make the assumption that the samples are independent and identically distributed (i.i.d.),
such as stochastic gradient descent.
eration for dataset shuffling. Pass an int for reproducible output across multiple function
calls. See Glossary.
remove [tuple] May contain any subset of (‘headers’, ‘footers’, ‘quotes’). Each of these are
kinds of text that will be detected and removed from the newsgroup posts, preventing clas-
sifiers from overfitting on metadata.

‘headers’ removes newsgroup headers, ‘footers’ removes blocks at the ends of posts that
look like signatures, and ‘quotes’ removes lines that appear to be quoting another post.
‘headers’ follows an exact standard; the other filters are not always correct.
download_if_missing [optional, True by default] If False, raise an IOError if the data is not
locally available instead of trying to download the data from the source site.
Returns
bunch [Bunch object with the following attribute:]
• bunch.data: list, length [n_samples]
• bunch.target: array, shape [n_samples]
• bunch.filenames: list, length [n_samples]
• bunch.DESCR: a description of the dataset.
• bunch.target_names: a list of categories of the returned data, length [n_classes]. This
depends on the categories parameter.
Examples using sklearn.datasets.fetch_20newsgroups
• FeatureHasher and DictVectorizer Comparison
sklearn.datasets.fetch_20newsgroups_vectorized
sklearn.datasets.fetch_20newsgroups_vectorized(subset=’train’, remove=(),
data_home=None, down-
load_if_missing=True, re-
turn_X_y=False)
Load the 20 newsgroups dataset and vectorize it into token counts (classification).
This is a convenience function; the transformation is done using the default settings for sklearn.
feature_extraction.text.CountVectorizer. For more advanced usage (stopword
filtering, n-gram extraction, etc.), combine fetch_20newsgroups with a custom sklearn.
feature_extraction.text.CountVectorizer, sklearn.feature_extraction.text.
HashingVectorizer, sklearn.feature_extraction.text.TfidfTransformer or
sklearn.feature_extraction.text.TfidfVectorizer.
Classes 20
Samples total 18846
Features real


Parameters
subset [‘train’ or ‘test’, ‘all’, optional] Select the dataset to load: ‘train’ for the training set,
‘test’ for the test set, ‘all’ for both, with shuffled ordering.
remove [tuple] May contain any subset of (‘headers’, ‘footers’, ‘quotes’). Each of these are
kinds of text that will be detected and removed from the newsgroup posts, preventing clas-
sifiers from overfitting on metadata.
‘headers’ removes newsgroup headers, ‘footers’ removes blocks at the ends of posts that
look like signatures, and ‘quotes’ removes lines that appear to be quoting another post.
data_home [optional, default: None] Specify an download and cache folder for the datasets. If
None, all scikit-learn data is stored in ‘~/scikit_learn_data’ subfolders.
download_if_missing [optional, True by default] If False, raise an IOError if the data is not
return_X_y [boolean, default=False.] If True, returns (data.data, data.target) in-
stead of a Bunch object.
Returns
bunch [Bunch object with the following attribute:]
• bunch.data: sparse matrix, shape [n_samples, n_features]
• bunch.target: array, shape [n_samples]
• bunch.target_names: a list of categories of the returned data, length [n_classes].
• bunch.DESCR: a description of the dataset.
(data, target) [tuple if return_X_y is True] New in version 0.20.
Examples using sklearn.datasets.fetch_20newsgroups_vectorized

sklearn.datasets.fetch_california_housing
sklearn.datasets.fetch_california_housing(data_home=None, download_if_missing=True,
return_X_y=False)
Load the California housing dataset (regression).
Samples total 20640

Dimensionality 8
Features real
Target real 0.15 - 5.

Parameters

data_home [optional, default: None] Specify another download and cache folder for the
datasets. By default all scikit-learn data is stored in ‘~/scikit_learn_data’ subfolders.
download_if_missing [optional, default=True] If False, raise a IOError if the data is not locally
available instead of trying to download the data from the source site.
Returns
dataset [dict-like object with the following attributes:]
dataset.data [ndarray, shape [20640, 8]] Each row corresponding to the 8 feature values in
order.
dataset.target [numpy array of shape (20640,)] Each value corresponds to the average house
value in units of 100,000.
dataset.feature_names [array of length 8] Array of ordered feature names used in the dataset.
dataset.DESCR [string] Description of the California housing dataset.
Notes
This dataset consists of 20,640 samples and 9 features.
Examples using sklearn.datasets.fetch_california_housing

• Compare the effect of different scalers on data with outliers
sklearn.datasets.fetch_covtype
sklearn.datasets.fetch_covtype(data_home=None, download_if_missing=True, ran-

dom_state=None, shuffle=False, return_X_y=False)
Load the covertype dataset (classification).
Classes 7
Dimensionality 54
Features int

Parameters
data_home [string, optional] Specify another download and cache folder for the datasets. By
default all scikit-learn data is stored in ‘~/scikit_learn_data’ subfolders.

download_if_missing [boolean, default=True] If False, raise a IOError if the data is not locally
shuffle [bool, default=False] Whether to shuffle dataset.
Returns
dataset.data [numpy array of shape (581012, 54)] Each row corresponds to the 54 features in
the dataset.
dataset.target [numpy array of shape (581012,)] Each value corresponds to one of the 7 forest
covertypes with values ranging between 1 to 7.
dataset.DESCR [string] Description of the forest covertype dataset.
sklearn.datasets.fetch_kddcup99
sklearn.datasets.fetch_kddcup99(subset=None, data_home=None, shuffle=False,

random_state=None, percent10=True, down-
load_if_missing=True, return_X_y=False)
Load the kddcup99 dataset (classification).
Classes 23
Dimensionality 41

Parameters
subset [None, ‘SA’, ‘SF’, ‘http’, ‘smtp’] To return the corresponding classical subsets of kddcup
99. If None, return the entire kddcup 99 dataset.
default all scikit-learn data is stored in ‘~/scikit_learn_data’ subfolders. .. versionadded::
0.19
eration for dataset shuffling and for selection of abnormal samples if subset=’SA’. Pass an
int for reproducible output across multiple function calls. See Glossary.
percent10 [bool, default=True] Whether to load only 10 percent of the data.

download_if_missing [bool, default=True] If False, raise a IOError if the data is not locally
return_X_y [boolean, default=False.] If True, returns (data, target) instead of a Bunch
object. See below for more information about the data and target object.
Returns
data [Bunch]
Dictionary-like object, the interesting attributes are:
• ‘data’, the data to learn.
• ‘target’, the regression target for each sample.
• ‘DESCR’, a description of the dataset.
sklearn.datasets.fetch_lfw_pairs
sklearn.datasets.fetch_lfw_pairs(subset=’train’, data_home=None, funneled=True, re-

size=0.5, color=False, slice_=(slice(70, 195, None), slice(78,
172, None)), download_if_missing=True)
Load the Labeled Faces in the Wild (LFW) pairs dataset (classification).
Classes 5749
Samples total 13233
Dimensionality 5828
In the official README.txt this task is described as the “Restricted” task. As I am not sure as to implement the
“Unrestricted” variant correctly, I left it as unsupported for now.
The original images are 250 x 250 pixels, but the default slice and resize arguments reduce them to 62 x 47.
Parameters
subset [optional, default: ‘train’] Select the dataset to load: ‘train’ for the development training
set, ‘test’ for the development test set, and ‘10_folds’ for the official evaluation set that is
meant to be used with a 10-folds cross validation.
funneled [boolean, optional, default: True] Download and use the funneled variant of the
dataset.
resize [float, optional, default 0.5] Ratio used to resize the each face picture.
color [boolean, optional, default False] Keep the 3 RGB channels instead of averaging them to
a single gray level channel. If color is True the shape of the data has one more dimension
than the shape with color = False.

slice_ [optional] Provide a custom 2D slice (height, width) to extract the ‘interesting’ part of
the jpeg files and avoid use statistical correlation from the background
download_if_missing [optional, True by default] If False, raise a IOError if the data is not
Returns
The data is returned as a Bunch object with the following attributes:
data [numpy array of shape (2200, 5828). Shape depends on subset.] Each row corresponds
to 2 ravel’d face images of original size 62 x 47 pixels. Changing the slice_, resize or
subset parameters will change the shape of the output.
pairs [numpy array of shape (2200, 2, 62, 47). Shape depends on subset] Each row has 2
face images corresponding to same or different person from the dataset containing 5749
people. Changing the slice_, resize or subset parameters will change the shape of
the output.
target [numpy array of shape (2200,). Shape depends on subset.] Labels associated to each
pair of images. The two label values being different persons or the same person.
DESCR [string] Description of the Labeled Faces in the Wild (LFW) dataset.
sklearn.datasets.fetch_lfw_people
sklearn.datasets.fetch_lfw_people(data_home=None, funneled=True, resize=0.5,

min_faces_per_person=0, color=False, slice_=(slice(70,
195, None), slice(78, 172, None)), down-
load_if_missing=True, return_X_y=False)
Load the Labeled Faces in the Wild (LFW) people dataset (classification).
Classes 5749
Samples total 13233
Dimensionality 5828

Parameters
funneled [boolean, optional, default: True] Download and use the funneled variant of the
dataset.
resize [float, optional, default 0.5] Ratio used to resize the each face picture.
min_faces_per_person [int, optional, default None] The extracted dataset will only retain pic-
tures of people that have at least min_faces_per_person different pictures.
color [boolean, optional, default False] Keep the 3 RGB channels instead of averaging them to
a single gray level channel. If color is True the shape of the data has one more dimension
than the shape with color = False.
slice_ [optional] Provide a custom 2D slice (height, width) to extract the ‘interesting’ part of
the jpeg files and avoid use statistical correlation from the background

return_X_y [boolean, default=False.] If True, returns (dataset.data, dataset.
target) instead of a Bunch object. See below for more information about the dataset.data
and dataset.target object.
Returns
dataset.data [numpy array of shape (13233, 2914)] Each row corresponds to a ravelled face
image of original size 62 x 47 pixels. Changing the slice_ or resize parameters will
change the shape of the output.
dataset.images [numpy array of shape (13233, 62, 47)] Each row is a face image corresponding
to one of the 5749 people in the dataset. Changing the slice_ or resize parameters will
change the shape of the output.
dataset.target [numpy array of shape (13233,)] Labels associated to each face image. Those
labels range from 0-5748 and correspond to the person IDs.
dataset.DESCR [string] Description of the Labeled Faces in the Wild (LFW) dataset.
Examples using sklearn.datasets.fetch_lfw_people
sklearn.datasets.fetch_olivetti_faces
sklearn.datasets.fetch_olivetti_faces(data_home=None, shuffle=False, random_state=0,

download_if_missing=True)
Load the Olivetti faces data-set from AT&T (classification).
Classes 40
Samples total 400
Dimensionality 4096

Parameters
shuffle [boolean, optional] If True the order of the dataset is shuffled to avoid having images of
the same person grouped.
random_state [int, RandomState instance or None (default=0)] Determines random number
generation for dataset shuffling. Pass an int for reproducible output across multiple function

Returns
An object with the following attributes:
data [numpy array of shape (400, 4096)] Each row corresponds to a ravelled face image of
original size 64 x 64 pixels.
images [numpy array of shape (400, 64, 64)] Each row is a face image corresponding to one of
the 40 subjects of the dataset.
target [numpy array of shape (400, )] Labels associated to each face image. Those labels are
ranging from 0-39 and correspond to the Subject IDs.
DESCR [string] Description of the modified Olivetti Faces Dataset.
Examples using sklearn.datasets.fetch_olivetti_faces

sklearn.datasets.fetch_openml
sklearn.datasets.fetch_openml(name=None, version=’active’, data_id=None, data_home=None,

target_column=’default-target’, cache=True, return_X_y=False)
Fetch dataset from openml by name or dataset id.
Datasets are uniquely identified by either an integer ID or by a combination of name and version (i.e. there
might be multiple versions of the ‘iris’ dataset). Please give either name or data_id (not both). In case a name is
given, a version can also be provided.
Note: EXPERIMENTAL
The API is experimental in version 0.20 (particularly the return value structure), and might have small backward-
incompatible changes in future releases.
Parameters
name [str or None] String identifier of the dataset. Note that OpenML can have multiple
datasets with the same name.
version [integer or ‘active’, default=’active’] Version of the dataset. Can only be provided if
also name is given. If ‘active’ the oldest version that’s still active is used. Since there
may be more than one active version of a dataset, and those versions may fundamentally be
different from one another, setting an exact version is highly recommended.
data_id [int or None] OpenML ID of the dataset. The most specific way of retrieving a dataset.
If data_id is not given, name (and potential version) are used to obtain a dataset.

data_home [string or None, default None] Specify another download and cache folder for the
data sets. By default all scikit-learn data is stored in ‘~/scikit_learn_data’ subfolders.
target_column [string, list or None, default ‘default-target’] Specify the column name in the
data to use as target. If ‘default-target’, the standard target column a stored on the server is
used. If None, all columns are returned as data and the target is None. If list (of strings), all
columns with these names are returned as multi-target (Note: not all scikit-learn classifiers
can handle all types of multi-output combinations)
cache [boolean, default=True] Whether to cache downloaded datasets using joblib.
object. See below for more information about the data and target objects.
Returns
data [Bunch] Dictionary-like object, with attributes:
data [np.array or scipy.sparse.csr_matrix of floats] The feature matrix. Categorical features
are encoded as ordinals.
target [np.array] The regression target or classification labels, if applicable. Dtype is float
if numeric, and object if categorical.
DESCR [str] The full description of the dataset
feature_names [list] The names of the dataset columns
categories [dict] Maps each categorical feature name to a list of values, such that the value
encoded as i is ith in the list.
details [dict] More metadata from OpenML
(data, target) [tuple if return_X_y is True]
Note: EXPERIMENTAL
This interface is experimental as at version 0.20 and subsequent releases may change
attributes without notice (although there should only be minor changes to data and
target).
Missing values in the ‘data’ are represented as NaN’s. Missing values in ‘target’ are repre-
sented as NaN’s (numerical target) or None (categorical target)
Examples using sklearn.datasets.fetch_openml
• Gaussian process regression (GPR) on Mauna Loa CO2 data.

• Early stopping of Stochastic Gradient Descent
• Classifier Chain

sklearn.datasets.fetch_rcv1
sklearn.datasets.fetch_rcv1(data_home=None, subset=’all’, download_if_missing=True, ran-

dom_state=None, shuffle=False, return_X_y=False)
Load the RCV1 multilabel dataset (classification).
Version: RCV1-v2, vectors, full sets, topics multilabels.
Classes 103

Parameters
default all scikit-learn data is stored in ‘~/scikit_learn_data’ subfolders.
subset [string, ‘train’, ‘test’, or ‘all’, default=’all’] Select the dataset to load: ‘train’ for the
training set (23149 samples), ‘test’ for the test set (781265 samples), ‘all’ for both, with the
training samples first if shuffle is False. This follows the official LYRL2004 chronological
split.
download_if_missing [boolean, default=True] If False, raise a IOError if the data is not locally
return_X_y [boolean, default=False.] If True, returns (dataset.data, dataset.
target) instead of a Bunch object. See below for more information about the dataset.data
and dataset.target object.
Returns
dataset.data [scipy csr array, dtype np.float64, shape (804414, 47236)] The array has 0.16% of
non zero values.
dataset.target [scipy csr array, dtype np.uint8, shape (804414, 103)] Each sample has a value
of 1 in its categories, and 0 in others. The array has 3.15% of non zero values.
dataset.sample_id [numpy array, dtype np.uint32, shape (804414,)] Identification number of
each sample, as ordered in dataset.data.
dataset.target_names [numpy array, dtype object, length (103)] Names of each target (RCV1
topics), as ordered in dataset.target.
dataset.DESCR [string] Description of the RCV1 dataset.

sklearn.datasets.fetch_species_distributions
sklearn.datasets.fetch_species_distributions(data_home=None, down-
load_if_missing=True)
Loader for species distribution dataset from Phillips et. al. (2006)
Parameters
Returns
The data is returned as a Bunch object with the following attributes:
coverages [array, shape = [14, 1592, 1212]] These represent the 14 features measured at each
point of the map grid. The latitude/longitude values for the grid are discussed below. Miss-
ing data is represented by the value -9999.
train [record array, shape = (1624,)] The training points for the data. Each point has three fields:
• train[‘species’] is the species name
• train[‘dd long’] is the longitude, in degrees
• train[‘dd lat’] is the latitude, in degrees
test [record array, shape = (620,)] The test points for the data. Same format as the training data.
Nx, Ny [integers] The number of longitudes (x) and latitudes (y) in the grid
x_left_lower_corner, y_left_lower_corner [floats] The (x,y) position of the lower-left corner,
in degrees
grid_size [float] The spacing between points of the grid, in degrees
Notes
This dataset represents the geographic distribution of species. The dataset is provided by Phillips et. al. (2006).
• For an example of using this dataset with scikit-learn, see exam-
ples/applications/plot_species_distribution_modeling.py.
References
• “Maximum entropy modeling of species geographic distributions” S. J. Phillips, R. P. Anderson, R. E.

Schapire - Ecological Modelling, 190:231-259, 2006.

Examples using sklearn.datasets.fetch_species_distributions

• Kernel Density Estimate of Species Distributions
sklearn.datasets.get_data_home
sklearn.datasets.get_data_home(data_home=None)
Return the path of the scikit-learn data dir.
This folder is used by some large dataset loaders to avoid downloading the data several times.
By default the data dir is set to a folder named ‘scikit_learn_data’ in the user home folder.
Alternatively, it can be set by the ‘SCIKIT_LEARN_DATA’ environment variable or programmatically by giving
an explicit folder path. The ‘~’ symbol is expanded to the user home folder.
If the folder does not already exist, it is automatically created.
Parameters
data_home [str | None] The path to scikit-learn data dir.
Examples using sklearn.datasets.get_data_home
• Out-of-core classification of text documents
sklearn.datasets.load_boston
sklearn.datasets.load_boston(return_X_y=False)
Load and return the boston house-prices dataset (regression).
Samples total 506

Dimensionality 13
Features real, positive
Targets real 5. - 50.

Parameters
Returns
data [Bunch] Dictionary-like object, the interesting attributes are: ‘data’, the data to learn,
‘target’, the regression targets, ‘DESCR’, the full description of the dataset, and ‘filename’,
the physical location of boston csv dataset (added in version 0.20).

Notes
Changed in version 0.20: Fixed a wrong data point at [445, 0].
Examples

>>> print(boston.data.shape)
(506, 13)
Examples using sklearn.datasets.load_boston

• Plotting Cross-Validated Predictions
sklearn.datasets.load_breast_cancer
sklearn.datasets.load_breast_cancer(return_X_y=False)
Load and return the breast cancer wisconsin dataset (classification).
The breast cancer dataset is a classic and very easy binary classification dataset.
Classes 2
Samples per class 212(M),357(B)
Samples total 569
Dimensionality 30

Parameters
return_X_y [boolean, default=False] If True, returns (data, target) instead of a Bunch
Returns
‘target’, the classification labels, ‘target_names’, the meaning of the labels, ‘feature_names’,
the meaning of the features, and ‘DESCR’, the full description of the dataset, ‘filename’, the
physical location of breast cancer csv dataset (added in version 0.20).


The copy of UCI ML Breast Cancer Wisconsin (Diagnostic) dataset is
downloaded from:
https://goo.gl/U2Uwz2
Examples
Let’s say you are interested in the samples 10, 50, and 85, and want to know their class name.

>>> data = load_breast_cancer()
>>> data.target[[10, 50, 85]]
array([0, 1, 0])
>>> list(data.target_names)
['malignant', 'benign']
sklearn.datasets.load_diabetes
sklearn.datasets.load_diabetes(return_X_y=False)
Load and return the diabetes dataset (regression).
Samples total 442

Dimensionality 10
Features real, -.2 < x < .2
Targets integer 25 - 346

Parameters
Returns
data [Bunch] Dictionary-like object, the interesting attributes are: ‘data’, the data to learn, ‘tar-
get’, the regression target for each sample, ‘data_filename’, the physical location of diabetes
data csv dataset, and ‘target_filename’, the physical location of diabetes targets csv datataset
(added in version 0.20).
Examples using sklearn.datasets.load_diabetes


• Sparsity Example: Fitting only features 1 and 2

sklearn.datasets.load_digits
sklearn.datasets.load_digits(n_class=10, return_X_y=False)
Load and return the digits dataset (classification).
Each datapoint is a 8x8 image of a digit.
Classes 10
Samples per class ~180
Samples total 1797
Dimensionality 64
Features integers 0-16

Parameters
n_class [integer, between 0 and 10, optional (default=10)] The number of classes to return.
Returns
‘images’, the images corresponding to each sample, ‘target’, the classification labels for
each sample, ‘target_names’, the meaning of the labels, and ‘DESCR’, the full description
of the dataset.
This is a copy of the test set of the UCI ML hand-written digits datasets
http://archive.ics.uci.edu/ml/datasets/Optical+Recognition+of+Handwritten+Digits
Examples
To load the data and visualize the images:

>>> digits = load_digits()
>>> print(digits.data.shape)
(1797, 64)
>>> plt.gray()
>>> plt.matshow(digits.images[0])
>>> plt.show()

Examples using sklearn.datasets.load_digits

• Recognizing hand-written digits
• The Digit Dataset
• Digits Classification Exercise
• Cross-validation on Digits Dataset Exercise
• Recursive feature elimination
• Manifold learning on handwritten digits: Locally Linear Embedding, Isomap. . .
• Parameter estimation using grid search with cross-validation
• Kernel Density Estimation
sklearn.datasets.load_files
sklearn.datasets.load_files(container_path, description=None, categories=None,

load_content=True, shuffle=True, encoding=None, de-
code_error=’strict’, random_state=0)
Load text files with categories as subfolder names.
Individual samples are assumed to be files stored a two levels folder structure such as the following:
container_folder/
category_1_folder/ file_1.txt file_2.txt . . . file_42.txt
category_2_folder/ file_43.txt file_44.txt . . .

The folder names are used as supervised signal label names. The individual file names are not important.
This function does not try to extract features into a numpy array or scipy sparse matrix. In addition, if
load_content is false it does not try to load the files in memory.
To use text files in a scikit-learn classification or clustering algorithm, you will need to use the
sklearn.feature_extraction.text module to build a feature extraction transformer that suits your problem.
If you set load_content=True, you should also specify the encoding of the text using the ‘encoding’ parame-
ter. For many modern text files, ‘utf-8’ will be the correct encoding. If you leave encoding equal to None,
then the content will be made of bytes instead of Unicode, and you will not be able to use most functions in
sklearn.feature_extraction.text.
Similar feature extractors should be built for other kind of unstructured data input such as images, audio, video,
...
Parameters
container_path [string or unicode] Path to the main folder holding one subfolder per category
description [string or unicode, optional (default=None)] A paragraph describing the character-
istic of the dataset: its source, reference, etc.
categories [A collection of strings or None, optional (default=None)] If None (default), load all
the categories. If not None, list of category names to load (other categories ignored).
load_content [boolean, optional (default=True)] Whether to load or not the content of the dif-
ferent files. If true a ‘data’ attribute containing the text information is present in the data
structure returned. If not, a filenames attribute gives the path to the files.
shuffle [bool, optional (default=True)] Whether or not to shuffle the data: might be important
for models that make the assumption that the samples are independent and identically dis-
tributed (i.i.d.), such as stochastic gradient descent.
encoding [string or None (default is None)] If None, do not try to decode the content of the files
(e.g. for images or other non-text content). If not None, encoding to use to decode text files
to Unicode if load_content is True.
decode_error [{‘strict’, ‘ignore’, ‘replace’}, optional] Instruction on what to do if a byte se-
quence is given to analyze that contains characters not of the given encoding. Passed as
keyword argument ‘errors’ to bytes.decode.
random_state [int, RandomState instance or None (default=0)] Determines random number
generation for dataset shuffling. Pass an int for reproducible output across multiple function
Returns
data [Bunch] Dictionary-like object, the interesting attributes are: either data, the raw text data
to learn, or ‘filenames’, the files holding it, ‘target’, the classification labels (integer index),
‘target_names’, the meaning of the labels, and ‘DESCR’, the full description of the dataset.
sklearn.datasets.load_iris
sklearn.datasets.load_iris(return_X_y=False)
Load and return the iris dataset (classification).
The iris dataset is a classic and very easy multi-class classification dataset.

Classes 3
Samples per class 50
Samples total 150
Dimensionality 4

Parameters
Returns
the meaning of the features, ‘DESCR’, the full description of the dataset, ‘filename’, the
physical location of iris csv dataset (added in version 0.20).
Notes
Changed in version 0.20: Fixed two wrong data points according to Fisher’s paper. The new version is the same
as in R, but not as in the UCI Machine Learning Repository.
Examples
>>> data = load_iris()
>>> data.target[[10, 25, 50]]
array([0, 0, 1])
['setosa', 'versicolor', 'virginica']
Examples using sklearn.datasets.load_iris
• Plot classification probability

• K-means Clustering
• The Iris Dataset
• PCA example with Iris Data-set
• Incremental PCA
• Plot the decision boundaries of a VotingClassifier


• SVM Exercise
• Test with permutations the significance of a classification score
• Gaussian process classification (GPC) on iris dataset
• Logistic Regression 3-class Classifier
• GMM covariances
• Receiver Operating Characteristic (ROC) with cross validation
• Nested versus non-nested cross-validation
• Confusion matrix
• Receiver Operating Characteristic (ROC)
• Precision-Recall
• Nearest Neighbors Classification
• Nearest Centroid Classification
• SVM with custom kernel
• Understanding the decision tree structure
sklearn.datasets.load_linnerud
sklearn.datasets.load_linnerud(return_X_y=False)
Load and return the linnerud dataset (multivariate regression).
Samples total 20
Dimensionality 3 (for both data and target)
Features integer
Targets integer

Parameters


Returns
data [Bunch] Dictionary-like object, the interesting attributes are: ‘data’ and ‘targets’, the two
multivariate datasets, with ‘data’ corresponding to the exercise and ‘targets’ corresponding
to the physiological measurements, as well as ‘feature_names’ and ‘target_names’. In addi-
tion, you will also have access to ‘data_filename’, the physical location of linnerud data csv
dataset, and ‘target_filename’, the physical location of linnerud targets csv datataset (added
in version 0.20).
sklearn.datasets.load_sample_image
sklearn.datasets.load_sample_image(image_name)
Load the numpy array of a single sample image
Parameters
image_name [{china.jpg, flower.jpg}] The name of the sample image loaded
Returns
img [3D array] The image as a numpy array: height x width x color
Examples
>>> from sklearn.datasets import load_sample_image

>>> china = load_sample_image('china.jpg')
>>> china.dtype
dtype('uint8')
>>> china.shape
(427, 640, 3)
>>> flower = load_sample_image('flower.jpg')
>>> flower.dtype
dtype('uint8')
>>> flower.shape
(427, 640, 3)
Examples using sklearn.datasets.load_sample_image
sklearn.datasets.load_sample_images
sklearn.datasets.load_sample_images()
Load sample images for image manipulation.
Loads both, china and flower.


Returns
data [Bunch] Dictionary-like object with the following attributes : ‘images’, the two sample
images, ‘filenames’, the file names for the images, and ‘DESCR’ the full description of the
dataset.
Examples
To load the data and visualize the images:
>>> from sklearn.datasets import load_sample_images

>>> dataset = load_sample_images()
>>> len(dataset.images)
2
>>> first_img_data = dataset.images[0]
>>> first_img_data.shape
(427, 640, 3)
>>> first_img_data.dtype
dtype('uint8')
sklearn.datasets.load_svmlight_file
sklearn.datasets.load_svmlight_file(f, n_features=None, dtype=<class ‘numpy.float64’>,

multilabel=False, zero_based=’auto’, query_id=False,
offset=0, length=-1)
Load datasets in the svmlight / libsvm format into sparse CSR matrix
This format is a text-based format, with one sample per line. It does not store zero valued features hence is
suitable for sparse dataset.
The first element of each line can be used to store a target variable to predict.
This format is used as the default format for both svmlight and the libsvm command line programs.
Parsing a text based source can be expensive. When working on repeatedly on the same dataset, it is recom-
mended to wrap this loader with joblib.Memory.cache to store a memmapped backup of the CSR results of the
first call and benefit from the near instantaneous loading of memmapped structures for the subsequent calls.
In case the file contains a pairwise preference constraint (known as “qid” in the svmlight format) these are
ignored unless the query_id parameter is set to True. These pairwise preference constraints can be used to
constraint the combination of samples when using pairwise loss functions (as is the case in some learning to
rank problems) so that only pairs with the same query_id value are considered.
This implementation is written in Cython and is reasonably fast. However, a faster API-compatible loader is
also available at:
https://github.com/mblondel/svmlight-loader
Parameters
f [{str, file-like, int}] (Path to) a file to load. If a path ends in “.gz” or “.bz2”, it will be uncom-
pressed on the fly. If an integer is passed, it is assumed to be a file descriptor. A file-like
or file descriptor will not be closed by this function. A file-like object must be opened in
binary mode.

n_features [int or None] The number of features to use. If None, it will be inferred. This
argument is useful to load several files that are subsets of a bigger sliced dataset: each
subset might not have examples of every feature, hence the inferred shape might vary from
one slice to another. n_features is only required if offset or length are passed a non-
default value.
dtype [numpy data type, default np.float64] Data type of dataset to be loaded. This will be the
data type of the output numpy arrays X and y.
multilabel [boolean, optional, default False] Samples may have several labels each (see https:
//www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multilabel.html)
zero_based [boolean or “auto”, optional, default “auto”] Whether column indices in f are zero-
based (True) or one-based (False). If column indices are one-based, they are transformed
to zero-based to match Python/NumPy conventions. If set to “auto”, a heuristic check is
applied to determine this from the file contents. Both kinds of files occur “in the wild”, but
they are unfortunately not self-identifying. Using “auto” or True should always be safe when
no offset or length is passed. If offset or length are passed, the “auto” mode falls
back to zero_based=True to avoid having the heuristic check yield inconsistent results
on different segments of the file.
query_id [boolean, default False] If True, will return the query_id array for each file.
offset [integer, optional, default 0] Ignore the offset first bytes by seeking forward, then dis-
carding the following bytes up until the next new line character.
length [integer, optional, default -1] If strictly positive, stop reading any new line of data once
the position in the file has reached the (offset + length) bytes threshold.
Returns
X [scipy.sparse matrix of shape (n_samples, n_features)]
y [ndarray of shape (n_samples,), or, in the multilabel a list of] tuples of length n_samples.
query_id [array of shape (n_samples,)] query_id for each sample. Only returned when query_id
is set to True.
See also:
load_svmlight_files similar function for loading multiple files in this
format, enforcing
Examples
To use joblib.Memory to cache the svmlight file:

from sklearn.datasets import load_svmlight_file
mem = Memory("./mycache")
@mem.cache
def get_data():
data = load_svmlight_file("mysvmlightfile")
return data[0], data[1]
X, y = get_data()

sklearn.datasets.load_svmlight_files
sklearn.datasets.load_svmlight_files(files, n_features=None, dtype=<class

‘numpy.float64’>, multilabel=False,
zero_based=’auto’, query_id=False, offset=0, length=-
1)
Load dataset from multiple files in SVMlight format
This function is equivalent to mapping load_svmlight_file over a list of files, except that the results are concate-
nated into a single, flat list and the samples vectors are constrained to all have the same number of features.
In case the file contains a pairwise preference constraint (known as “qid” in the svmlight format) these are
ignored unless the query_id parameter is set to True. These pairwise preference constraints can be used to
constraint the combination of samples when using pairwise loss functions (as is the case in some learning to
rank problems) so that only pairs with the same query_id value are considered.
Parameters
files [iterable over {str, file-like, int}] (Paths of) files to load. If a path ends in “.gz” or “.bz2”, it
will be uncompressed on the fly. If an integer is passed, it is assumed to be a file descriptor.
File-likes and file descriptors will not be closed by this function. File-like objects must be
opened in binary mode.
n_features [int or None] The number of features to use. If None, it will be inferred from the
maximum column index occurring in any of the files.
This can be set to a higher value than the actual number of features in any of the input files,
but setting it to a lower value will cause an exception to be raised.
dtype [numpy data type, default np.float64] Data type of dataset to be loaded. This will be the
data type of the output numpy arrays X and y.
multilabel [boolean, optional] Samples may have several labels each (see https://www.csie.ntu.
edu.tw/~cjlin/libsvmtools/datasets/multilabel.html)
zero_based [boolean or “auto”, optional] Whether column indices in f are zero-based (True) or
one-based (False). If column indices are one-based, they are transformed to zero-based to
match Python/NumPy conventions. If set to “auto”, a heuristic check is applied to determine
this from the file contents. Both kinds of files occur “in the wild”, but they are unfortunately
not self-identifying. Using “auto” or True should always be safe when no offset or length
is passed. If offset or length are passed, the “auto” mode falls back to zero_based=True to
avoid having the heuristic check yield inconsistent results on different segments of the file.
query_id [boolean, defaults to False] If True, will return the query_id array for each file.
offset [integer, optional, default 0] Ignore the offset first bytes by seeking forward, then dis-
carding the following bytes up until the next new line character.
length [integer, optional, default -1] If strictly positive, stop reading any new line of data once
the position in the file has reached the (offset + length) bytes threshold.
Returns
[X1, y1, . . . , Xn, yn]
where each (Xi, yi) pair is the result from load_svmlight_file(files[i]).
If query_id is set to True, this will return instead [X1, y1, q1,
. . . , Xn, yn, qn] where (Xi, yi, qi) is the result from
load_svmlight_file(files[i])

See also:
load_svmlight_file
Notes
When fitting a model to a matrix X_train and evaluating it against a matrix X_test, it is essential that X_train
and X_test have the same number of features (X_train.shape[1] == X_test.shape[1]). This may not be the case
if you load the files individually with load_svmlight_file.
sklearn.datasets.load_wine
sklearn.datasets.load_wine(return_X_y=False)
Load and return the wine dataset (classification).
The wine dataset is a classic and very easy multi-class classification dataset.
Classes 3
Samples per class [59,71,48]
Samples total 178
Dimensionality 13

Parameters
Returns
the meaning of the features, and ‘DESCR’, the full description of the dataset.
(data, target) [tuple if return_X_y is True]
The copy of UCI ML Wine Data Set dataset is downloaded and modified to fit
standard format from:
https://archive.ics.uci.edu/ml/machine-learning-databases/wine/wine.data
Examples
>>> from sklearn.datasets import load_wine

>>> data = load_wine()
>>> data.target[[10, 80, 140]]
array([0, 1, 2])
['class_0', 'class_1', 'class_2']

Examples using sklearn.datasets.load_wine
• Importance of Feature Scaling
6.8.2 Samples generator
datasets.make_biclusters(shape, n_clusters) Generate an array with constant block diagonal structure

for biclustering.
datasets.make_blobs([n_samples, n_features, . . . ]) Generate isotropic Gaussian blobs for clustering.
datasets.make_checkerboard(shape, n_clusters) Generate an array with block checkerboard structure for bi-
clustering.
datasets.make_circles([n_samples, shuffle, . . . ]) Make a large circle containing a smaller circle in 2d.
datasets.make_classification([n_samples, Generate a random n-class classification problem.
. . . ])
datasets.make_friedman1([n_samples, . . . ]) Generate the “Friedman #1” regression problem
datasets.make_friedman2([n_samples, noise, Generate the “Friedman #2” regression problem
. . . ])
datasets.make_friedman3([n_samples, noise, Generate the “Friedman #3” regression problem
. . . ])
datasets.make_gaussian_quantiles([mean, Generate isotropic Gaussian and label samples by quantile
. . . ])
datasets.make_hastie_10_2([n_samples, . . . ]) Generates data for binary classification used in Hastie et al.
datasets.make_low_rank_matrix([n_samples, Generate a mostly low rank matrix with bell-shaped singu-
. . . ]) lar values
datasets.make_moons([n_samples, shuffle, . . . ]) Make two interleaving half circles
datasets.make_multilabel_classification([. .Generate . ]) a random multilabel classification problem.
datasets.make_regression([n_samples, . . . ]) Generate a random regression problem.
datasets.make_s_curve([n_samples, noise, . . . ]) Generate an S curve dataset.
datasets.make_sparse_coded_signal(n_samples, Generate a signal as a sparse combination of dictionary el-
...) ements.
datasets.make_sparse_spd_matrix([dim, . . . ]) Generate a sparse symmetric definite positive matrix.
datasets.make_sparse_uncorrelated([. . . ]) Generate a random regression problem with sparse uncor-
related design
datasets.make_spd_matrix(n_dim[, ran- Generate a random symmetric, positive-definite matrix.
dom_state])
datasets.make_swiss_roll([n_samples, noise, Generate a swiss roll dataset.
. . . ])
sklearn.datasets.make_biclusters
sklearn.datasets.make_biclusters(shape, n_clusters, noise=0.0, minval=10, maxval=100, shuf-

fle=True, random_state=None)
Generate an array with constant block diagonal structure for biclustering.
Parameters
shape [iterable (n_rows, n_cols)] The shape of the result.
n_clusters [integer] The number of biclusters.
noise [float, optional (default=0.0)] The standard deviation of the gaussian noise.

minval [int, optional (default=10)] Minimum value of a bicluster.

maxval [int, optional (default=100)] Maximum value of a bicluster.
shuffle [boolean, optional (default=True)] Shuffle the samples.
eration for dataset creation. Pass an int for reproducible output across multiple function
Returns
X [array of shape shape] The generated array.
rows [array of shape (n_clusters, X.shape[0],)] The indicators for cluster membership of each
row.
cols [array of shape (n_clusters, X.shape[1],)] The indicators for cluster membership of each
column.
See also:
make_checkerboard
References
[1]
Examples using sklearn.datasets.make_biclusters
• A demo of the Spectral Co-Clustering algorithm
sklearn.datasets.make_blobs
sklearn.datasets.make_blobs(n_samples=100, n_features=2, centers=None, cluster_std=1.0,

center_box=(-10.0, 10.0), shuffle=True, random_state=None)
Generate isotropic Gaussian blobs for clustering.
Parameters
n_samples [int or array-like, optional (default=100)] If int, it is the total number of points
equally divided among clusters. If array-like, each element of the sequence indicates the
number of samples per cluster.
n_features [int, optional (default=2)] The number of features for each sample.
centers [int or array of shape [n_centers, n_features], optional] (default=None) The number of
centers to generate, or the fixed center locations. If n_samples is an int and centers is None,
3 centers are generated. If n_samples is array-like, centers must be either None or an array
of length equal to the length of n_samples.
cluster_std [float or sequence of floats, optional (default=1.0)] The standard deviation of the
clusters.
center_box [pair of floats (min, max), optional (default=(-10.0, 10.0))] The bounding box for
each cluster center when centers are generated at random.

Returns
X [array of shape [n_samples, n_features]] The generated samples.
y [array of shape [n_samples]] The integer labels for cluster membership of each sample.
See also:
make_classification a more intricate variant
Examples
>>> from sklearn.datasets.samples_generator import make_blobs

>>> X, y = make_blobs(n_samples=10, centers=3, n_features=2,
... random_state=0)
>>> print(X.shape)
(10, 2)
>>> y
array([0, 0, 1, 0, 2, 2, 2, 1, 1, 0])
>>> X, y = make_blobs(n_samples=[3, 3, 4], centers=None, n_features=2,
... random_state=0)
>>> print(X.shape)
(10, 2)
>>> y
array([0, 1, 2, 0, 2, 2, 2, 1, 1, 0])
Examples using sklearn.datasets.make_blobs

• Normal and Shrinkage Linear Discriminant Analysis for classification
• Demonstration of k-means assumptions
• Plot randomly generated classification dataset
• SGD: Maximum margin separating hyperplane


• SVM: Maximum margin separating hyperplane
sklearn.datasets.make_checkerboard
sklearn.datasets.make_checkerboard(shape, n_clusters, noise=0.0, minval=10, maxval=100,

shuffle=True, random_state=None)
Generate an array with block checkerboard structure for biclustering.
Parameters
shape [iterable (n_rows, n_cols)] The shape of the result.
n_clusters [integer or iterable (n_row_clusters, n_column_clusters)] The number of row and
column clusters.
minval [int, optional (default=10)] Minimum value of a bicluster.
maxval [int, optional (default=100)] Maximum value of a bicluster.
Returns
X [array of shape shape] The generated array.
rows [array of shape (n_clusters, X.shape[0],)] The indicators for cluster membership of each
row.
cols [array of shape (n_clusters, X.shape[1],)] The indicators for cluster membership of each
column.
See also:
make_biclusters
References
[1]
Examples using sklearn.datasets.make_checkerboard
• A demo of the Spectral Biclustering algorithm

sklearn.datasets.make_circles
sklearn.datasets.make_circles(n_samples=100, shuffle=True, noise=None, random_state=None,

factor=0.8)
Make a large circle containing a smaller circle in 2d.
A simple toy dataset to visualize clustering and classification algorithms.
Parameters
n_samples [int, optional (default=100)] The total number of points generated. If odd, the inner
circle will have one point more than the outer circle.
shuffle [bool, optional (default=True)] Whether to shuffle the samples.
noise [double or None (default=None)] Standard deviation of Gaussian noise added to the data.
eration for dataset shuffling and noise. Pass an int for reproducible output across multiple
function calls. See Glossary.
factor [0 < double < 1 (default=.8)] Scale factor between inner and outer circle.
Returns
X [array of shape [n_samples, 2]] The generated samples.
y [array of shape [n_samples]] The integer labels (0 or 1) for class membership of each sample.
Examples using sklearn.datasets.make_circles
• Kernel PCA
• t-SNE: The effect of various perplexity values on the shape
sklearn.datasets.make_classification
sklearn.datasets.make_classification(n_samples=100, n_features=20, n_informative=2,

n_redundant=2, n_repeated=0, n_classes=2,
n_clusters_per_class=2, weights=None, flip_y=0.01,
class_sep=1.0, hypercube=True, shift=0.0, scale=1.0,
shuffle=True, random_state=None)
Generate a random n-class classification problem.

This initially creates clusters of points normally distributed (std=1) about vertices of an n_informative-
dimensional hypercube with sides of length 2*class_sep and assigns an equal number of clusters to each
class. It introduces interdependence between these features and adds various types of further noise to the data.
Without shuffling, X horizontally stacks features in the following order: the primary n_informative fea-
tures, followed by n_redundant linear combinations of the informative features, followed by n_repeated
duplicates, drawn randomly with replacement from the informative and redundant features. The remaining fea-
tures are filled with random noise. Thus, without shuffling, all useful features are contained in the columns
X[:, :n_informative + n_redundant + n_repeated].
Parameters
n_samples [int, optional (default=100)] The number of samples.
n_features [int, optional (default=20)] The total number of fea-
tures. These comprise n_informative informative features,
n_redundant redundant features, n_repeated duplicated features and
n_features-n_informative-n_redundant-n_repeated useless features
drawn at random.
n_informative [int, optional (default=2)] The number of informative features. Each class is
composed of a number of gaussian clusters each located around the vertices of a hypercube
in a subspace of dimension n_informative. For each cluster, informative features are
drawn independently from N(0, 1) and then randomly linearly combined within each cluster
in order to add covariance. The clusters are then placed on the vertices of the hypercube.
n_redundant [int, optional (default=2)] The number of redundant features. These features are
generated as random linear combinations of the informative features.
n_repeated [int, optional (default=0)] The number of duplicated features, drawn randomly
from the informative and the redundant features.
n_classes [int, optional (default=2)] The number of classes (or labels) of the classification prob-
lem.
n_clusters_per_class [int, optional (default=2)] The number of clusters per class.
weights [list of floats or None (default=None)] The proportions of samples assigned to each
class. If None, then classes are balanced. Note that if len(weights) == n_classes
- 1, then the last class weight is automatically inferred. More than n_samples samples
may be returned if the sum of weights exceeds 1.
flip_y [float, optional (default=0.01)] The fraction of samples whose class are randomly ex-
changed. Larger values introduce noise in the labels and make the classification task harder.
class_sep [float, optional (default=1.0)] The factor multiplying the hypercube size. Larger val-
ues spread out the clusters/classes and make the classification task easier.
hypercube [boolean, optional (default=True)] If True, the clusters are put on the vertices of a
hypercube. If False, the clusters are put on the vertices of a random polytope.
shift [float, array of shape [n_features] or None, optional (default=0.0)] Shift features by the
specified value. If None, then features are shifted by a random value drawn in [-class_sep,
class_sep].
scale [float, array of shape [n_features] or None, optional (default=1.0)] Multiply features by
the specified value. If None, then features are scaled by a random value drawn in [1, 100].
Note that scaling happens after shifting.
shuffle [boolean, optional (default=True)] Shuffle the samples and the features.

Returns
y [array of shape [n_samples]] The integer labels for class membership of each sample.
See also:
make_blobs simplified variant

make_multilabel_classification unrelated generator for multilabel tasks
Notes
The algorithm is adapted from Guyon [1] and was designed to generate the “Madelon” dataset.
References
[1]
Examples using sklearn.datasets.make_classification

• Recursive feature elimination with cross-validation
• Scaling the regularization parameter for SVCs
sklearn.datasets.make_friedman1
sklearn.datasets.make_friedman1(n_samples=100, n_features=10, noise=0.0, ran-

dom_state=None)
Generate the “Friedman #1” regression problem
This dataset is described in Friedman [1] and Breiman [2].
Inputs X are independent features uniformly distributed on the interval [0, 1]. The output y is created according
to the formula:

y(X) = 10 * sin(pi * X[:, 0] * X[:, 1]) + 20 * (X[:, 2] - 0.5) ** 2 + 10 * X[:,

˓→3] + 5 * X[:, 4] + noise * N(0, 1).
Out of the n_features features, only 5 are actually used to compute y. The remaining features are independent
of y.
The number of features has to be >= 5.
Parameters
n_features [int, optional (default=10)] The number of features. Should be at least 5.
noise [float, optional (default=0.0)] The standard deviation of the gaussian noise applied to the
output.
eration for dataset noise. Pass an int for reproducible output across multiple function calls.
See Glossary.
Returns
X [array of shape [n_samples, n_features]] The input samples.
y [array of shape [n_samples]] The output values.
References
[1], [2]
sklearn.datasets.make_friedman2(n_samples=100, noise=0.0, random_state=None)

Inputs X are 4 independent features uniformly distributed on the intervals:
0 <= X[:, 0] <= 100,

40 * pi <= X[:, 1] <= 560 * pi,
0 <= X[:, 2] <= 1,
1 <= X[:, 3] <= 11.
The output y is created according to the formula:
y(X) = (X[:, 0] ** 2 + (X[:, 1] * X[:, 2] - 1 / (X[:, 1] * X[:, 3])) ** 2) ** 0.

˓→5 + noise * N(0, 1).

Parameters
output.

See Glossary.
Returns
X [array of shape [n_samples, 4]] The input samples.
References
[1], [2]
sklearn.datasets.make_friedman3(n_samples=100, noise=0.0, random_state=None)

Inputs X are 4 independent features uniformly distributed on the intervals:
0 <= X[:, 0] <= 100,

40 * pi <= X[:, 1] <= 560 * pi,
0 <= X[:, 2] <= 1,
1 <= X[:, 3] <= 11.
The output y is created according to the formula:
y(X) = arctan((X[:, 1] * X[:, 2] - 1 / (X[:, 1] * X[:, 3])) / X[:, 0]) + noise *

˓→N(0, 1).

Parameters
output.
See Glossary.
Returns
References
[1], [2]

sklearn.datasets.make_gaussian_quantiles
sklearn.datasets.make_gaussian_quantiles(mean=None, cov=1.0, n_samples=100,

n_features=2, n_classes=3, shuffle=True,
random_state=None)
Generate isotropic Gaussian and label samples by quantile
This classification dataset is constructed by taking a multi-dimensional standard normal distribution and defining
classes separated by nested concentric multi-dimensional spheres such that roughly equal numbers of samples
are in each class (quantiles of the 𝜒2 distribution).
Parameters
mean [array of shape [n_features], optional (default=None)] The mean of the multi-
dimensional normal distribution. If None then use the origin (0, 0, . . . ).
cov [float, optional (default=1.)] The covariance matrix will be this value times the unit matrix.
This dataset only produces symmetric normal distributions.
n_samples [int, optional (default=100)] The total number of points equally divided among
classes.
n_features [int, optional (default=2)] The number of features for each sample.
n_classes [int, optional (default=3)] The number of classes
Returns
y [array of shape [n_samples]] The integer labels for quantile membership of each sample.
Notes
The dataset is from Zhu et al [1].
References
[1]
Examples using sklearn.datasets.make_gaussian_quantiles

• Two-class AdaBoost
• Multi-class AdaBoosted Decision Trees

sklearn.datasets.make_hastie_10_2
sklearn.datasets.make_hastie_10_2(n_samples=12000, random_state=None)
Generates data for binary classification used in Hastie et al. 2009, Example 10.2.
The ten features are standard independent Gaussian and the target y is defined by:
y[i] = 1 if np.sum(X[i] ** 2) > 9.34 else -1

Parameters
Returns
See also:
make_gaussian_quantiles a generalization of this dataset approach
References
[1]
Examples using sklearn.datasets.make_hastie_10_2

• Discrete versus Real AdaBoost
• Demonstration of multi-metric evaluation on cross_val_score and GridSearchCV
sklearn.datasets.make_low_rank_matrix
sklearn.datasets.make_low_rank_matrix(n_samples=100, n_features=100, effective_rank=10,

tail_strength=0.5, random_state=None)
Generate a mostly low rank matrix with bell-shaped singular values
Most of the variance can be explained by a bell-shaped curve of width effective_rank: the low rank part of the
singular values profile is:
(1 - tail_strength) * exp(-1.0 * (i / effective_rank) ** 2)
The remaining singular values’ tail is fat, decreasing as:
tail_strength * exp(-0.1 * i / effective_rank).

The low rank part of the profile can be considered the structured signal part of the data while the tail can be
considered the noisy part of the data that cannot be summarized by a low number of linear components (singular
vectors).
This kind of singular profiles is often seen in practice, for instance:
• gray level pictures of faces
• TF-IDF vectors of text documents crawled from the web
Parameters
n_features [int, optional (default=100)] The number of features.
effective_rank [int, optional (default=10)] The approximate number of singular vectors re-
quired to explain most of the data by linear combinations.
tail_strength [float between 0.0 and 1.0, optional (default=0.5)] The relative importance of the
fat noisy tail of the singular values profile.
Returns
X [array of shape [n_samples, n_features]] The matrix.
sklearn.datasets.make_moons
sklearn.datasets.make_moons(n_samples=100, shuffle=True, noise=None, random_state=None)

Make two interleaving half circles
A simple toy dataset to visualize clustering and classification algorithms. Read more in the User Guide.
Parameters
n_samples [int, optional (default=100)] The total number of points generated.
shuffle [bool, optional (default=True)] Whether to shuffle the samples.
noise [double or None (default=None)] Standard deviation of Gaussian noise added to the data.
eration for dataset shuffling and noise. Pass an int for reproducible output across multiple
function calls. See Glossary.
Returns
X [array of shape [n_samples, 2]] The generated samples.
y [array of shape [n_samples]] The integer labels (0 or 1) for class membership of each sample.
Examples using sklearn.datasets.make_moons



sklearn.datasets.make_multilabel_classification
sklearn.datasets.make_multilabel_classification(n_samples=100, n_features=20,
n_classes=5, n_labels=2, length=50,
allow_unlabeled=True, sparse=False,
return_indicator=’dense’, re-
turn_distributions=False, ran-
dom_state=None)
Generate a random multilabel classification problem.
For each sample, the generative process is:
• pick the number of labels: n ~ Poisson(n_labels)
• n times, choose a class c: c ~ Multinomial(theta)
• pick the document length: k ~ Poisson(length)
• k times, choose a word: w ~ Multinomial(theta_c)
In the above process, rejection sampling is used to make sure that n is never zero or more than n_classes, and
that the document length is never zero. Likewise, we reject classes which have already been chosen.
Parameters
n_features [int, optional (default=20)] The total number of features.
n_classes [int, optional (default=5)] The number of classes of the classification problem.
n_labels [int, optional (default=2)] The average number of labels per instance. More precisely,
the number of labels per sample is drawn from a Poisson distribution with n_labels as its
expected value, but samples are bounded (using rejection sampling) by n_classes, and
must be nonzero if allow_unlabeled is False.
length [int, optional (default=50)] The sum of the features (number of words if documents) is
drawn from a Poisson distribution with this expected value.
allow_unlabeled [bool, optional (default=True)] If True, some instances might not belong to
any class.
sparse [bool, optional (default=False)] If True, return a sparse feature matrix
New in version 0.17: parameter to allow sparse output.
return_indicator [‘dense’ (default) | ‘sparse’ | False] If dense return Y in the dense binary
indicator format. If 'sparse' return Y in the sparse binary indicator format. False
returns a list of lists of labels.
return_distributions [bool, optional (default=False)] If True, return the prior class probability
and conditional probabilities of features given classes, from which the data was drawn.

Returns
Y [array or sparse CSR matrix of shape [n_samples, n_classes]] The label sets.
p_c [array, shape [n_classes]] The probability of each class being drawn. Only returned if
return_distributions=True.
p_w_c [array, shape [n_features, n_classes]] The probability of each feature being drawn given
each class. Only returned if return_distributions=True.
Examples using sklearn.datasets.make_multilabel_classification
• Plot randomly generated multilabel dataset
sklearn.datasets.make_regression
sklearn.datasets.make_regression(n_samples=100, n_features=100, n_informative=10,

n_targets=1, bias=0.0, effective_rank=None,
tail_strength=0.5, noise=0.0, shuffle=True, coef=False,
random_state=None)
Generate a random regression problem.
The input set can either be well conditioned (by default) or have a low rank-fat tail singular profile. See
make_low_rank_matrix for more details.
The output is generated by applying a (potentially biased) random linear regression model with n_informative
nonzero regressors to the previously generated input and some gaussian centered noise with some adjustable
scale.
Parameters
n_informative [int, optional (default=10)] The number of informative features, i.e., the number
of features used to build the linear model used to generate the output.
n_targets [int, optional (default=1)] The number of regression targets, i.e., the dimension of the
y output vector associated with a sample. By default, the output is a scalar.
bias [float, optional (default=0.0)] The bias term in the underlying linear model.
effective_rank [int or None, optional (default=None)]
if not None: The approximate number of singular vectors required to explain most of the
input data by linear combinations. Using this kind of singular spectrum in the input allows
the generator to reproduce the correlations often observed in practice.
if None: The input set is well conditioned, centered and gaussian with unit variance.

tail_strength [float between 0.0 and 1.0, optional (default=0.5)] The relative importance of the
fat noisy tail of the singular values profile if effective_rank is not None.
output.
shuffle [boolean, optional (default=True)] Shuffle the samples and the features.
coef [boolean, optional (default=False)] If True, the coefficients of the underlying linear model
are returned.
Returns
y [array of shape [n_samples] or [n_samples, n_targets]] The output values.
coef [array of shape [n_features] or [n_features, n_targets], optional] The coefficient of the
underlying linear model. It is returned only if coef is True.
Examples using sklearn.datasets.make_regression
• Plot Ridge coefficients as a function of the L2 regularization
• Lasso on dense and sparse data
sklearn.datasets.make_s_curve
sklearn.datasets.make_s_curve(n_samples=100, noise=0.0, random_state=None)

Generate an S curve dataset.
Parameters
n_samples [int, optional (default=100)] The number of sample points on the S curve.
Returns
X [array of shape [n_samples, 3]] The points.
t [array of shape [n_samples]] The univariate position of the sample according to the main
dimension of the points in the manifold.

Examples using sklearn.datasets.make_s_curve

• Comparison of Manifold Learning methods
sklearn.datasets.make_sparse_coded_signal
sklearn.datasets.make_sparse_coded_signal(n_samples, n_components, n_features,

n_nonzero_coefs, random_state=None)
Generate a signal as a sparse combination of dictionary elements.
Returns a matrix Y = DX, such as D is (n_features, n_components), X is (n_components, n_samples) and each
column of X has exactly n_nonzero_coefs non-zero elements.
Parameters
n_samples [int] number of samples to generate
n_components [int,] number of components in the dictionary
n_features [int] number of features of the dataset to generate
n_nonzero_coefs [int] number of active (non-zero) coefficients in each sample
Returns
data [array of shape [n_features, n_samples]] The encoded signal (Y).
dictionary [array of shape [n_features, n_components]] The dictionary with normalized com-
ponents (D).
code [array of shape [n_components, n_samples]] The sparse code such that each column of
this matrix has exactly n_nonzero_coefs non-zero items (X).
Examples using sklearn.datasets.make_sparse_coded_signal
sklearn.datasets.make_sparse_spd_matrix
sklearn.datasets.make_sparse_spd_matrix(dim=1, alpha=0.95, norm_diag=False,

smallest_coef=0.1, largest_coef=0.9, ran-
dom_state=None)
Generate a sparse symmetric definite positive matrix.
Parameters
dim [integer, optional (default=1)] The size of the random matrix to generate.
alpha [float between 0 and 1, optional (default=0.95)] The probability that a coefficient is zero
(see notes). Larger values enforce more sparsity.

norm_diag [boolean, optional (default=False)] Whether to normalize the output matrix to make
the leading diagonal elements all 1
smallest_coef [float between 0 and 1, optional (default=0.1)] The value of the smallest coeffi-
cient.
largest_coef [float between 0 and 1, optional (default=0.9)] The value of the largest coefficient.
Returns
prec [sparse matrix of shape (dim, dim)] The generated matrix.
See also:
make_spd_matrix
Notes
The sparsity is actually imposed on the cholesky factor of the matrix. Thus alpha does not translate directly into
the filling fraction of the matrix itself.
Examples using sklearn.datasets.make_sparse_spd_matrix
sklearn.datasets.make_sparse_uncorrelated
sklearn.datasets.make_sparse_uncorrelated(n_samples=100, n_features=10, ran-

dom_state=None)
Generate a random regression problem with sparse uncorrelated design
This dataset is described in Celeux et al [1]. as:
X ~ N(0, 1)
y(X) = X[:, 0] + 2 * X[:, 1] - 2 * X[:, 2] - 1.5 * X[:, 3]
Only the first 4 features are informative. The remaining features are useless.
Parameters
Returns

References
[1]
sklearn.datasets.make_spd_matrix
sklearn.datasets.make_spd_matrix(n_dim, random_state=None)
Generate a random symmetric, positive-definite matrix.
Parameters
n_dim [int] The matrix dimension.
Returns
X [array of shape [n_dim, n_dim]] The random symmetric, positive-definite matrix.
See also:
make_sparse_spd_matrix
sklearn.datasets.make_swiss_roll
sklearn.datasets.make_swiss_roll(n_samples=100, noise=0.0, random_state=None)

Generate a swiss roll dataset.
Parameters
n_samples [int, optional (default=100)] The number of sample points on the S curve.
Returns
X [array of shape [n_samples, 3]] The points.
t [array of shape [n_samples]] The univariate position of the sample according to the main
dimension of the points in the manifold.
Notes
The algorithm is from Marsland [1].
References
[1]

Examples using sklearn.datasets.make_swiss_roll

• Swiss Roll reduction with LLE
6.9 sklearn.decomposition: Matrix Decomposition
The sklearn.decomposition module includes matrix decomposition algorithms, including among others PCA,
NMF or ICA. Most of the algorithms of this module can be regarded as dimensionality reduction techniques.
User guide: See the Decomposing signals in components (matrix factorization problems) section for further details.
decomposition.DictionaryLearning([. . . ]) Dictionary learning

decomposition.FactorAnalysis([n_components, Factor Analysis (FA)
. . . ])
decomposition.FastICA([n_components, . . . ]) FastICA: a fast algorithm for Independent Component
Analysis.
decomposition.IncrementalPCA([n_components, Incremental principal components analysis (IPCA).
. . . ])
decomposition.KernelPCA([n_components, . . . ]) Kernel Principal component analysis (KPCA)
decomposition.LatentDirichletAllocation([. .Latent
. ]) Dirichlet Allocation with online variational Bayes
algorithm
decomposition.MiniBatchDictionaryLearning([. Mini-batch
. . ]) dictionary learning
decomposition.MiniBatchSparsePCA([. . . ]) Mini-batch Sparse Principal Components Analysis
decomposition.NMF([n_components, init, . . . ]) Non-Negative Matrix Factorization (NMF)
decomposition.PCA([n_components, copy, . . . ]) Principal component analysis (PCA)
decomposition.SparsePCA([n_components, . . . ]) Sparse Principal Components Analysis (SparsePCA)
decomposition.SparseCoder(dictionary[, . . . ]) Sparse coding
decomposition.TruncatedSVD([n_components, Dimensionality reduction using truncated SVD (aka LSA).
. . . ])
6.9.1 sklearn.decomposition.DictionaryLearning
class sklearn.decomposition.DictionaryLearning(n_components=None, alpha=1,

max_iter=1000, tol=1e-08,
fit_algorithm=’lars’, trans-
form_algorithm=’omp’, trans-
form_n_nonzero_coefs=None, trans-
form_alpha=None, n_jobs=None,
code_init=None, dict_init=None, ver-
bose=False, split_sign=False, ran-
dom_state=None, positive_code=False,
positive_dict=False)
Dictionary learning
Finds a dictionary (a set of atoms) that can best be used to represent data using a sparse code.
Solves the optimization problem:
(U^*,V^*) = argmin 0.5 || Y - U V ||_2^2 + alpha * || U ||_1
(U,V)
with || V_k ||_2 = 1 for all 0 <= k < n_components


Parameters
n_components [int,] number of dictionary elements to extract
alpha [float,] sparsity controlling parameter
max_iter [int,] maximum number of iterations to perform
tol [float,] tolerance for numerical error
fit_algorithm [{‘lars’, ‘cd’}] lars: uses the least angle regression method to solve the lasso
problem (linear_model.lars_path) cd: uses the coordinate descent method to compute the
Lasso solution (linear_model.Lasso). Lars will be faster if the estimated components are
sparse.
New in version 0.17: cd coordinate descent method to improve speed.
transform_algorithm [{‘lasso_lars’, ‘lasso_cd’, ‘lars’, ‘omp’, ‘threshold’}] Algorithm used
to transform the data lars: uses the least angle regression method (linear_model.lars_path)
lasso_lars: uses Lars to compute the Lasso solution lasso_cd: uses the coordinate descent
method to compute the Lasso solution (linear_model.Lasso). lasso_lars will be faster if
the estimated components are sparse. omp: uses orthogonal matching pursuit to estimate
the sparse solution threshold: squashes to zero all coefficients less than alpha from the
projection dictionary * X'
New in version 0.17: lasso_cd coordinate descent method to improve speed.
transform_n_nonzero_coefs [int, 0.1 * n_features by default] Number of nonzero co-
efficients to target in each column of the solution. This is only used by algorithm=’lars’ and
algorithm=’omp’ and is overridden by alpha in the omp case.
transform_alpha [float, 1. by default] If algorithm=’lasso_lars’ or algorithm=’lasso_cd’, al-
pha is the penalty applied to the L1 norm. If algorithm=’threshold’, alpha is the abso-
lute value of the threshold below which coefficients will be squashed to zero. If algo-
rithm=’omp’, alpha is the tolerance parameter: the value of the reconstruction error tar-
geted. In this case, it overrides n_nonzero_coefs.
n_jobs [int or None, optional (default=None)] Number of parallel jobs to run. None means 1
code_init [array of shape (n_samples, n_components),] initial value for the code, for warm
restart
dict_init [array of shape (n_components, n_features),] initial values for the dictionary, for warm
restart
verbose [bool, optional (default: False)] To control the verbosity of the procedure.
split_sign [bool, False by default] Whether to split the sparse feature vector into the concate-
nation of its negative part and its positive part. This can improve the performance of down-
stream classifiers.
6.9. sklearn.decomposition: Matrix Decomposition 1547

positive_code [bool] Whether to enforce positivity when finding the code.

positive_dict [bool] Whether to enforce positivity when finding the dictionary
Attributes
components_ [array, [n_components, n_features]] dictionary atoms extracted from the data
error_ [array] vector of errors at each iteration
See also:
SparseCoder, MiniBatchDictionaryLearning, SparsePCA, MiniBatchSparsePCA
Notes
References:
J. Mairal, F. Bach, J. Ponce, G. Sapiro, 2009: Online dictionary learning for sparse coding (http://www.di.ens.
fr/sierra/pdfs/icml09.pdf)
Methods
fit(X[, y]) Fit the model from data in X.

transform(X) Encode the data as a sparse combination of the dictio-
nary atoms.
__init__(n_components=None, alpha=1, max_iter=1000, tol=1e-08, fit_algorithm=’lars’, trans-

form_algorithm=’omp’, transform_n_nonzero_coefs=None, transform_alpha=None,
n_jobs=None, code_init=None, dict_init=None, verbose=False, split_sign=False, ran-
dom_state=None, positive_code=False, positive_dict=False)
fit(X, y=None)
Fit the model from data in X.
Parameters
X [array-like, shape (n_samples, n_features)] Training vector, where n_samples in the num-
y [Ignored]
Returns
self [object] Returns the object itself

Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
transform(X)
Encode the data as a sparse combination of the dictionary atoms.
Coding method is determined by the object parameter transform_algorithm.
Parameters
X [array of shape (n_samples, n_features)] Test data to be transformed, must have the same
number of features as the data used to train the model.
Returns
X_new [array, shape (n_samples, n_components)] Transformed data
6.9.2 sklearn.decomposition.FactorAnalysis
class sklearn.decomposition.FactorAnalysis(n_components=None, tol=0.01, copy=True,

max_iter=1000, noise_variance_init=None,
svd_method=’randomized’, iterated_power=3,
random_state=0)
Factor Analysis (FA)
A simple linear generative model with Gaussian latent variables.
The observations are assumed to be caused by a linear transformation of lower dimensional latent factors and
added Gaussian noise. Without loss of generality the factors are distributed according to a Gaussian with zero
mean and unit covariance. The noise is also zero mean and has an arbitrary diagonal covariance matrix.
If we would restrict the model further, by assuming that the Gaussian noise is even isotropic (all diagonal entries
are the same) we would obtain PPCA.

FactorAnalysis performs a maximum likelihood estimate of the so-called loading matrix, the transformation of
the latent variables to the observed ones, using expectation-maximization (EM).
Parameters
n_components [int | None] Dimensionality of latent space, the number of components of X that
are obtained after transform. If None, n_components is set to the number of features.
tol [float] Stopping tolerance for EM algorithm.
copy [bool] Whether to make a copy of X. If False, the input X gets overwritten during fitting.
max_iter [int] Maximum number of iterations.
noise_variance_init [None | array, shape=(n_features,)] The initial guess of the noise variance
for each feature. If None, it defaults to np.ones(n_features)
svd_method [{‘lapack’, ‘randomized’}] Which SVD method to use. If ‘lapack’ use standard
SVD from scipy.linalg, if ‘randomized’ use fast randomized_svd function. Defaults to
‘randomized’. For most applications ‘randomized’ will be sufficiently precise while pro-
viding significant speed gains. Accuracy can also be improved by setting higher values for
iterated_power. If this is not sufficient, for maximum precision you should choose ‘lapack’.
iterated_power [int, optional] Number of iterations for the power method. 3 by default. Only
used if svd_method equals ‘randomized’
random_state [int, RandomState instance or None, optional (default=0)] If int, random_state
instance used by np.random. Only used when svd_method equals ‘randomized’.
Attributes
components_ [array, [n_components, n_features]] Components with maximum variance.
loglike_ [list, [n_iterations]] The log likelihood at each iteration.
noise_variance_ [array, shape=(n_features,)] The estimated noise variance for each feature.
See also:
PCA Principal component analysis is also a latent linear variable model which however assumes equal noise
variance for each feature. This extra assumption makes probabilistic PCA faster as it can be computed in
closed form.
FastICA Independent component analysis, a latent variable model with non-Gaussian latent variables.
References
Examples

>>> from sklearn.decomposition import FactorAnalysis
>>> X, _ = load_digits(return_X_y=True)
>>> transformer = FactorAnalysis(n_components=7, random_state=0)
>>> X_transformed = transformer.fit_transform(X)

>>> X_transformed.shape
(1797, 7)
Methods
fit(X[, y]) Fit the FactorAnalysis model to X using EM

get_covariance() Compute data covariance with the FactorAnalysis
model.
get_precision() Compute data precision matrix with the FactorAnalysis
model.
score(X[, y]) Compute the average log-likelihood of the samples
score_samples(X) Compute the log-likelihood of each sample
transform(X) Apply dimensionality reduction to X using the model.
__init__(n_components=None, tol=0.01, copy=True, max_iter=1000, noise_variance_init=None,

svd_method=’randomized’, iterated_power=3, random_state=0)
fit(X, y=None)
Fit the FactorAnalysis model to X using EM
Parameters
y [Ignored]
Returns
self
Parameters
Returns
get_covariance()
Compute data covariance with the FactorAnalysis model.
cov = components_.T * components_ + diag(noise_variance)
Returns
cov [array, shape (n_features, n_features)] Estimated covariance of data.

Parameters
Returns
get_precision()
Compute data precision matrix with the FactorAnalysis model.
Returns
precision [array, shape (n_features, n_features)] Estimated precision of data.
score(X, y=None)
Compute the average log-likelihood of the samples
Parameters
X [array, shape (n_samples, n_features)] The data
y [Ignored]
Returns
ll [float] Average log-likelihood of the samples under the current model
score_samples(X)
Compute the log-likelihood of each sample
Parameters
X [array, shape (n_samples, n_features)] The data
Returns
ll [array, shape (n_samples,)] Log-likelihood of each sample under the current model
of a nested object.
Returns
self
transform(X)
Apply dimensionality reduction to X using the model.
Compute the expected mean of the latent variables. See Barber, 21.2.33 (or Bishop, 12.66).
Parameters
Returns
X_new [array-like, shape (n_samples, n_components)] The latent variables of X.

Examples using sklearn.decomposition.FactorAnalysis

6.9.3 sklearn.decomposition.FastICA
class sklearn.decomposition.FastICA(n_components=None, algorithm=’parallel’, whiten=True,

fun=’logcosh’, fun_args=None, max_iter=200, tol=0.0001,
w_init=None, random_state=None)
FastICA: a fast algorithm for Independent Component Analysis.
Parameters
n_components [int, optional] Number of components to use. If none is passed, all are used.
algorithm [{‘parallel’, ‘deflation’}] Apply parallel or deflational algorithm for FastICA.
whiten [boolean, optional] If whiten is false, the data is already considered to be whitened, and
no whitening is performed.
fun [string or function, optional. Default: ‘logcosh’] The functional form of the G function used
in the approximation to neg-entropy. Could be either ‘logcosh’, ‘exp’, or ‘cube’. You can
also provide your own function. It should return a tuple containing the value of the function,
and of its derivative, in the point. Example:
def my_g(x): return x ** 3, (3 * x ** 2).mean(axis=-1)
fun_args [dictionary, optional] Arguments to send to the functional form. If empty and if
fun=’logcosh’, fun_args will take value {‘alpha’ : 1.0}.
max_iter [int, optional] Maximum number of iterations during fit.
tol [float, optional] Tolerance on update at each iteration.
w_init [None of an (n_components, n_components) ndarray] The mixing matrix to be used to
initialize the algorithm.
Attributes
components_ [2D array, shape (n_components, n_features)] The unmixing matrix.
mixing_ [array, shape (n_features, n_components)] The mixing matrix.
n_iter_ [int] If the algorithm is “deflation”, n_iter is the maximum number of iterations run
across all components. Else they are just the number of iterations taken to converge.
Notes
Implementation based on A. Hyvarinen and E. Oja, Independent Component Analysis: Algorithms and Appli-
cations, Neural Networks, 13(4-5), 2000, pp. 411-430

Examples

>>> from sklearn.decomposition import FastICA
>>> transformer = FastICA(n_components=7,
... random_state=0)
(1797, 7)
Methods
fit(X[, y]) Fit the model to X.

fit_transform(X[, y]) Fit the model and recover the sources from X.
inverse_transform(X[, copy]) Transform the sources back to the mixed data (apply
mixing matrix).
transform(X[, y, copy]) Recover the sources from X (apply the unmixing ma-
trix).
__init__(n_components=None, algorithm=’parallel’, whiten=True, fun=’logcosh’, fun_args=None,

max_iter=200, tol=0.0001, w_init=None, random_state=None)
fit(X, y=None)
Fit the model to X.
Parameters
X [array-like, shape (n_samples, n_features)] Training data, where n_samples is the number
of samples and n_features is the number of features.
y [Ignored]
Returns
self
Fit the model and recover the sources from X.
Parameters
y [Ignored]
Returns
X_new [array-like, shape (n_samples, n_components)]
Parameters

Returns
inverse_transform(X, copy=True)
Transform the sources back to the mixed data (apply mixing matrix).
Parameters
X [array-like, shape (n_samples, n_components)] Sources, where n_samples is the number
of samples and n_components is the number of components.
copy [bool (optional)] If False, data passed to fit are overwritten. Defaults to True.
Returns
X_new [array-like, shape (n_samples, n_features)]
of a nested object.
Returns
self
transform(X, y=’deprecated’, copy=True)
Recover the sources from X (apply the unmixing matrix).
Parameters
X [array-like, shape (n_samples, n_features)] Data to transform, where n_samples is the
number of samples and n_features is the number of features.
y [(ignored)] Deprecated since version 0.19: This parameter will be removed in 0.21.
copy [bool (optional)] If False, data passed to fit are overwritten. Defaults to True.
Returns
Examples using sklearn.decomposition.FastICA

6.9.4 sklearn.decomposition.IncrementalPCA
class sklearn.decomposition.IncrementalPCA(n_components=None, whiten=False, copy=True,

batch_size=None)
Incremental principal components analysis (IPCA).
Linear dimensionality reduction using Singular Value Decomposition of centered data, keeping only the most
significant singular vectors to project the data to a lower dimensional space.

Depending on the size of the input data, this algorithm can be much more memory efficient than a PCA.
This algorithm has constant memory complexity, on the order of batch_size, enabling use of np.memmap
files without loading the entire file into memory.
The computational overhead of each SVD is O(batch_size * n_features ** 2), but only 2 *
batch_size samples remain in memory at a time. There will be n_samples / batch_size SVD compu-
tations to get the principal components, versus 1 large SVD of complexity O(n_samples * n_features
** 2) for PCA.
Parameters
n_components [int or None, (default=None)] Number of components to keep. If
n_components `` is ``None, then n_components is set to min(n_samples,
n_features).
whiten [bool, optional] When True (False by default) the components_ vectors are di-
vided by n_samples times components_ to ensure uncorrelated outputs with unit
component-wise variances.
Whitening will remove some information from the transformed signal (the relative variance
scales of the components) but can sometimes improve the predictive accuracy of the down-
stream estimators by making data respect some hard-wired assumptions.
copy [bool, (default=True)] If False, X will be overwritten. copy=False can be used to save
memory but is unsafe for general use.
batch_size [int or None, (default=None)] The number of samples to use for each batch. Only
used when calling fit. If batch_size is None, then batch_size is inferred from the
data and set to 5 * n_features, to provide a balance between approximation accuracy
and memory consumption.
Attributes
components_ [array, shape (n_components, n_features)] Components with maximum variance.
explained_variance_ [array, shape (n_components,)] Variance explained by each of the se-
lected components.
explained_variance_ratio_ [array, shape (n_components,)] Percentage of variance explained
by each of the selected components. If all components are stored, the sum of explained
variances is equal to 1.0.
singular_values_ [array, shape (n_components,)] The singular values corresponding to each
of the selected components. The singular values are equal to the 2-norms of the
n_components variables in the lower-dimensional space.
mean_ [array, shape (n_features,)] Per-feature empirical mean, aggregate over calls to
partial_fit.
var_ [array, shape (n_features,)] Per-feature empirical variance, aggregate over calls to
partial_fit.
noise_variance_ [float] The estimated noise covariance following the Probabilistic PCA model
from Tipping and Bishop 1999. See “Pattern Recognition and Machine Learning” by C.
Bishop, 12.2.1 p. 574 or http://www.miketipping.com/papers/met-mppca.pdf.
n_components_ [int] The estimated number of components. Relevant when
n_components=None.

n_samples_seen_ [int] The number of samples processed by the estimator. Will be reset on
new calls to fit, but increments across partial_fit calls.
See also:
PCA, KernelPCA, SparsePCA, TruncatedSVD
Notes
Implements the incremental PCA model from: D. Ross, J. Lim, R. Lin, M. Yang, Incremental Learning for
Robust Visual Tracking, International Journal of Computer Vision, Volume 77, Issue 1-3, pp. 125-141, May
2008. See http://www.cs.toronto.edu/~dross/ivt/RossLimLinYang_ijcv.pdf
This model is an extension of the Sequential Karhunen-Loeve Transform from: A. Levy and M. Lindenbaum, Se-
quential Karhunen-Loeve Basis Extraction and its Application to Images, IEEE Transactions on Image Process-
ing, Volume 9, Number 8, pp. 1371-1374, August 2000. See http://www.cs.technion.ac.il/~mic/doc/skl-ip.pdf
We have specifically abstained from an optimization used by authors of both papers, a QR decomposition used
in specific situations to reduce the algorithmic complexity of the SVD. The source for this technique is Matrix
Computations, Third Edition, G. Holub and C. Van Loan, Chapter 5, section 5.4.4, pp 252-253.. This technique
has been omitted because it is advantageous only when decomposing a matrix with n_samples (rows) >=
5/3 * n_features (columns), and hurts the readability of the implemented algorithm. This would be a good
opportunity for future optimization, if it is deemed necessary.
References
4. Ross, J. Lim, R. Lin, M. Yang. Incremental Learning for Robust Visual Tracking, International Jour-
nal of Computer Vision, Volume 77, Issue 1-3, pp. 125-141, May 2008.
7. Golub and C. Van Loan. Matrix Computations, Third Edition, Chapter 5, Section 5.4.4, pp. 252-
253.
Examples

>>> from sklearn.decomposition import IncrementalPCA
>>> transformer = IncrementalPCA(n_components=7, batch_size=200)
>>> # either partially fit on smaller batches of data
>>> transformer.partial_fit(X[:100, :])
IncrementalPCA(batch_size=200, copy=True, n_components=7, whiten=False)
>>> # or let the fit function itself divide the data into batches
(1797, 7)
Methods
fit(X[, y]) Fit the model with X, using minibatches of size

batch_size.


get_covariance() Compute data covariance with the generative model.
get_precision() Compute data precision matrix with the generative
model.
inverse_transform(X) Transform data back to its original space.
partial_fit(X[, y, check_input]) Incremental fit with X.
transform(X) Apply dimensionality reduction to X.
__init__(n_components=None, whiten=False, copy=True, batch_size=None)

fit(X, y=None)
Fit the model with X, using minibatches of size batch_size.
Parameters
y [Ignored]
Returns
self [object] Returns the instance itself.
Parameters
Returns
get_covariance()
Compute data covariance with the generative model.
cov = components_.T * S**2 * components_ + sigma2 * eye(n_features)
where S**2 contains the explained variances, and sigma2 contains the noise variances.
Returns
cov [array, shape=(n_features, n_features)] Estimated covariance of data.
Parameters
Returns

get_precision()
Compute data precision matrix with the generative model.
Equals the inverse of the covariance but computed with the matrix inversion lemma for efficiency.
Returns
precision [array, shape=(n_features, n_features)] Estimated precision of data.
inverse_transform(X)
Transform data back to its original space.
In other words, return an input X_original whose transform would be X.
Parameters
X [array-like, shape (n_samples, n_components)] New data, where n_samples is the number
Returns
X_original array-like, shape (n_samples, n_features)
Notes
If whitening is enabled, inverse_transform will compute the exact inverse operation, which includes re-
versing whitening.
partial_fit(X, y=None, check_input=True)
Incremental fit with X. All of X is processed as a single batch.
Parameters
check_input [bool] Run check_array on X.
y [Ignored]
Returns
of a nested object.
Returns
self
transform(X)
Apply dimensionality reduction to X.
X is projected on the first principal components previously extracted from a training set.
Parameters
X [array-like, shape (n_samples, n_features)] New data, where n_samples is the number of
samples and n_features is the number of features.

Returns
Examples

>>> X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]])
>>> ipca = IncrementalPCA(n_components=2, batch_size=3)
>>> ipca.fit(X)
>>> ipca.transform(X)
Examples using sklearn.decomposition.IncrementalPCA
• Incremental PCA
6.9.5 sklearn.decomposition.KernelPCA
class sklearn.decomposition.KernelPCA(n_components=None, kernel=’linear’, gamma=None,

degree=3, coef0=1, kernel_params=None, alpha=1.0,
fit_inverse_transform=False, eigen_solver=’auto’,
tol=0, max_iter=None, remove_zero_eig=False, ran-
dom_state=None, copy_X=True, n_jobs=None)
Kernel Principal component analysis (KPCA)
Non-linear dimensionality reduction through the use of kernels (see Pairwise metrics, Affinities and Kernels).
Parameters
n_components [int, default=None] Number of components. If None, all non-zero components
are kept.
kernel [“linear” | “poly” | “rbf” | “sigmoid” | “cosine” | “precomputed”] Kernel. De-
fault=”linear”.
gamma [float, default=1/n_features] Kernel coefficient for rbf, poly and sigmoid kernels. Ig-
nored by other kernels.
degree [int, default=3] Degree for poly kernels. Ignored by other kernels.
coef0 [float, default=1] Independent term in poly and sigmoid kernels. Ignored by other kernels.
kernel_params [mapping of string to any, default=None] Parameters (keyword arguments) and
values for kernel passed as callable object. Ignored by other kernels.
alpha [int, default=1.0] Hyperparameter of the ridge regression that learns the inverse transform
(when fit_inverse_transform=True).
fit_inverse_transform [bool, default=False] Learn the inverse transform for non-precomputed
kernels. (i.e. learn to find the pre-image of a point)
eigen_solver [string [‘auto’|’dense’|’arpack’], default=’auto’] Select eigensolver to use. If
n_components is much less than the number of training samples, arpack may be more effi-
cient than the dense eigensolver.

tol [float, default=0] Convergence tolerance for arpack. If 0, optimal value will be chosen by
arpack.
max_iter [int, default=None] Maximum number of iterations for arpack. If None, optimal value
will be chosen by arpack.
remove_zero_eig [boolean, default=False] If True, then all components with zero eigenvalues
are removed, so that the number of components in the output may be < n_components
(and sometimes even zero due to numerical instability). When n_components is None, this
parameter is ignored and components with zero eigenvalues are removed regardless.
the RandomState instance used by np.random. Used when eigen_solver == ‘arpack’.
copy_X [boolean, default=True] If True, input X is copied and stored by the model in the X_fit_
attribute. If no further changes will be done to X, setting copy_X=False saves memory by
storing a reference.
Attributes
lambdas_ [array, (n_components,)] Eigenvalues of the centered kernel matrix in decreasing
order. If n_components and remove_zero_eig are not set, then all values are stored.
alphas_ [array, (n_samples, n_components)] Eigenvectors of the centered kernel matrix. If
n_components and remove_zero_eig are not set, then all components are stored.
dual_coef_ [array, (n_samples, n_features)] Inverse transform matrix. Only available when
fit_inverse_transform is True.
X_transformed_fit_ [array, (n_samples, n_components)] Projection of the fitted data on the
kernel principal components. Only available when fit_inverse_transform is True.
X_fit_ [(n_samples, n_features)] The data used to fit the model. If copy_X=False, then X_fit_
is a reference. This attribute is used for the calls to transform.
References
Kernel PCA was introduced in: Bernhard Schoelkopf, Alexander J. Smola, and Klaus-Robert Mueller. 1999.
Kernel principal component analysis. In Advances in kernel methods, MIT Press, Cambridge, MA, USA
327-352.
Examples

>>> from sklearn.decomposition import KernelPCA
>>> transformer = KernelPCA(n_components=7, kernel='linear')


(1797, 7)
Methods

fit_transform(X[, y]) Fit the model from data in X and transform X.
inverse_transform(X) Transform X back to original space.
transform(X) Transform X.
__init__(n_components=None, kernel=’linear’, gamma=None, degree=3, coef0=1, ker-

nel_params=None, alpha=1.0, fit_inverse_transform=False, eigen_solver=’auto’,
tol=0, max_iter=None, remove_zero_eig=False, random_state=None, copy_X=True,
n_jobs=None)
fit(X, y=None)
Parameters
Returns
fit_transform(X, y=None, **params)
Fit the model from data in X and transform X.
Parameters
Returns
Parameters
Returns
Transform X back to original space.
Parameters
X [array-like, shape (n_samples, n_components)]

Returns
X_new [array-like, shape (n_samples, n_features)]
References
“Learning to Find Pre-Images”, G BakIr et al, 2004.

of a nested object.
Returns
self
transform(X)
Transform X.
Parameters
Returns
Examples using sklearn.decomposition.KernelPCA
• Kernel PCA
6.9.6 sklearn.decomposition.LatentDirichletAllocation
class sklearn.decomposition.LatentDirichletAllocation(n_components=10,
doc_topic_prior=None,
topic_word_prior=None,
learning_method=’batch’,
learning_decay=0.7,
learning_offset=10.0,
max_iter=10, batch_size=128,
evaluate_every=-1, to-
tal_samples=1000000.0,
perp_tol=0.1,
mean_change_tol=0.001,
max_doc_update_iter=100,
n_jobs=None, verbose=0,
random_state=None,
n_topics=None)
Latent Dirichlet Allocation with online variational Bayes algorithm
Parameters

n_components [int, optional (default=10)] Number of topics.

doc_topic_prior [float, optional (default=None)] Prior of document topic distribution theta. If
the value is None, defaults to 1 / n_components. In [Re25e5648fc37-1], this is called alpha.
topic_word_prior [float, optional (default=None)] Prior of topic word distribution beta. If the
value is None, defaults to 1 / n_components. In [Re25e5648fc37-1], this is called eta.
learning_method [‘batch’ | ‘online’, default=’batch’] Method used to update _component.
Only used in fit method. In general, if the data size is large, the online update will be
much faster than the batch update.
Valid options:
'batch': Batch variational Bayes method. Use all training data in

each EM update.
Old `components_` will be overwritten in each iteration.
'online': Online variational Bayes method. In each EM update, use
mini-batch of training data to update the ``components_``
variable incrementally. The learning rate is controlled by the
``learning_decay`` and the ``learning_offset`` parameters.
Changed in version 0.20: The default learning method is now "batch".

learning_decay [float, optional (default=0.7)] It is a parameter that control learning rate in the
online learning method. The value should be set between (0.5, 1.0] to guarantee asymptotic
convergence. When the value is 0.0 and batch_size is n_samples, the update method is
same as batch learning. In the literature, this is called kappa.
learning_offset [float, optional (default=10.)] A (positive) parameter that downweights early
iterations in online learning. It should be greater than 1.0. In the literature, this is called
tau_0.
max_iter [integer, optional (default=10)] The maximum number of iterations.
batch_size [int, optional (default=128)] Number of documents to use in each EM iteration.
Only used in online learning.
evaluate_every [int, optional (default=0)] How often to evaluate perplexity. Only used in fit
method. set it to 0 or negative number to not evalute perplexity in training at all. Evaluating
perplexity can help you check convergence in training process, but it will also increase total
training time. Evaluating perplexity in every iteration might increase training time up to
two-fold.
total_samples [int, optional (default=1e6)] Total number of documents. Only used in the par-
tial_fit method.
perp_tol [float, optional (default=1e-1)] Perplexity tolerance in batch learning. Only used when
evaluate_every is greater than 0.
mean_change_tol [float, optional (default=1e-3)] Stopping tolerance for updating document
topic distribution in E-step.
max_doc_update_iter [int (default=100)] Max number of iterations for updating document
topic distribution in the E-step.
n_jobs [int or None, optional (default=None)] The number of jobs to use in the E-step. None
means 1 unless in a joblib.parallel_backend context. -1 means using all proces-
sors. See Glossary for more details.
verbose [int, optional (default=0)] Verbosity level.


n_topics [int, optional (default=None)] This parameter has been renamed to n_components and
will be removed in version 0.21. .. deprecated:: 0.19
Attributes
components_ [array, [n_components, n_features]] Variational parameters for topic word dis-
tribution. Since the complete conditional for topic word distribution is a Dirichlet,
components_[i, j] can be viewed as pseudocount that represents the number of
times word j was assigned to topic i. It can also be viewed as distribution over the words
for each topic after normalization: model.components_ / model.components_.
sum(axis=1)[:, np.newaxis].
n_batch_iter_ [int] Number of iterations of the EM step.
n_iter_ [int] Number of passes over the dataset.
References
[1] “Online Learning for Latent Dirichlet Allocation”, Matthew D. Hoffman, David M. Blei, Francis
Bach, 2010
[2] “Stochastic Variational Inference”, Matthew D. Hoffman, David M. Blei, Chong Wang, John Paisley,
2013
[3] Matthew D. Hoffman’s onlineldavb code. Link: https://github.com/blei-lab/onlineldavb
Examples
>>> from sklearn.decomposition import LatentDirichletAllocation

>>> from sklearn.datasets import make_multilabel_classification
>>> # This produces a feature matrix of token counts, similar to what
>>> # CountVectorizer would produce on text.
>>> X, _ = make_multilabel_classification(random_state=0)
>>> lda = LatentDirichletAllocation(n_components=5,
... random_state=0)
>>> lda.fit(X)
LatentDirichletAllocation(...)
>>> # get topics for some given samples:
>>> lda.transform(X[-2:])
array([[0.00360392, 0.25499205, 0.0036211 , 0.64236448, 0.09541846],
[0.15297572, 0.00362644, 0.44412786, 0.39568399, 0.003586 ]])
Methods
fit(X[, y]) Learn model for the data X with variational Bayes
method.


partial_fit(X[, y]) Online VB with Mini-Batch update.
perplexity(X[, doc_topic_distr, sub_sampling]) Calculate approximate perplexity for data X.
score(X[, y]) Calculate approximate log-likelihood as score.
transform(X) Transform data X according to the fitted model.
__init__(n_components=10, doc_topic_prior=None, topic_word_prior=None, learn-

ing_method=’batch’, learning_decay=0.7, learning_offset=10.0, max_iter=10,
batch_size=128, evaluate_every=-1, total_samples=1000000.0, perp_tol=0.1,
mean_change_tol=0.001, max_doc_update_iter=100, n_jobs=None, verbose=0, ran-
dom_state=None, n_topics=None)
fit(X, y=None)
Learn model for the data X with variational Bayes method.
When learning_method is ‘online’, use mini-batch update. Otherwise, use batch update.
Parameters
X [array-like or sparse matrix, shape=(n_samples, n_features)] Document word matrix.
y [Ignored]
Returns
self
Parameters
Returns
Parameters
Returns
partial_fit(X, y=None)
Online VB with Mini-Batch update.
Parameters
y [Ignored]
Returns

self
perplexity(X, doc_topic_distr=’deprecated’, sub_sampling=False)
Calculate approximate perplexity for data X.
Perplexity is defined as exp(-1. * log-likelihood per word)
Changed in version 0.19: doc_topic_distr argument has been deprecated and is ignored because user no
longer has access to unnormalized distribution
Parameters
X [array-like or sparse matrix, [n_samples, n_features]] Document word matrix.
doc_topic_distr [None or array, shape=(n_samples, n_components)] Document topic dis-
tribution. This argument is deprecated and is currently being ignored.
Deprecated since version 0.19.
sub_sampling [bool] Do sub-sampling or not.
Returns
score [float] Perplexity score.
score(X, y=None)
Calculate approximate log-likelihood as score.
Parameters
y [Ignored]
Returns
score [float] Use approximate bound as score.
of a nested object.
Returns
self
transform(X)
Transform data X according to the fitted model.
Changed in version 0.18: doc_topic_distr is now normalized
Parameters
Returns
doc_topic_distr [shape=(n_samples, n_components)] Document topic distribution for X.
Examples using sklearn.decomposition.LatentDirichletAllocation

6.9.7 sklearn.decomposition.MiniBatchDictionaryLearning
class sklearn.decomposition.MiniBatchDictionaryLearning(n_components=None, al-

pha=1, n_iter=1000,
fit_algorithm=’lars’,
n_jobs=None,
batch_size=3, shuffle=True,
dict_init=None, trans-
form_algorithm=’omp’, trans-
form_n_nonzero_coefs=None,
transform_alpha=None, ver-
bose=False, split_sign=False,
random_state=None, pos-
itive_code=False, posi-
tive_dict=False)
Mini-batch dictionary learning
Finds a dictionary (a set of atoms) that can best be used to represent data using a sparse code.
Solves the optimization problem:
(U^*,V^*) = argmin 0.5 || Y - U V ||_2^2 + alpha * || U ||_1

(U,V)

Parameters
n_components [int,] number of dictionary elements to extract
alpha [float,] sparsity controlling parameter
n_iter [int,] total number of iterations to perform
fit_algorithm [{‘lars’, ‘cd’}] lars: uses the least angle regression method to solve the lasso
problem (linear_model.lars_path) cd: uses the coordinate descent method to compute the
Lasso solution (linear_model.Lasso). Lars will be faster if the estimated components are
sparse.
batch_size [int,] number of samples in each mini-batch
shuffle [bool,] whether to shuffle the samples before forming batches
dict_init [array of shape (n_components, n_features),] initial value of the dictionary for warm
restart scenarios
transform_algorithm [{‘lasso_lars’, ‘lasso_cd’, ‘lars’, ‘omp’, ‘threshold’}] Algorithm used to
transform the data. lars: uses the least angle regression method (linear_model.lars_path)
projection dictionary * X’


stream classifiers.
positive_dict [bool] Whether to enforce positivity when finding the dictionary.
Attributes
components_ [array, [n_components, n_features]] components extracted from the data
inner_stats_ [tuple of (A, B) ndarrays] Internal sufficient statistics that are kept by the algo-
rithm. Keeping them is useful in online settings, to avoid loosing the history of the evolu-
tion, but they shouldn’t have any use for the end user. A (n_components, n_components) is
the dictionary covariance matrix. B (n_features, n_components) is the data approximation
matrix
See also:
SparseCoder, DictionaryLearning, SparsePCA, MiniBatchSparsePCA
Notes
References:
J. Mairal, F. Bach, J. Ponce, G. Sapiro, 2009: Online dictionary learning for sparse coding (http://www.di.ens.
fr/sierra/pdfs/icml09.pdf)
Methods

partial_fit(X[, y, iter_offset]) Updates the model using the data in X as a mini-batch.


nary atoms.
__init__(n_components=None, alpha=1, n_iter=1000, fit_algorithm=’lars’, n_jobs=None,

batch_size=3, shuffle=True, dict_init=None, transform_algorithm=’omp’, trans-
form_n_nonzero_coefs=None, transform_alpha=None, verbose=False, split_sign=False,
random_state=None, positive_code=False, positive_dict=False)
fit(X, y=None)
Parameters
y [Ignored]
Returns
Parameters
Returns
Parameters
Returns
partial_fit(X, y=None, iter_offset=None)
Updates the model using the data in X as a mini-batch.
Parameters
y [Ignored]
iter_offset [integer, optional] The number of iteration on data batches that has been per-
formed before this call to partial_fit. This is optional: if no number is passed, the memory
of the object is used.
Returns


of a nested object.
Returns
self
transform(X)
Parameters
Returns
Examples using sklearn.decomposition.MiniBatchDictionaryLearning

6.9.8 sklearn.decomposition.MiniBatchSparsePCA
class sklearn.decomposition.MiniBatchSparsePCA(n_components=None, alpha=1,

ridge_alpha=0.01, n_iter=100, call-
back=None, batch_size=3, ver-
bose=False, shuffle=True, n_jobs=None,
method=’lars’, random_state=None,
normalize_components=False)
Mini-batch Sparse Principal Components Analysis
Finds the set of sparse components that can optimally reconstruct the data. The amount of sparseness is control-
lable by the coefficient of the L1 penalty, given by the parameter alpha.
Parameters
n_components [int,] number of sparse atoms to extract
alpha [int,] Sparsity controlling parameter. Higher values lead to sparser components.
ridge_alpha [float,] Amount of ridge shrinkage to apply in order to improve conditioning when
calling the transform method.
n_iter [int,] number of iterations to perform for each mini batch
callback [callable or None, optional (default: None)] callable that gets invoked every five iter-
ations
batch_size [int,] the number of features to take in each mini batch

verbose [int] Controls the verbosity; the higher, the more messages. Defaults to 0.
shuffle [boolean,] whether to shuffle the data before splitting it in batches
method [{‘lars’, ‘cd’}] lars: uses the least angle regression method to solve the lasso prob-
lem (linear_model.lars_path) cd: uses the coordinate descent method to compute the Lasso
solution (linear_model.Lasso). Lars will be faster if the estimated components are sparse.
normalize_components [boolean, optional (default=False)]
• if False, use a version of Sparse PCA without components normalization and without data
centering. This is likely a bug and even though it’s the default for backward compatibility,
this should not be used.
• if True, use a version of Sparse PCA with components normalization and data centering.
Deprecated since version 0.22: normalize_components was added and set to False
for backward compatibility. It would be set to True from 0.22 onwards.
Attributes
components_ [array, [n_components, n_features]] Sparse components extracted from the data.
mean_ [array, shape (n_features,)] Per-feature empirical mean, estimated from the training set.
Equal to X.mean(axis=0).
See also:
PCA, SparsePCA, DictionaryLearning
Examples

>>> from sklearn.decomposition import MiniBatchSparsePCA
>>> X, _ = make_friedman1(n_samples=200, n_features=30, random_state=0)
>>> transformer = MiniBatchSparsePCA(n_components=5,
... batch_size=50,
... normalize_components=True,
... random_state=0)
>>> transformer.fit(X)
MiniBatchSparsePCA(...)
>>> X_transformed = transformer.transform(X)
(200, 5)
>>> # most values in the components_ are zero (sparsity)
>>> np.mean(transformer.components_ == 0)
0.94

Methods

transform(X[, ridge_alpha]) Least Squares projection of the data onto the sparse
components.
__init__(n_components=None, alpha=1, ridge_alpha=0.01, n_iter=100, callback=None,

batch_size=3, verbose=False, shuffle=True, n_jobs=None, method=’lars’, ran-
dom_state=None, normalize_components=False)
fit(X, y=None)
Parameters
y [Ignored]
Returns
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self

transform(X, ridge_alpha=’deprecated’)
Least Squares projection of the data onto the sparse components.
To avoid instability issues in case the system is under-determined, regularization can be applied (Ridge
regression) via the ridge_alpha parameter.
Note that Sparse PCA components orthogonality is not enforced as in PCA hence one cannot use a simple
linear projection.
Parameters
ridge_alpha [float, default: 0.01] Amount of ridge shrinkage to apply in order to improve
conditioning.
Deprecated since version 0.19: This parameter will be removed in 0.21. Specify
ridge_alpha in the SparsePCA constructor.
Returns
X_new array, shape (n_samples, n_components) Transformed data.
Examples using sklearn.decomposition.MiniBatchSparsePCA
6.9.9 sklearn.decomposition.NMF
class sklearn.decomposition.NMF(n_components=None, init=None, solver=’cd’,

beta_loss=’frobenius’, tol=0.0001, max_iter=200, ran-
dom_state=None, alpha=0.0, l1_ratio=0.0, verbose=0, shuf-
fle=False)
Non-Negative Matrix Factorization (NMF)
Find two non-negative matrices (W, H) whose product approximates the non- negative matrix X. This factoriza-
tion can be used for example for dimensionality reduction, source separation or topic extraction.
The objective function is:
0.5 * ||X - WH||_Fro^2

+ alpha * l1_ratio * ||vec(W)||_1
+ alpha * l1_ratio * ||vec(H)||_1
+ 0.5 * alpha * (1 - l1_ratio) * ||H||_Fro^2
Where:
||A||_Fro^2 = \sum_{i,j} A_{ij}^2 (Frobenius norm)

||vec(A)||_1 = \sum_{i,j} abs(A_{ij}) (Elementwise L1 norm)
For multiplicative-update (‘mu’) solver, the Frobenius norm (0.5 * ||X - WH||_Fro^2) can be changed into
another beta-divergence loss, by changing the beta_loss parameter.
The objective function is minimized with an alternating minimization of W and H.
Parameters

n_components [int or None] Number of components, if n_components is not set all features
are kept.
init [‘random’ | ‘nndsvd’ | ‘nndsvda’ | ‘nndsvdar’ | ‘custom’] Method used to initialize the pro-
cedure. Default: ‘nndsvd’ if n_components < n_features, otherwise random. Valid options:
• ‘random’: non-negative random matrices, scaled with: sqrt(X.mean() /
n_components)
• ‘nndsvd’: Nonnegative Double Singular Value Decomposition (NNDSVD)
initialization (better for sparseness)
• ‘nndsvda’: NNDSVD with zeros filled with the average of X (better when sparsity is
not desired)
• ‘nndsvdar’: NNDSVD with zeros filled with small random values (generally faster,
less accurate alternative to NNDSVDa for when sparsity is not desired)
• ‘custom’: use custom matrices W and H
solver [‘cd’ | ‘mu’] Numerical solver to use: ‘cd’ is a Coordinate Descent solver. ‘mu’ is a
Multiplicative Update solver.
New in version 0.17: Coordinate Descent solver.
New in version 0.19: Multiplicative Update solver.
beta_loss [float or string, default ‘frobenius’] String must be in {‘frobenius’, ‘kullback-leibler’,
‘itakura-saito’}. Beta divergence to be minimized, measuring the distance between X and
the dot product WH. Note that values different from ‘frobenius’ (or 2) and ‘kullback-leibler’
(or 1) lead to significantly slower fits. Note that for beta_loss <= 0 (or ‘itakura-saito’), the
input matrix X cannot contain zeros. Used only in ‘mu’ solver.
tol [float, default: 1e-4] Tolerance of the stopping condition.
max_iter [integer, default: 200] Maximum number of iterations before timing out.
random_state [int, RandomState instance or None, optional, default: None] If int, ran-
alpha [double, default: 0.] Constant that multiplies the regularization terms. Set it to zero to
have no regularization.
New in version 0.17: alpha used in the Coordinate Descent solver.
l1_ratio [double, default: 0.] The regularization mixing parameter, with 0 <= l1_ratio <= 1. For
l1_ratio = 0 the penalty is an elementwise L2 penalty (aka Frobenius Norm). For l1_ratio =
1 it is an elementwise L1 penalty. For 0 < l1_ratio < 1, the penalty is a combination of L1
and L2.
New in version 0.17: Regularization parameter l1_ratio used in the Coordinate Descent
solver.
verbose [bool, default=False] Whether to be verbose.
shuffle [boolean, default: False] If true, randomize the order of coordinates in the CD solver.
New in version 0.17: shuffle parameter used in the Coordinate Descent solver.
Attributes

components_ [array, [n_components, n_features]] Factorization matrix, sometimes called ‘dic-

tionary’.
reconstruction_err_ [number] Frobenius norm of the matrix difference, or beta-divergence,
between the training data X and the reconstructed data WH from the fitted model.
n_iter_ [int] Actual number of iterations.
References
Cichocki, Andrzej, and P. H. A. N. Anh-Huy. “Fast local algorithms for large scale nonnegative matrix and ten-
sor factorizations.” IEICE transactions on fundamentals of electronics, communications and computer sciences
92.3: 708-721, 2009.
Fevotte, C., & Idier, J. (2011). Algorithms for nonnegative matrix factorization with the beta-divergence. Neural
Computation, 23(9).
Examples

>>> X = np.array([[1, 1], [2, 1], [3, 1.2], [4, 1], [5, 0.8], [6, 1]])
>>> from sklearn.decomposition import NMF
>>> model = NMF(n_components=2, init='random', random_state=0)
>>> W = model.fit_transform(X)
>>> H = model.components_
Methods
fit(X[, y]) Learn a NMF model for the data X.

fit_transform(X[, y, W, H]) Learn a NMF model for the data X and returns the trans-
formed data.
inverse_transform(W) Transform data back to its original space.
transform(X) Transform the data X according to the fitted NMF model
__init__(n_components=None, init=None, solver=’cd’, beta_loss=’frobenius’, tol=0.0001,

max_iter=200, random_state=None, alpha=0.0, l1_ratio=0.0, verbose=0, shuffle=False)
fit(X, y=None, **params)
Learn a NMF model for the data X.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Data matrix to be decom-
posed
y [Ignored]
Returns
self
fit_transform(X, y=None, W=None, H=None)
Learn a NMF model for the data X and returns the transformed data.

This is more efficient than calling fit followed by transform.

Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Data matrix to be decom-
posed
y [Ignored]
W [array-like, shape (n_samples, n_components)] If init=’custom’, it is used as initial guess
for the solution.
H [array-like, shape (n_components, n_features)] If init=’custom’, it is used as initial guess
for the solution.
Returns
W [array, shape (n_samples, n_components)] Transformed data.
Parameters
Returns
inverse_transform(W)
Parameters
W [{array-like, sparse matrix}, shape (n_samples, n_components)] Transformed data matrix
Returns
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Data matrix of original shape
.. versionadded:: 0.18
of a nested object.
Returns
self
transform(X)
Transform the data X according to the fitted NMF model
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Data matrix to be trans-
formed by the model
Returns
W [array, shape (n_samples, n_components)] Transformed data

Examples using sklearn.decomposition.NMF
6.9.10 sklearn.decomposition.PCA
class sklearn.decomposition.PCA(n_components=None, copy=True, whiten=False,

svd_solver=’auto’, tol=0.0, iterated_power=’auto’, ran-
dom_state=None)
Principal component analysis (PCA)
Linear dimensionality reduction using Singular Value Decomposition of the data to project it to a lower dimen-
sional space.
It uses the LAPACK implementation of the full SVD or a randomized truncated SVD by the method of Halko
et al. 2009, depending on the shape of the input data and the number of components to extract.
It can also use the scipy.sparse.linalg ARPACK implementation of the truncated SVD.
Notice that this class does not support sparse input. See TruncatedSVD for an alternative with sparse data.
Parameters
n_components [int, float, None or string] Number of components to keep. if n_components is
not set all components are kept:
n_components == min(n_samples, n_features)
If n_components == 'mle' and svd_solver == 'full', Minka’s MLE is used

to guess the dimension. Use of n_components == 'mle' will interpret svd_solver
== 'auto' as svd_solver == 'full'.
If 0 < n_components < 1 and svd_solver == 'full', select the number of
components such that the amount of variance that needs to be explained is greater than the
percentage specified by n_components.
If svd_solver == 'arpack', the number of components must be strictly less than the
minimum of n_features and n_samples.
Hence, the None case results in:
n_components == min(n_samples, n_features) - 1
copy [bool (default True)] If False, data passed to fit are overwritten and running
fit(X).transform(X) will not yield the expected results, use fit_transform(X) instead.
whiten [bool, optional (default False)] When True (False by default) the components_ vectors
are multiplied by the square root of n_samples and then divided by the singular values to
ensure uncorrelated outputs with unit component-wise variances.
Whitening will remove some information from the transformed signal (the relative variance
scales of the components) but can sometime improve the predictive accuracy of the down-
stream estimators by making their data respect some hard-wired assumptions.
svd_solver [string {‘auto’, ‘full’, ‘arpack’, ‘randomized’}]

auto : the solver is selected by a default policy based on X.shape and n_components: if the
input data is larger than 500x500 and the number of components to extract is lower than
80% of the smallest dimension of the data, then the more efficient ‘randomized’ method is
enabled. Otherwise the exact full SVD is computed and optionally truncated afterwards.
full : run exact full SVD calling the standard LAPACK solver via scipy.linalg.svd and select
the components by postprocessing
arpack : run SVD truncated to n_components calling ARPACK solver via
scipy.sparse.linalg.svds. It requires strictly 0 < n_components < min(X.shape)
randomized : run randomized SVD by the method of Halko et al.
tol [float >= 0, optional (default .0)] Tolerance for singular values computed by svd_solver ==
‘arpack’.
iterated_power [int >= 0, or ‘auto’, (default ‘auto’)] Number of iterations for the power method
computed by svd_solver == ‘randomized’.
random_state [int, RandomState instance or None, optional (default None)] If int, ran-
the RandomState instance used by np.random. Used when svd_solver == ‘arpack’ or
‘randomized’.
Attributes
components_ [array, shape (n_components, n_features)] Principal axes in feature space, rep-
resenting the directions of maximum variance in the data. The components are sorted by
explained_variance_.
explained_variance_ [array, shape (n_components,)] The amount of variance explained by
each of the selected components.
Equal to n_components largest eigenvalues of the covariance matrix of X.
by each of the selected components.
If n_components is not set then all components are stored and the sum of the ratios is
equal to 1.0.
n_components_ [int] The estimated number of components. When n_components is set to
‘mle’ or a number between 0 and 1 (with svd_solver == ‘full’) this number is estimated
from input data. Otherwise it equals the parameter n_components, or the lesser value of
n_features and n_samples if n_components is None.

noise_variance_ [float] The estimated noise covariance following the Probabilistic PCA model
from Tipping and Bishop 1999. See “Pattern Recognition and Machine Learning” by C.
Bishop, 12.2.1 p. 574 or http://www.miketipping.com/papers/met-mppca.pdf. It is required
to compute the estimated data covariance and score samples.
Equal to the average of (min(n_features, n_samples) - n_components) smallest eigenvalues
of the covariance matrix of X.
See also:
KernelPCA, SparsePCA, TruncatedSVD, IncrementalPCA
References
For n_components == ‘mle’, this class uses the method of Minka, T. P. “Automatic choice of dimensionality for
PCA”. In NIPS, pp. 598-604
Implements the probabilistic PCA model from: ‘Tipping, M. E., and Bishop, C. M. (1999). “Probabilistic
principal component analysis”. Journal of the Royal Statistical Society: Series B (Statistical Methodology),
61(3), 611-622. via the score and score_samples methods. See http://www.miketipping.com/papers/met-mppca.
pdf
For svd_solver == ‘arpack’, refer to scipy.sparse.linalg.svds.
For svd_solver == ‘randomized’, see: Halko, N., Martinsson, P. G., and Tropp, J. A. (2011). “Finding structure
with randomness: Probabilistic algorithms for constructing approximate matrix decompositions”. SIAM review,
53(2), 217-288. and also Martinsson, P. G., Rokhlin, V., and Tygert, M. (2011). “A randomized algorithm for
the decomposition of matrices”. Applied and Computational Harmonic Analysis, 30(1), 47-68.
Examples

>>> X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]])
>>> pca = PCA(n_components=2)
>>> pca.fit(X)
PCA(copy=True, iterated_power='auto', n_components=2, random_state=None,
>>> print(pca.explained_variance_ratio_)
[0.9924... 0.0075...]
>>> print(pca.singular_values_)
[6.30061... 0.54980...]
>>> pca = PCA(n_components=2, svd_solver='full')

>>> pca.fit(X)
svd_solver='full', tol=0.0, whiten=False)
[0.9924... 0.00755...]
[6.30061... 0.54980...]
>>> pca = PCA(n_components=1, svd_solver='arpack')

>>> pca.fit(X)
svd_solver='arpack', tol=0.0, whiten=False)

[0.99244...]
[6.30061...]
Methods
fit(X[, y]) Fit the model with X.

fit_transform(X[, y]) Fit the model with X and apply the dimensionality re-
duction on X.
get_covariance() Compute data covariance with the generative model.
get_precision() Compute data precision matrix with the generative
model.
inverse_transform(X) Transform data back to its original space.
score(X[, y]) Return the average log-likelihood of all samples.
score_samples(X) Return the log-likelihood of each sample.
transform(X) Apply dimensionality reduction to X.
__init__(n_components=None, copy=True, whiten=False, svd_solver=’auto’, tol=0.0, iter-

ated_power=’auto’, random_state=None)
fit(X, y=None)
Fit the model with X.
Parameters
y [Ignored]
Returns
Fit the model with X and apply the dimensionality reduction on X.
Parameters
y [Ignored]
Returns
get_covariance()
Compute data covariance with the generative model.
cov = components_.T * S**2 * components_ + sigma2 * eye(n_features)
where S**2 contains the explained variances, and sigma2 contains the noise variances.
Returns

cov [array, shape=(n_features, n_features)] Estimated covariance of data.

Parameters
Returns
get_precision()
Compute data precision matrix with the generative model.
Equals the inverse of the covariance but computed with the matrix inversion lemma for efficiency.
Returns
precision [array, shape=(n_features, n_features)] Estimated precision of data.
In other words, return an input X_original whose transform would be X.
Parameters
X [array-like, shape (n_samples, n_components)] New data, where n_samples is the number
Returns
X_original array-like, shape (n_samples, n_features)
Notes
If whitening is enabled, inverse_transform will compute the exact inverse operation, which includes re-
versing whitening.
score(X, y=None)
Return the average log-likelihood of all samples.
See. “Pattern Recognition and Machine Learning” by C. Bishop, 12.2.1 p. 574 or http://www.miketipping.
com/papers/met-mppca.pdf
Parameters
X [array, shape(n_samples, n_features)] The data.
y [Ignored]
Returns
ll [float] Average log-likelihood of the samples under the current model
score_samples(X)
Return the log-likelihood of each sample.
See. “Pattern Recognition and Machine Learning” by C. Bishop, 12.2.1 p. 574 or http://www.miketipping.
com/papers/met-mppca.pdf
Parameters

X [array, shape(n_samples, n_features)] The data.

Returns
ll [array, shape (n_samples,)] Log-likelihood of each sample under the current model
of a nested object.
Returns
self
transform(X)
Apply dimensionality reduction to X.
X is projected on the first principal components previously extracted from a training set.
Parameters
X [array-like, shape (n_samples, n_features)] New data, where n_samples is the number of
samples and n_features is the number of features.
Returns
Examples

>>> X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]])
>>> ipca = IncrementalPCA(n_components=2, batch_size=3)
>>> ipca.fit(X)
>>> ipca.transform(X)
Examples using sklearn.decomposition.PCA
• The Iris Dataset
• PCA example with Iris Data-set
• Incremental PCA


• Principal components analysis (PCA)
• Kernel PCA
• Multi-dimensional scaling
• Using FunctionTransformer to select columns
6.9.11 sklearn.decomposition.SparsePCA
class sklearn.decomposition.SparsePCA(n_components=None, alpha=1, ridge_alpha=0.01,

max_iter=1000, tol=1e-08, method=’lars’,
n_jobs=None, U_init=None, V_init=None,
verbose=False, random_state=None, normal-
ize_components=False)
Sparse Principal Components Analysis (SparsePCA)
Finds the set of sparse components that can optimally reconstruct the data. The amount of sparseness is control-
lable by the coefficient of the L1 penalty, given by the parameter alpha.
Parameters
n_components [int,] Number of sparse atoms to extract.
alpha [float,] Sparsity controlling parameter. Higher values lead to sparser components.
ridge_alpha [float,] Amount of ridge shrinkage to apply in order to improve conditioning when
calling the transform method.
max_iter [int,] Maximum number of iterations to perform.
tol [float,] Tolerance for the stopping condition.
U_init [array of shape (n_samples, n_components),] Initial values for the loadings for warm
restart scenarios.
V_init [array of shape (n_components, n_features),] Initial values for the components for warm
restart scenarios.
verbose [int] Controls the verbosity; the higher, the more messages. Defaults to 0.


normalize_components [boolean, optional (default=False)]
• if False, use a version of Sparse PCA without components normalization and without data
centering. This is likely a bug and even though it’s the default for backward compatibility,
this should not be used.
• if True, use a version of Sparse PCA with components normalization and data centering.
Deprecated since version 0.22: normalize_components was added and set to False
for backward compatibility. It would be set to True from 0.22 onwards.
Attributes
components_ [array, [n_components, n_features]] Sparse components extracted from the data.
error_ [array] Vector of errors at each iteration.
See also:
PCA, MiniBatchSparsePCA, DictionaryLearning
Examples

>>> from sklearn.decomposition import SparsePCA
>>> X, _ = make_friedman1(n_samples=200, n_features=30, random_state=0)
>>> transformer = SparsePCA(n_components=5,
... normalize_components=True,
... random_state=0)
>>> transformer.fit(X)
SparsePCA(...)
>>> X_transformed = transformer.transform(X)
(200, 5)
>>> # most values in the components_ are zero (sparsity)
>>> np.mean(transformer.components_ == 0)
0.9666...
Methods



transform(X[, ridge_alpha]) Least Squares projection of the data onto the sparse
components.
__init__(n_components=None, alpha=1, ridge_alpha=0.01, max_iter=1000, tol=1e-08,

method=’lars’, n_jobs=None, U_init=None, V_init=None, verbose=False, ran-
dom_state=None, normalize_components=False)
fit(X, y=None)
Parameters
y [Ignored]
Returns
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
transform(X, ridge_alpha=’deprecated’)
Least Squares projection of the data onto the sparse components.
To avoid instability issues in case the system is under-determined, regularization can be applied (Ridge
regression) via the ridge_alpha parameter.

Note that Sparse PCA components orthogonality is not enforced as in PCA hence one cannot use a simple
linear projection.
Parameters
ridge_alpha [float, default: 0.01] Amount of ridge shrinkage to apply in order to improve
conditioning.
Deprecated since version 0.19: This parameter will be removed in 0.21. Specify
ridge_alpha in the SparsePCA constructor.
Returns
X_new array, shape (n_samples, n_components) Transformed data.
6.9.12 sklearn.decomposition.SparseCoder
class sklearn.decomposition.SparseCoder(dictionary, transform_algorithm=’omp’,

transform_n_nonzero_coefs=None, trans-
form_alpha=None, split_sign=False, n_jobs=None,
positive_code=False)
Sparse coding
Finds a sparse representation of data against a fixed, precomputed dictionary.
Each row of the result is the solution to a sparse coding problem. The goal is to find a sparse array code such
that:
X ~= code * dictionary

Parameters
dictionary [array, [n_components, n_features]] The dictionary atoms used for sparse coding.
Lines are assumed to be normalized to unit norm.
transform_algorithm [{‘lasso_lars’, ‘lasso_cd’, ‘lars’, ‘omp’, ‘threshold’}] Algorithm used to
transform the data: lars: uses the least angle regression method (linear_model.lars_path)
projection dictionary * X'
stream classifiers.

Attributes
components_ [array, [n_components, n_features]] The unchanged dictionary atoms
See also:
DictionaryLearning, MiniBatchDictionaryLearning, SparsePCA,
MiniBatchSparsePCA, sparse_encode
Methods
fit(X[, y]) Do nothing and return the estimator unchanged

nary atoms.
__init__(dictionary, transform_algorithm=’omp’, transform_n_nonzero_coefs=None, trans-

form_alpha=None, split_sign=False, n_jobs=None, positive_code=False)
fit(X, y=None)
Do nothing and return the estimator unchanged
This method is just there to implement the usual API and hence work in pipelines.
Parameters
X [Ignored]
y [Ignored]
Returns
self [object] Returns the object itself
Parameters
Returns
Parameters

Returns
of a nested object.
Returns
self
transform(X)
Parameters
Returns
Examples using sklearn.decomposition.SparseCoder
• Sparse coding with a precomputed dictionary
6.9.13 sklearn.decomposition.TruncatedSVD
class sklearn.decomposition.TruncatedSVD(n_components=2, algorithm=’randomized’,

n_iter=5, random_state=None, tol=0.0)
Dimensionality reduction using truncated SVD (aka LSA).
This transformer performs linear dimensionality reduction by means of truncated singular value decomposition
(SVD). Contrary to PCA, this estimator does not center the data before computing the singular value decompo-
sition. This means it can work with scipy.sparse matrices efficiently.
In particular, truncated SVD works on term count/tf-idf matrices as returned by the vectorizers in
sklearn.feature_extraction.text. In that context, it is known as latent semantic analysis (LSA).
This estimator supports two algorithms: a fast randomized SVD solver, and a “naive” algorithm that uses
ARPACK as an eigensolver on (X * X.T) or (X.T * X), whichever is more efficient.
Parameters
n_components [int, default = 2] Desired dimensionality of output data. Must be strictly less
than the number of features. The default value is useful for visualisation. For LSA, a value
of 100 is recommended.

algorithm [string, default = “randomized”] SVD solver to use. Either “arpack” for the
ARPACK wrapper in SciPy (scipy.sparse.linalg.svds), or “randomized” for the randomized
algorithm due to Halko (2009).
n_iter [int, optional (default 5)] Number of iterations for randomized SVD solver. Not used by
ARPACK. The default is larger than the default in randomized_svd to handle sparse matrices
that may have large slowly decaying spectrum.
random_state [int, RandomState instance or None, optional, default = None] If int, ran-
tol [float, optional] Tolerance for ARPACK. 0 means machine precision. Ignored by random-
ized SVD solver.
Attributes
components_ [array, shape (n_components, n_features)]
explained_variance_ [array, shape (n_components,)] The variance of the training samples
transformed by a projection to each component.
by each of the selected components.
See also:
PCA
Notes
SVD suffers from a problem called “sign indeterminacy”, which means the sign of the components_ and the
output from transform depend on the algorithm and random state. To work around this, fit instances of this class
to data once, then keep the instance around to do transformations.
References
Finding structure with randomness: Stochastic algorithms for constructing approximate matrix decompositions
Halko, et al., 2009 (arXiv:909) https://arxiv.org/pdf/0909.4061.pdf
Examples
>>> from sklearn.decomposition import TruncatedSVD

>>> from sklearn.random_projection import sparse_random_matrix
>>> X = sparse_random_matrix(100, 100, density=0.01, random_state=42)
>>> svd = TruncatedSVD(n_components=5, n_iter=7, random_state=42)
>>> svd.fit(X)
TruncatedSVD(algorithm='randomized', n_components=5, n_iter=7,
random_state=42, tol=0.0)
>>> print(svd.explained_variance_ratio_)
[0.0606... 0.0584... 0.0497... 0.0434... 0.0372...]

>>> print(svd.explained_variance_ratio_.sum())
0.249...
>>> print(svd.singular_values_)
[2.5841... 2.5245... 2.3201... 2.1753... 2.0443...]
Methods
fit(X[, y]) Fit LSI model on training data X.

fit_transform(X[, y]) Fit LSI model to X and perform dimensionality reduc-
tion on X.
inverse_transform(X) Transform X back to its original space.
transform(X) Perform dimensionality reduction on X.
__init__(n_components=2, algorithm=’randomized’, n_iter=5, random_state=None, tol=0.0)

fit(X, y=None)
Fit LSI model on training data X.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Training data.
y [Ignored]
Returns
self [object] Returns the transformer object.
Fit LSI model to X and perform dimensionality reduction on X.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Training data.
y [Ignored]
Returns
X_new [array, shape (n_samples, n_components)] Reduced version of X. This will always
be a dense array.
Parameters
Returns
Transform X back to its original space.
Returns an array X_original whose transform would be X.

Parameters
X [array-like, shape (n_samples, n_components)] New data.
Returns
X_original [array, shape (n_samples, n_features)] Note that this is always a dense array.
of a nested object.
Returns
self
transform(X)
Perform dimensionality reduction on X.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] New data.
Returns
X_new [array, shape (n_samples, n_components)] Reduced version of X. This will always
be a dense array.
Examples using sklearn.decomposition.TruncatedSVD

decomposition.dict_learning(X, Solves a dictionary learning matrix factorization problem.

n_components, . . . )
decomposition.dict_learning_online(X[, Solves a dictionary learning matrix factorization problem
. . . ]) online.
decomposition.fastica(X[, n_components, . . . ]) Perform Fast Independent Component Analysis.
decomposition.sparse_encode(X, dictionary[, Sparse coding
. . . ])
6.9.14 sklearn.decomposition.dict_learning
sklearn.decomposition.dict_learning(X, n_components, alpha, max_iter=100, tol=1e-

08, method=’lars’, n_jobs=None, dict_init=None,
code_init=None, callback=None, verbose=False,
random_state=None, return_n_iter=False, posi-
tive_dict=False, positive_code=False)
Solves a dictionary learning matrix factorization problem.
Finds the best dictionary and the corresponding sparse code for approximating the data matrix X by solving:

(U^*, V^*) = argmin 0.5 || X - U V ||_2^2 + alpha * || U ||_1

(U,V)
where V is the dictionary and U is the sparse code.

Parameters
X [array of shape (n_samples, n_features)] Data matrix.
n_components [int,] Number of dictionary atoms to extract.
alpha [int,] Sparsity controlling parameter.
max_iter [int,] Maximum number of iterations to perform.
tol [float,] Tolerance for the stopping condition.
dict_init [array of shape (n_components, n_features),] Initial value for the dictionary for warm
restart scenarios.
code_init [array of shape (n_samples, n_components),] Initial value for the sparse code for
warm restart scenarios.
callback [callable or None, optional (default: None)] Callable that gets invoked every five iter-
ations
return_n_iter [bool] Whether or not to return the number of iterations.
Returns
code [array of shape (n_samples, n_components)] The sparse code factor in the matrix factor-
ization.
dictionary [array of shape (n_components, n_features),] The dictionary factor in the matrix
factorization.
errors [array] Vector of errors at each iteration.
n_iter [int] Number of iterations run. Returned only if return_n_iter is set to True.

See also:
dict_learning_online, DictionaryLearning, MiniBatchDictionaryLearning,
SparsePCA, MiniBatchSparsePCA
6.9.15 sklearn.decomposition.dict_learning_online
sklearn.decomposition.dict_learning_online(X, n_components=2, alpha=1, n_iter=100,

return_code=True, dict_init=None, call-
back=None, batch_size=3, verbose=False,
shuffle=True, n_jobs=None, method=’lars’,
iter_offset=0, random_state=None, re-
turn_inner_stats=False, inner_stats=None,
return_n_iter=False, positive_dict=False,
positive_code=False)
Solves a dictionary learning matrix factorization problem online.
Finds the best dictionary and the corresponding sparse code for approximating the data matrix X by solving:
(U^*, V^*) = argmin 0.5 || X - U V ||_2^2 + alpha * || U ||_1

(U,V)
where V is the dictionary and U is the sparse code. This is accomplished by repeatedly iterating over mini-
batches by slicing the input data.
Parameters
X [array of shape (n_samples, n_features)] Data matrix.
n_components [int,] Number of dictionary atoms to extract.
alpha [float,] Sparsity controlling parameter.
n_iter [int,] Number of iterations to perform.
return_code [boolean,] Whether to also return the code U or just the dictionary V.
dict_init [array of shape (n_components, n_features),] Initial value for the dictionary for warm
restart scenarios.
callback [callable or None, optional (default: None)] callable that gets invoked every five iter-
ations
batch_size [int,] The number of samples to take in each batch.
shuffle [boolean,] Whether to shuffle the data before splitting it in batches.
iter_offset [int, default 0] Number of previous iterations completed on the dictionary used for
initialization.


return_inner_stats [boolean, optional] Return the inner statistics A (dictionary covariance)
and B (data approximation). Useful to restart the algorithm in an online setting. If re-
turn_inner_stats is True, return_code is ignored
inner_stats [tuple of (A, B) ndarrays] Inner sufficient statistics that are kept by the algorithm.
Passing them at initialization is useful in online settings, to avoid loosing the history of
the evolution. A (n_components, n_components) is the dictionary covariance matrix. B
(n_features, n_components) is the data approximation matrix
return_n_iter [bool] Whether or not to return the number of iterations.
Returns
code [array of shape (n_samples, n_components),] the sparse code (only returned if re-
turn_code=True)
dictionary [array of shape (n_components, n_features),] the solutions to the dictionary learning
problem
See also:
dict_learning, DictionaryLearning, MiniBatchDictionaryLearning, SparsePCA,
MiniBatchSparsePCA
6.9.16 sklearn.decomposition.fastica
sklearn.decomposition.fastica(X, n_components=None, algorithm=’parallel’, whiten=True,

fun=’logcosh’, fun_args=None, max_iter=200, tol=0.0001,
w_init=None, random_state=None, return_X_mean=False,
compute_sources=True, return_n_iter=False)
Perform Fast Independent Component Analysis.
Parameters
X [array-like, shape (n_samples, n_features)] Training vector, where n_samples is the number
n_components [int, optional] Number of components to extract. If None no dimension reduc-
tion is performed.
algorithm [{‘parallel’, ‘deflation’}, optional] Apply a parallel or deflational FASTICA algo-
rithm.
whiten [boolean, optional] If True perform an initial whitening of the data. If False, the data
is assumed to have already been preprocessed: it should be centered, normed and white.

Otherwise you will get incorrect results. In this case the parameter n_components will be
ignored.
fun [string or function, optional. Default: ‘logcosh’] The functional form of the G function
used in the approximation to neg-entropy. Could be either ‘logcosh’, ‘exp’, or ‘cube’. You
can also provide your own function. It should return a tuple containing the value of the
function, and of its derivative, in the point. The derivative should be averaged along its last
dimension. Example:
def my_g(x): return x ** 3, np.mean(3 * x ** 2, axis=-1)
fun_args [dictionary, optional] Arguments to send to the functional form. If empty or None
and if fun=’logcosh’, fun_args will take value {‘alpha’ : 1.0}
max_iter [int, optional] Maximum number of iterations to perform.
tol [float, optional] A positive scalar giving the tolerance at which the un-mixing matrix is
considered to have converged.
w_init [(n_components, n_components) array, optional] Initial un-mixing array of dimension
(n.comp,n.comp). If None (default) then an array of normal r.v.’s is used.
return_X_mean [bool, optional] If True, X_mean is returned too.
compute_sources [bool, optional] If False, sources are not computed, but only the rotation
matrix. This can save memory when working with big data. Defaults to True.
Returns
K [array, shape (n_components, n_features) | None.] If whiten is ‘True’, K is the pre-whitening
matrix that projects data onto the first n_components principal components. If whiten is
‘False’, K is ‘None’.
W [array, shape (n_components, n_components)] Estimated un-mixing matrix. The mixing
matrix can be obtained by:
w = np.dot(W, K.T)
A = w.T * (w * w.T).I
S [array, shape (n_samples, n_components) | None] Estimated source matrix

X_mean [array, shape (n_features, )] The mean over features. Returned only if return_X_mean
is True.
n_iter [int] If the algorithm is “deflation”, n_iter is the maximum number of iterations run
across all components. Else they are just the number of iterations taken to converge. This is
returned only when return_n_iter is set to True.
Notes
The data matrix X is considered to be a linear combination of non-Gaussian (independent) components i.e. X
= AS where columns of S contain the independent components and A is a linear mixing matrix. In short ICA
attempts to un-mix’ the data by estimating an un-mixing matrix W where ‘‘S = W K X.‘

This implementation was originally made for data of shape [n_features, n_samples]. Now the input is transposed
before the algorithm is applied. This makes it slightly faster for Fortran-ordered input.
Implemented using FastICA: A. Hyvarinen and E. Oja, Independent Component Analysis: Algorithms and
Applications, Neural Networks, 13(4-5), 2000, pp. 411-430
6.9.17 sklearn.decomposition.sparse_encode
sklearn.decomposition.sparse_encode(X, dictionary, gram=None, cov=None, algo-

rithm=’lasso_lars’, n_nonzero_coefs=None, al-
pha=None, copy_cov=True, init=None, max_iter=1000,
n_jobs=None, check_input=True, verbose=0, posi-
tive=False)
Sparse coding
Each row of the result is the solution to a sparse coding problem. The goal is to find a sparse array code such
that:
X ~= code * dictionary

Parameters
X [array of shape (n_samples, n_features)] Data matrix
dictionary [array of shape (n_components, n_features)] The dictionary matrix against which to
solve the sparse coding of the data. Some of the algorithms assume normalized rows for
meaningful output.
gram [array, shape=(n_components, n_components)] Precomputed Gram matrix, dictionary *
dictionary’
cov [array, shape=(n_components, n_samples)] Precomputed covariance, dictionary’ * X
algorithm [{‘lasso_lars’, ‘lasso_cd’, ‘lars’, ‘omp’, ‘threshold’}] lars: uses the least angle re-
gression method (linear_model.lars_path) lasso_lars: uses Lars to compute the Lasso so-
lution lasso_cd: uses the coordinate descent method to compute the Lasso solution (lin-
ear_model.Lasso). lasso_lars will be faster if the estimated components are sparse. omp:
uses orthogonal matching pursuit to estimate the sparse solution threshold: squashes to zero
all coefficients less than alpha from the projection dictionary * X’
n_nonzero_coefs [int, 0.1 * n_features by default] Number of nonzero coefficients to target in
each column of the solution. This is only used by algorithm=’lars’ and algorithm=’omp’
and is overridden by alpha in the omp case.
alpha [float, 1. by default] If algorithm=’lasso_lars’ or algorithm=’lasso_cd’, alpha is the
penalty applied to the L1 norm. If algorithm=’threshold’, alpha is the absolute value of
the threshold below which coefficients will be squashed to zero. If algorithm=’omp’, alpha
is the tolerance parameter: the value of the reconstruction error targeted. In this case, it
overrides n_nonzero_coefs.
copy_cov [boolean, optional] Whether to copy the precomputed covariance matrix; if False, it
may be overwritten.
init [array of shape (n_samples, n_components)] Initialization value of the sparse codes. Only
used if algorithm=’lasso_cd’.
max_iter [int, 1000 by default] Maximum number of iterations to perform if algo-
rithm=’lasso_cd’.

check_input [boolean, optional] If False, the input arrays X and dictionary will not be checked.
verbose [int, optional] Controls the verbosity; the higher, the more messages. Defaults to 0.
positive [boolean, optional] Whether to enforce positivity when finding the encoding.
Returns
code [array of shape (n_samples, n_components)] The sparse codes
See also:
sklearn.linear_model.lars_path, sklearn.linear_model.orthogonal_mp, sklearn.
linear_model.Lasso, SparseCoder
6.10 sklearn.discriminant_analysis: Discriminant Analysis
Linear Discriminant Analysis and Quadratic Discriminant Analysis

User guide: See the Linear and Quadratic Discriminant Analysis section for further details.
discriminant_analysis. Linear Discriminant Analysis

LinearDiscriminantAnalysis([. . . ])
discriminant_analysis. Quadratic Discriminant Analysis
QuadraticDiscriminantAnalysis([. . . ])
6.10.1 sklearn.discriminant_analysis.LinearDiscriminantAnalysis
class sklearn.discriminant_analysis.LinearDiscriminantAnalysis(solver=’svd’,
shrinkage=None,
priors=None,
n_components=None,
store_covariance=False,
tol=0.0001)
Linear Discriminant Analysis
A classifier with a linear decision boundary, generated by fitting class conditional densities to the data and using
Bayes’ rule.
The model fits a Gaussian density to each class, assuming that all classes share the same covariance matrix.
The fitted model can also be used to reduce the dimensionality of the input by projecting it to the most discrim-
inative directions.
New in version 0.17: LinearDiscriminantAnalysis.
Parameters
solver [string, optional]
Solver to use, possible values:

• ‘svd’: Singular value decomposition (default). Does not compute the covariance ma-
trix, therefore this solver is recommended for data with a large number of features.
• ‘lsqr’: Least squares solution, can be combined with shrinkage.
• ‘eigen’: Eigenvalue decomposition, can be combined with shrinkage.
shrinkage [string or float, optional]
Shrinkage parameter, possible values:
• None: no shrinkage (default).
• ‘auto’: automatic shrinkage using the Ledoit-Wolf lemma.
• float between 0 and 1: fixed shrinkage parameter.
Note that shrinkage works only with ‘lsqr’ and ‘eigen’ solvers.
priors [array, optional, shape (n_classes,)] Class priors.
n_components [int, optional] Number of components (< n_classes - 1) for dimensionality re-
duction.
store_covariance [bool, optional] Additionally compute class covariance matrix (default
False), used only in ‘svd’ solver.
tol [float, optional, (default 1.0e-4)] Threshold used for rank estimation in SVD solver.
Attributes
coef_ [array, shape (n_features,) or (n_classes, n_features)] Weight vector(s).
intercept_ [array, shape (n_features,)] Intercept term.
covariance_ [array-like, shape (n_features, n_features)] Covariance matrix (shared by all
classes).
by each of the selected components. If n_components is not set then all components are
stored and the sum of explained variances is equal to 1.0. Only available when eigen or svd
solver is used.
means_ [array-like, shape (n_classes, n_features)] Class means.
priors_ [array-like, shape (n_classes,)] Class priors (sum to 1).
scalings_ [array-like, shape (rank, n_classes - 1)] Scaling of the features in the space spanned
by the class centroids.
xbar_ [array-like, shape (n_features,)] Overall mean.
classes_ [array-like, shape (n_classes,)] Unique class labels.
See also:
sklearn.discriminant_analysis.QuadraticDiscriminantAnalysis Quadratic Discrimi-

nant Analysis
6.10. sklearn.discriminant_analysis: Discriminant Analysis 1599

Notes
The default solver is ‘svd’. It can perform both classification and transform, and it does not rely on the calcu-
lation of the covariance matrix. This can be an advantage in situations where the number of features is large.
However, the ‘svd’ solver cannot be used with shrinkage.
The ‘lsqr’ solver is an efficient algorithm that only works for classification. It supports shrinkage.
The ‘eigen’ solver is based on the optimization of the between class scatter to within class scatter ratio. It can
be used for both classification and transform, and it supports shrinkage. However, the ‘eigen’ solver needs to
compute the covariance matrix, so it might not be suitable for situations with a high number of features.
Examples

>>> from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
>>> X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]])
>>> y = np.array([1, 1, 1, 2, 2, 2])
>>> clf = LinearDiscriminantAnalysis()
>>> clf.fit(X, y)
LinearDiscriminantAnalysis(n_components=None, priors=None, shrinkage=None,
solver='svd', store_covariance=False, tol=0.0001)
[1]
Methods

fit(X, y) Fit LinearDiscriminantAnalysis model according to the
given training data and parameters.
predict_log_proba(X) Estimate log probability.
predict_proba(X) Estimate probability.
labels.
transform(X) Project data to maximize class separation.
__init__(solver=’svd’, shrinkage=None, priors=None, n_components=None,

store_covariance=False, tol=0.0001)
Parameters
Returns


fit(X, y)
Fit LinearDiscriminantAnalysis model according to the given training data and parameters.
Changed in version 0.19: store_covariance has been moved to main constructor.
Changed in version 0.19: tol has been moved to main constructor.
Parameters
y [array, shape (n_samples,)] Target values.

Parameters
Returns
Parameters
Returns
predict(X)
Parameters
Returns
Estimate log probability.
Parameters
X [array-like, shape (n_samples, n_features)] Input data.
Returns
C [array, shape (n_samples, n_classes)] Estimated log probabilities.

predict_proba(X)
Estimate probability.
Parameters
Returns
C [array, shape (n_samples, n_classes)] Estimated probabilities.
Parameters
Returns
of a nested object.
Returns
self
transform(X)
Project data to maximize class separation.
Parameters
Returns
X_new [array, shape (n_samples, n_components)] Transformed data.
Examples using sklearn.discriminant_analysis.LinearDiscriminantAnalysis
• Normal and Shrinkage Linear Discriminant Analysis for classification

• Linear and Quadratic Discriminant Analysis with covariance ellipsoid

6.10.2 sklearn.discriminant_analysis.QuadraticDiscriminantAnalysis
class sklearn.discriminant_analysis.QuadraticDiscriminantAnalysis(priors=None,
reg_param=0.0,
tol=0.0001,
store_covariances=None)
Quadratic Discriminant Analysis
A classifier with a quadratic decision boundary, generated by fitting class conditional densities to the data and
using Bayes’ rule.
The model fits a Gaussian density to each class.
New in version 0.17: QuadraticDiscriminantAnalysis
Parameters
priors [array, optional, shape = [n_classes]] Priors on classes
reg_param [float, optional] Regularizes the covariance estimate as
(1-reg_param)*Sigma + reg_param*np.eye(n_features)
store_covariance [boolean] If True the covariance matrices are computed and stored in the
self.covariance_ attribute.
tol [float, optional, default 1.0e-4] Threshold used for rank estimation.
store_covariances [boolean] Deprecated, use store_covariance.
Attributes
covariance_ [list of array-like, shape = [n_features, n_features]] Covariance matrices of each
class.
means_ [array-like, shape = [n_classes, n_features]] Class means.
priors_ [array-like, shape = [n_classes]] Class priors (sum to 1).
rotations_ [list of arrays] For each class k an array of shape [n_features, n_k], with n_k =
min(n_features, number of elements in class k) It is the rotation of the
Gaussian distribution, i.e. its principal axis.
scalings_ [list of arrays] For each class k an array of shape [n_k]. It contains the scaling of the
Gaussian distributions along its principal axes, i.e. the variance in the rotated coordinate
system.
See also:
sklearn.discriminant_analysis.LinearDiscriminantAnalysis Linear Discriminant

Analysis
Examples

>>> from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis

>>> X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]])
>>> y = np.array([1, 1, 1, 2, 2, 2])
>>> clf = QuadraticDiscriminantAnalysis()
>>> clf.fit(X, y)
...
QuadraticDiscriminantAnalysis(priors=None, reg_param=0.0,
store_covariances=None, tol=0.0001)
[1]
Methods
decision_function(X) Apply decision function to an array of samples.

fit(X, y) Fit the model according to the given training data and
parameters.
predict(X) Perform classification on an array of test vectors X.
predict_log_proba(X) Return posterior probabilities of classification.
predict_proba(X) Return posterior probabilities of classification.
labels.
__init__(priors=None, reg_param=0.0, store_covariance=False, tol=0.0001,

store_covariances=None)
covariances_
DEPRECATED: Attribute covariances_ was deprecated in version 0.19 and will be removed in 0.21.
Use covariance_ instead
Apply decision function to an array of samples.
Parameters
X [array-like, shape = [n_samples, n_features]] Array of samples (test vectors).
Returns
C [array, shape = [n_samples, n_classes] or [n_samples,]] Decision function values related
to each class, per sample. In the two-class case, the shape is [n_samples,], giving the log
likelihood ratio of the positive class.
fit(X, y)
Fit the model according to the given training data and parameters.
Changed in version 0.19: store_covariances has been moved to main constructor as
store_covariance
Changed in version 0.19: tol has been moved to main constructor.
Parameters

X [array-like, shape = [n_samples, n_features]] Training vector, where n_samples is the

y [array, shape = [n_samples]] Target values (integers)
Parameters
Returns
predict(X)
Perform classification on an array of test vectors X.
The predicted class C for each sample in X is returned.
Parameters
Returns
C [array, shape = [n_samples]]
Return posterior probabilities of classification.
Parameters
X [array-like, shape = [n_samples, n_features]] Array of samples/test vectors.
Returns
C [array, shape = [n_samples, n_classes]] Posterior log-probabilities of classification per
class.
predict_proba(X)
Return posterior probabilities of classification.
Parameters
X [array-like, shape = [n_samples, n_features]] Array of samples/test vectors.
Returns
C [array, shape = [n_samples, n_classes]] Posterior probabilities of classification per class.
Parameters
Returns


of a nested object.
Returns
self
Examples using sklearn.discriminant_analysis.QuadraticDiscriminantAnalysis
• Linear and Quadratic Discriminant Analysis with covariance ellipsoid
6.11 sklearn.dummy: Dummy estimators
User guide: See the Model evaluation: quantifying the quality of predictions section for further details.
dummy.DummyClassifier([strategy, . . . ]) DummyClassifier is a classifier that makes predictions us-

ing simple rules.
dummy.DummyRegressor([strategy, constant, . . . ]) DummyRegressor is a regressor that makes predictions us-
ing simple rules.
6.11.1 sklearn.dummy.DummyClassifier
class sklearn.dummy.DummyClassifier(strategy=’stratified’, random_state=None, constant=None)

DummyClassifier is a classifier that makes predictions using simple rules.
This classifier is useful as a simple baseline to compare with other (real) classifiers. Do not use it for real
problems.
Parameters
strategy [str, default=”stratified”] Strategy to use to generate predictions.
• “stratified”: generates predictions by respecting the training set’s class distribution.
• “most_frequent”: always predicts the most frequent label in the training set.
• “prior”: always predicts the class that maximizes the class prior (like “most_frequent”)
and predict_proba returns the class prior.
• “uniform”: generates predictions uniformly at random.
• “constant”: always predicts a constant label that is provided by the user. This is useful for
metrics that evaluate a non-majority class
New in version 0.17: Dummy Classifier now supports prior fitting strategy using param-
eter prior.

random_state [int, RandomState instance or None, optional, default=None] If int, ran-

constant [int or str or array of shape = [n_outputs]] The explicit constant as predicted by the
“constant” strategy. This parameter is useful only for the “constant” strategy.
Attributes
classes_ [array or list of array of shape = [n_classes]] Class labels for each output.
n_classes_ [array or list of array of shape = [n_classes]] Number of label for each output.
class_prior_ [array or list of array of shape = [n_classes]] Probability of each class for each
output.
n_outputs_ [int,] Number of outputs.
sparse_output_ [bool,] True if the array returned from predict is to be in sparse CSC format.
Is automatically set to True if the input y is passed in sparse format.
Methods
fit(X, y[, sample_weight]) Fit the random classifier.

predict(X) Perform classification on test vectors X.
predict_log_proba(X) Return log probability estimates for the test vectors X.
predict_proba(X) Return probability estimates for the test vectors X.
labels.
__init__(strategy=’stratified’, random_state=None, constant=None)

Fit the random classifier.
Parameters
X [{array-like, object with finite length or shape}] Training data, requires length =
n_samples
y [array-like, shape = [n_samples] or [n_samples, n_outputs]] Target values.
sample_weight [array-like of shape = [n_samples], optional] Sample weights.
Returns
self [object]
Parameters
Returns
6.11. sklearn.dummy: Dummy estimators 1607


predict(X)
Perform classification on test vectors X.
Parameters
n_samples
Returns
y [array, shape = [n_samples] or [n_samples, n_outputs]] Predicted target values for X.
Return log probability estimates for the test vectors X.
Parameters
n_samples
Returns
P [array-like or list of array-like of shape = [n_samples, n_classes]] Returns the log proba-
bility of the sample for each class in the model, where classes are ordered arithmetically
for each output.
predict_proba(X)
Return probability estimates for the test vectors X.
Parameters
n_samples
Returns
P [array-like or list of array-lke of shape = [n_samples, n_classes]] Returns the probability
of the sample for each class in the model, where classes are ordered arithmetically, for
each output.
Parameters
X [{array-like, None}] Test samples with shape = (n_samples, n_features) or None. Passing
None as test samples gives the same result as passing real test samples, since DummyClas-
sifier operates independently of the sampled observations.
Returns

of a nested object.
Returns
self
6.11.2 sklearn.dummy.DummyRegressor
class sklearn.dummy.DummyRegressor(strategy=’mean’, constant=None, quantile=None)

DummyRegressor is a regressor that makes predictions using simple rules.
This regressor is useful as a simple baseline to compare with other (real) regressors. Do not use it for real
problems.
Parameters
strategy [str] Strategy to use to generate predictions.
• “mean”: always predicts the mean of the training set
• “median”: always predicts the median of the training set
• “quantile”: always predicts a specified quantile of the training set, provided with the
quantile parameter.
• “constant”: always predicts a constant value that is provided by the user.
constant [int or float or array of shape = [n_outputs]] The explicit constant as predicted by the
“constant” strategy. This parameter is useful only for the “constant” strategy.
quantile [float in [0.0, 1.0]] The quantile to predict using the “quantile” strategy. A quantile of
0.5 corresponds to the median, while 0.0 to the minimum and 1.0 to the maximum.
Attributes
constant_ [float or array of shape [n_outputs]] Mean or median or quantile of the training tar-
gets or constant value given by the user.
n_outputs_ [int,] Number of outputs.
Methods
fit(X, y[, sample_weight]) Fit the random regressor.

predict(X[, return_std]) Perform classification on test vectors X.
diction.
__init__(strategy=’mean’, constant=None, quantile=None)

Fit the random regressor.
Parameters
6.11. sklearn.dummy: Dummy estimators 1609

n_samples
y [array-like, shape = [n_samples] or [n_samples, n_outputs]] Target values.
Returns
self [object]
Parameters
Returns
predict(X, return_std=False)
Perform classification on test vectors X.
Parameters
n_samples
return_std [boolean, optional] Whether to return the standard deviation of posterior predic-
tion. All zeros in this case.
Returns
y [array, shape = [n_samples] or [n_samples, n_outputs]] Predicted target values for X.
y_std [array, shape = [n_samples] or [n_samples, n_outputs]] Standard deviation of predic-
tive distribution of query points.
Parameters
X [{array-like, None}] Test samples with shape = (n_samples, n_features) or None. For
some estimators this may be a precomputed kernel matrix instead, shape = (n_samples,
n_samples_fitted], where n_samples_fitted is the number of samples used in the fitting
for the estimator. Passing None as test samples gives the same result as passing real test
samples, since DummyRegressor operates independently of the sampled observations.
Returns

of a nested object.
Returns
self
6.12 sklearn.ensemble: Ensemble Methods
The sklearn.ensemble module includes ensemble-based methods for classification, regression and anomaly de-
tection.
User guide: See the Ensemble methods section for further details.
ensemble.AdaBoostClassifier([. . . ]) An AdaBoost classifier.

ensemble.AdaBoostRegressor([base_estimator, An AdaBoost regressor.
. . . ])
ensemble.BaggingClassifier([base_estimator, A Bagging classifier.
. . . ])
ensemble.BaggingRegressor([base_estimator, A Bagging regressor.
. . . ])
ensemble.ExtraTreesClassifier([. . . ]) An extra-trees classifier.
ensemble.ExtraTreesRegressor([n_estimators, An extra-trees regressor.
. . . ])
ensemble.GradientBoostingClassifier([loss, Gradient Boosting for classification.
. . . ])
ensemble.GradientBoostingRegressor([loss, Gradient Boosting for regression.
. . . ])
ensemble.IsolationForest([n_estimators, . . . ]) Isolation Forest Algorithm
ensemble.RandomForestClassifier([. . . ]) A random forest classifier.
ensemble.RandomForestRegressor([. . . ]) A random forest regressor.
ensemble.RandomTreesEmbedding([. . . ]) An ensemble of totally random trees.
ensemble.VotingClassifier(estimators[, . . . ]) Soft Voting/Majority Rule classifier for unfitted estimators.
6.12.1 sklearn.ensemble.AdaBoostClassifier
class sklearn.ensemble.AdaBoostClassifier(base_estimator=None, n_estimators=50, learn-

ing_rate=1.0, algorithm=’SAMME.R’, ran-
dom_state=None)
An AdaBoost classifier.
An AdaBoost [1] classifier is a meta-estimator that begins by fitting a classifier on the original dataset and then
fits additional copies of the classifier on the same dataset but where the weights of incorrectly classified instances
are adjusted such that subsequent classifiers focus more on difficult cases.
This class implements the algorithm known as AdaBoost-SAMME [2].
6.12. sklearn.ensemble: Ensemble Methods 1611

Parameters
base_estimator [object, optional (default=None)] The base estimator from which the boosted
ensemble is built. Support for sample weighting is required, as well as proper
classes_ and n_classes_ attributes. If None, then the base estimator is
DecisionTreeClassifier(max_depth=1)
n_estimators [integer, optional (default=50)] The maximum number of estimators at which
boosting is terminated. In case of perfect fit, the learning procedure is stopped early.
learning_rate [float, optional (default=1.)] Learning rate shrinks the contribution of each
classifier by learning_rate. There is a trade-off between learning_rate and
n_estimators.
algorithm [{‘SAMME’, ‘SAMME.R’}, optional (default=’SAMME.R’)] If ‘SAMME.R’ then
use the SAMME.R real boosting algorithm. base_estimator must support calculation
of class probabilities. If ‘SAMME’ then use the SAMME discrete boosting algorithm. The
SAMME.R algorithm typically converges faster than SAMME, achieving a lower test error
with fewer boosting iterations.
Attributes
estimators_ [list of classifiers] The collection of fitted sub-estimators.
classes_ [array of shape = [n_classes]] The classes labels.
n_classes_ [int] The number of classes.
estimator_weights_ [array of floats] Weights for each estimator in the boosted ensemble.
estimator_errors_ [array of floats] Classification error for each estimator in the boosted en-
semble.
See also:
AdaBoostRegressor, GradientBoostingClassifier, sklearn.tree.
DecisionTreeClassifier
References
[1], [2]
Methods

fit(X, y[, sample_weight]) Build a boosted classifier from the training set (X, y).
predict(X) Predict classes for X.


labels.
staged_decision_function(X) Compute decision function of X for each boosting itera-
tion.
staged_predict(X) Return staged predictions for X.
staged_predict_proba(X) Predict class probabilities for X.
staged_score(X, y[, sample_weight]) Return staged scores for X, y.
__init__(base_estimator=None, n_estimators=50, learning_rate=1.0, algorithm=’SAMME.R’, ran-

dom_state=None)
Parameters
X [{array-like, sparse matrix} of shape = [n_samples, n_features]] The training input sam-
ples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. DOK and LIL are converted
to CSR.
Returns
score [array, shape = [n_samples, k]] The decision function of the input samples. The order
of outputs is the same of that of the classes_ attribute. Binary classification is a special
cases with k == 1, otherwise k==n_classes. For binary classification, values closer
to -1 or 1 mean more like the first or second class in classes_, respectively.
Returns
Build a boosted classifier from the training set (X, y).
Parameters
to CSR.
y [array-like of shape = [n_samples]] The target values (class labels).
sample_weight [array-like of shape = [n_samples], optional] Sample weights. If None, the
sample weights are initialized to 1 / n_samples.
Returns
self [object]
Parameters

Returns
predict(X)
Predict classes for X.
The predicted class of an input sample is computed as the weighted mean prediction of the classifiers in
the ensemble.
Parameters
to CSR.
Returns
y [array of shape = [n_samples]] The predicted classes.
The predicted class log-probabilities of an input sample is computed as the weighted mean predicted class
log-probabilities of the classifiers in the ensemble.
Parameters
to CSR.
Returns
p [array of shape = [n_samples, n_classes]] The class probabilities of the input samples. The
order of outputs is the same of that of the classes_ attribute.
predict_proba(X)
The predicted class probabilities of an input sample is computed as the weighted mean predicted class
probabilities of the classifiers in the ensemble.
Parameters
to CSR.
Returns
Parameters


Returns
of a nested object.
Returns
self
staged_decision_function(X)
Compute decision function of X for each boosting iteration.
This method allows monitoring (i.e. determine error on testing set) after each boosting iteration.
Parameters
to CSR.
Returns
score [generator of array, shape = [n_samples, k]] The decision function of the input sam-
ples. The order of outputs is the same of that of the classes_ attribute. Binary classification
is a special cases with k == 1, otherwise k==n_classes. For binary classification,
values closer to -1 or 1 mean more like the first or second class in classes_, respectively.
staged_predict(X)
Return staged predictions for X.
The predicted class of an input sample is computed as the weighted mean prediction of the classifiers in
the ensemble.
This generator method yields the ensemble prediction after each iteration of boosting and therefore allows
monitoring, such as to determine the prediction on a test set after each boost.
Parameters
X [array-like of shape = [n_samples, n_features]] The input samples.
Returns
y [generator of array, shape = [n_samples]] The predicted classes.
staged_predict_proba(X)
The predicted class probabilities of an input sample is computed as the weighted mean predicted class
probabilities of the classifiers in the ensemble.
This generator method yields the ensemble predicted class probabilities after each iteration of boosting
and therefore allows monitoring, such as to determine the predicted class probabilities on a test set after
each boost.
Parameters

to CSR.
Returns
p [generator of array, shape = [n_samples]] The class probabilities of the input samples. The
staged_score(X, y, sample_weight=None)
Return staged scores for X, y.
This generator method yields the ensemble score after each iteration of boosting and therefore allows
monitoring, such as to determine the score on a test set after each boost.
Parameters
to CSR.
y [array-like, shape = [n_samples]] Labels for X.
Returns
z [float]
Examples using sklearn.ensemble.AdaBoostClassifier
6.12.2 sklearn.ensemble.AdaBoostRegressor
class sklearn.ensemble.AdaBoostRegressor(base_estimator=None, n_estimators=50, learn-

ing_rate=1.0, loss=’linear’, random_state=None)
An AdaBoost regressor.
An AdaBoost [1] regressor is a meta-estimator that begins by fitting a regressor on the original dataset and
then fits additional copies of the regressor on the same dataset but where the weights of instances are adjusted
according to the error of the current prediction. As such, subsequent regressors focus more on difficult cases.
This class implements the algorithm known as AdaBoost.R2 [2].
Parameters
base_estimator [object, optional (default=None)] The base estimator from which the boosted
ensemble is built. Support for sample weighting is required. If None, then the base estima-
tor is DecisionTreeRegressor(max_depth=3)
n_estimators [integer, optional (default=50)] The maximum number of estimators at which
boosting is terminated. In case of perfect fit, the learning procedure is stopped early.

learning_rate [float, optional (default=1.)] Learning rate shrinks the contribution of each
regressor by learning_rate. There is a trade-off between learning_rate and
n_estimators.
loss [{‘linear’, ‘square’, ‘exponential’}, optional (default=’linear’)] The loss function to use
when updating the weights after each boosting iteration.
Attributes
estimators_ [list of classifiers] The collection of fitted sub-estimators.
estimator_weights_ [array of floats] Weights for each estimator in the boosted ensemble.
estimator_errors_ [array of floats] Regression error for each estimator in the boosted ensem-
ble.
See also:
AdaBoostClassifier, GradientBoostingRegressor, sklearn.tree.
DecisionTreeRegressor
References
[1], [2]
Methods
fit(X, y[, sample_weight]) Build a boosted regressor from the training set (X, y).
predict(X) Predict regression value for X.
diction.
staged_predict(X) Return staged predictions for X.
staged_score(X, y[, sample_weight]) Return staged scores for X, y.
__init__(base_estimator=None, n_estimators=50, learning_rate=1.0, loss=’linear’, ran-

dom_state=None)
Returns
Build a boosted regressor from the training set (X, y).

Parameters
to CSR.
y [array-like of shape = [n_samples]] The target values (real numbers).
sample_weight [array-like of shape = [n_samples], optional] Sample weights. If None, the
sample weights are initialized to 1 / n_samples.
Returns
self [object]
Parameters
Returns
predict(X)
Predict regression value for X.
The predicted regression value of an input sample is computed as the weighted median prediction of the
classifiers in the ensemble.
Parameters
to CSR.
Returns
y [array of shape = [n_samples]] The predicted regression values.
Parameters
Returns

of a nested object.
Returns
self
staged_predict(X)
Return staged predictions for X.
The predicted regression value of an input sample is computed as the weighted median prediction of the
classifiers in the ensemble.
This generator method yields the ensemble prediction after each iteration of boosting and therefore allows
monitoring, such as to determine the prediction on a test set after each boost.
Parameters
to CSR.
Returns
y [generator of array, shape = [n_samples]] The predicted regression values.
staged_score(X, y, sample_weight=None)
Return staged scores for X, y.
This generator method yields the ensemble score after each iteration of boosting and therefore allows
monitoring, such as to determine the score on a test set after each boost.
Parameters
to CSR.
y [array-like, shape = [n_samples]] Labels for X.
Returns
z [float]
Examples using sklearn.ensemble.AdaBoostRegressor
• Decision Tree Regression with AdaBoost
6.12.3 sklearn.ensemble.BaggingClassifier
class sklearn.ensemble.BaggingClassifier(base_estimator=None, n_estimators=10,

max_samples=1.0, max_features=1.0, boot-
strap=True, bootstrap_features=False,
oob_score=False, warm_start=False, n_jobs=None,
random_state=None, verbose=0)
A Bagging classifier.

A Bagging classifier is an ensemble meta-estimator that fits base classifiers each on random subsets of the
original dataset and then aggregate their individual predictions (either by voting or by averaging) to form a
final prediction. Such a meta-estimator can typically be used as a way to reduce the variance of a black-box
estimator (e.g., a decision tree), by introducing randomization into its construction procedure and then making
an ensemble out of it.
This algorithm encompasses several works from the literature. When random subsets of the dataset are drawn
as random subsets of the samples, then this algorithm is known as Pasting [1]. If samples are drawn with re-
placement, then the method is known as Bagging [2]. When random subsets of the dataset are drawn as random
subsets of the features, then the method is known as Random Subspaces [3]. Finally, when base estimators are
built on subsets of both samples and features, then the method is known as Random Patches [4].
Parameters
base_estimator [object or None, optional (default=None)] The base estimator to fit on random
subsets of the dataset. If None, then the base estimator is a decision tree.
n_estimators [int, optional (default=10)] The number of base estimators in the ensemble.
max_samples [int or float, optional (default=1.0)] The number of samples to draw from X to
train each base estimator.
• If int, then draw max_samples samples.
• If float, then draw max_samples * X.shape[0] samples.
max_features [int or float, optional (default=1.0)] The number of features to draw from X to
• If int, then draw max_features features.
• If float, then draw max_features * X.shape[1] features.
bootstrap [boolean, optional (default=True)] Whether samples are drawn with replacement. If
False, sampling without replacement is performed.
bootstrap_features [boolean, optional (default=False)] Whether features are drawn with re-
placement.
generalization error.
ensemble. See the Glossary.
New in version 0.17: warm_start constructor parameter.
Attributes
base_estimator_ [estimator] The base estimator from which the ensemble is grown.

estimators_ [list of estimators] The collection of fitted base estimators.

estimators_samples_ [list of arrays] The subset of drawn samples for each base estima-
tor.
estimators_features_ [list of arrays] The subset of drawn features for each base estimator.
classes_ [array of shape = [n_classes]] The classes labels.
n_classes_ [int or list] The number of classes.
References
[1], [2], [3], [4]
Methods
decision_function(X) Average of the decision functions of the base classifiers.

fit(X, y[, sample_weight]) Build a Bagging ensemble of estimators from the train-
ing set (X, y).
labels.
__init__(base_estimator=None, n_estimators=10, max_samples=1.0, max_features=1.0, boot-

strap=True, bootstrap_features=False, oob_score=False, warm_start=False, n_jobs=None,
Average of the decision functions of the base classifiers.
Parameters
ples. Sparse matrices are accepted only if they are supported by the base estimator.
Returns
score [array, shape = [n_samples, k]] The decision function of the input samples. The
columns correspond to the classes in sorted order, as they appear in the attribute
classes_. Regression and binary classification are special cases with k == 1, oth-
erwise k==n_classes.
estimators_samples_
The subset of drawn samples for each base estimator.

Returns a dynamically generated list of indices identifying the samples used for fitting each member of the
ensemble, i.e., the in-bag samples.
Note: the list is re-created at each call to the property in order to reduce the object memory footprint by
not storing the sampling data. Thus fetching the property may be slower than expected.
Build a Bagging ensemble of estimators from the training set (X, y).
Parameters
y [array-like, shape = [n_samples]] The target values (class labels in classification, real num-
bers in regression).
samples are equally weighted. Note that this is supported only if the base estimator sup-
ports sample weighting.
Returns
self [object]
Parameters
Returns
predict(X)
The predicted class of an input sample is computed as the class with the highest mean predicted probability.
If base estimators do not implement a predict_proba method, then it resorts to voting.
Parameters
Returns
y [array of shape = [n_samples]] The predicted classes.
probabilities of the base estimators in the ensemble.
Parameters
Returns

p [array of shape = [n_samples, n_classes]] The class log-probabilities of the input samples.
The order of the classes corresponds to that in the attribute classes_.
predict_proba(X)
The predicted class probabilities of an input sample is computed as the mean predicted class probabilities
of the base estimators in the ensemble. If base estimators do not implement a predict_proba method,
then it resorts to voting and the predicted class probabilities of an input sample represents the proportion
of estimators predicting each class.
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
6.12.4 sklearn.ensemble.BaggingRegressor
class sklearn.ensemble.BaggingRegressor(base_estimator=None, n_estimators=10,

max_samples=1.0, max_features=1.0, boot-
strap=True, bootstrap_features=False,
oob_score=False, warm_start=False, n_jobs=None,
A Bagging regressor.
A Bagging regressor is an ensemble meta-estimator that fits base regressors each on random subsets of the
original dataset and then aggregate their individual predictions (either by voting or by averaging) to form a
final prediction. Such a meta-estimator can typically be used as a way to reduce the variance of a black-box

estimator (e.g., a decision tree), by introducing randomization into its construction procedure and then making
an ensemble out of it.
This algorithm encompasses several works from the literature. When random subsets of the dataset are drawn
as random subsets of the samples, then this algorithm is known as Pasting [1]. If samples are drawn with re-
placement, then the method is known as Bagging [2]. When random subsets of the dataset are drawn as random
subsets of the features, then the method is known as Random Subspaces [3]. Finally, when base estimators are
built on subsets of both samples and features, then the method is known as Random Patches [4].
Parameters
base_estimator [object or None, optional (default=None)] The base estimator to fit on random
subsets of the dataset. If None, then the base estimator is a decision tree.
max_samples [int or float, optional (default=1.0)] The number of samples to draw from X to
bootstrap [boolean, optional (default=True)] Whether samples are drawn with replacement. If
False, sampling without replacement is performed.
bootstrap_features [boolean, optional (default=False)] Whether features are drawn with re-
placement.
oob_score [bool] Whether to use out-of-bag samples to estimate the generalization error.
ensemble. See the Glossary.
Attributes
estimators_ [list of estimators] The collection of fitted sub-estimators.
tor.
estimators_features_ [list of arrays] The subset of drawn features for each base estimator.


on the training set. If n_estimators is small it might be possible that a data point was never
left out during the bootstrap. In this case, oob_prediction_ might contain NaN.
References
[1], [2], [3], [4]
Methods
fit(X, y[, sample_weight]) Build a Bagging ensemble of estimators from the train-
ing set (X, y).
diction.
__init__(base_estimator=None, n_estimators=10, max_samples=1.0, max_features=1.0, boot-

strap=True, bootstrap_features=False, oob_score=False, warm_start=False, n_jobs=None,
estimators_samples_
Build a Bagging ensemble of estimators from the training set (X, y).
Parameters
samples are equally weighted. Note that this is supported only if the base estimator sup-
ports sample weighting.
Returns
self [object]
Parameters

Returns
predict(X)
the estimators in the ensemble.
Parameters
Returns
y [array of shape = [n_samples]] The predicted values.
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.ensemble.BaggingRegressor

6.12.5 sklearn.ensemble.IsolationForest
class sklearn.ensemble.IsolationForest(n_estimators=100, max_samples=’auto’, contamina-

tion=’legacy’, max_features=1.0, bootstrap=False,
n_jobs=None, behaviour=’old’, random_state=None,
verbose=0)
Isolation Forest Algorithm
Return the anomaly score of each sample using the IsolationForest algorithm
The IsolationForest ‘isolates’ observations by randomly selecting a feature and then randomly selecting a split
value between the maximum and minimum values of the selected feature.
Since recursive partitioning can be represented by a tree structure, the number of splittings required to isolate a
sample is equivalent to the path length from the root node to the terminating node.
This path length, averaged over a forest of such random trees, is a measure of normality and our decision
function.
Random partitioning produces noticeably shorter paths for anomalies. Hence, when a forest of random trees
collectively produce shorter path lengths for particular samples, they are highly likely to be anomalies.
Parameters
max_samples [int or float, optional (default=”auto”)]
The number of samples to draw from X to train each base estimator.
• If “auto”, then max_samples=min(256, n_samples).
If max_samples is larger than the number of samples provided, all samples will be used for
all trees (no sampling).
data set, i.e. the proportion of outliers in the data set. Used when fitting to define the
threshold on the decision function. If ‘auto’, the decision function threshold is determined
as in the original paper.
Changed in version 0.20: The default value of contamination will change from 0.1 in
0.20 to 'auto' in 0.22.
bootstrap [boolean, optional (default=False)] If True, individual trees are fit on random subsets
of the training data sampled with replacement. If False, sampling without replacement is
performed.

behaviour [str, default=’old’] Behaviour of the decision_function which can be ei-

ther ‘old’ or ‘new’. Passing behaviour='new' makes the decision_function
change to match other anomaly detection algorithm API which will be the default be-
haviour in the future. As explained in details in the offset_ attribute documentation,
the decision_function becomes dependent on the contamination parameter, in such
a way that 0 becomes its natural threshold to detect outliers.
New in version 0.20: behaviour is added in 0.20 for back-compatibility purpose.
Deprecated since version 0.20: behaviour='old' is deprecated in 0.20 and will not be
possible in 0.22.
Deprecated since version 0.22: behaviour parameter will be deprecated in 0.22 and re-
moved in 0.24.
verbose [int, optional (default=0)] Controls the verbosity of the tree building process.
Attributes
tor.
max_samples_ [integer] The actual number of samples
offset_ [float] Offset used to define the decision function from the raw scores. We have the
relation: decision_function = score_samples - offset_. Assuming be-
haviour == ‘new’, offset_ is defined as follows. When the contamination parameter
is set to “auto”, the offset is equal to -0.5 as the scores of inliers are close to 0 and the
scores of outliers are close to -1. When a contamination parameter different than “auto”
is provided, the offset is defined in such a way we obtain the expected number of outliers
(samples with decision function < 0) in training. Assuming the behaviour parameter is set
to ‘old’, we always have offset_ = -0.5, making the decision function independent
from the contamination parameter.
References
[1], [2]
Methods
decision_function(X) Average anomaly score of X of the base classifiers.

fit(X[, y, sample_weight]) Fit estimator.
predict(X) Predict if a particular sample is an outlier or not.
score_samples(X) Opposite of the anomaly score defined in the original
paper.

__init__(n_estimators=100, max_samples=’auto’, contamination=’legacy’, max_features=1.0, boot-

strap=False, n_jobs=None, behaviour=’old’, random_state=None, verbose=0)
Average anomaly score of X of the base classifiers.
The anomaly score of an input sample is computed as the mean anomaly score of the trees in the forest.
The measure of normality of an observation given a tree is the depth of the leaf containing this observation,
which is equivalent to the number of splittings required to isolate this point. In case of several observations
n_left in the leaf, the average path length of a n_left samples isolation tree is added.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] The training input samples.
Sparse matrices are accepted only if they are supported by the base estimator.
Returns
scores [array, shape (n_samples,)] The anomaly score of the input samples. The lower, the
more abnormal. Negative scores represent outliers, positive scores represent inliers.
estimators_samples_
Fit estimator.
Parameters
X [array-like or sparse matrix, shape (n_samples, n_features)] The input samples. Use
dtype=np.float32 for maximum efficiency. Sparse matrices are also supported, use
sparse csc_matrix for maximum efficiency.
samples are equally weighted.
Returns
self [object]
Parameters
Returns

Parameters
Returns
predict(X)
Predict if a particular sample is an outlier or not.
Parameters
X [array-like or sparse matrix, shape (n_samples, n_features)] The input samples. Internally,
it will be converted to dtype=np.float32 and if a sparse matrix is provided to a sparse
csr_matrix.
Returns
is_inlier [array, shape (n_samples,)] For each observation, tells whether or not (+1 or -1) it
should be considered as an inlier according to the fitted model.
score_samples(X)
Opposite of the anomaly score defined in the original paper.
The anomaly score of an input sample is computed as the mean anomaly score of the trees in the forest.
The measure of normality of an observation given a tree is the depth of the leaf containing this observation,
which is equivalent to the number of splittings required to isolate this point. In case of several observations
n_left in the leaf, the average path length of a n_left samples isolation tree is added.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] The training input samples.
Sparse matrices are accepted only if they are supported by the base estimator.
Returns
scores [array, shape (n_samples,)] The anomaly score of the input samples. The lower, the
more abnormal.
of a nested object.
Returns
self
Examples using sklearn.ensemble.IsolationForest

• IsolationForest example

6.12.6 sklearn.ensemble.RandomTreesEmbedding
class sklearn.ensemble.RandomTreesEmbedding(n_estimators=’warn’, max_depth=5,

max_leaf_nodes=None,
min_impurity_split=None, sparse_output=True,
n_jobs=None, random_state=None, verbose=0,
warm_start=False)
An ensemble of totally random trees.
An unsupervised transformation of a dataset to a high-dimensional sparse representation. A datapoint is coded
according to which leaf of each tree it is sorted into. Using a one-hot encoding of the leaves, this leads to a
binary coding with as many ones as there are trees in the forest.
The dimensionality of the resulting representation is n_out <= n_estimators * max_leaf_nodes.
If max_leaf_nodes == None, the number of leaf nodes is at most n_estimators * 2 **
max_depth.
Parameters
n_estimators [integer, optional (default=10)] Number of trees in the forest.
max_depth [integer, optional (default=5)] The maximum depth of each tree. If None,
then nodes are expanded until all leaves are pure or until all leaves contain less than
• If float, then min_samples_split is a fraction and ceil(min_samples_split * n_samples) is
• If float, then min_samples_leaf is a fraction and ceil(min_samples_leaf * n_samples) is


the right child.
sparse_output [bool, optional (default=True)] Whether or not to return a sparse CSR matrix,
as default behavior, or to return a dense array compatible with dense pipeline operators.
Attributes
References
[1], [2]
Methods

fit(X[, y, sample_weight]) Fit estimator.
fit_transform(X[, y, sample_weight]) Fit estimator and transform dataset.


transform(X) Transform dataset.
__init__(n_estimators=’warn’, max_depth=5, min_samples_split=2, min_samples_leaf=1,

min_weight_fraction_leaf=0.0, max_leaf_nodes=None, min_impurity_decrease=0.0,
min_impurity_split=None, sparse_output=True, n_jobs=None, random_state=None,
verbose=0, warm_start=False)
apply(X)
Parameters
Returns
decision_path(X)
Parameters
Returns
Returns

Fit estimator.
Parameters
X [array-like or sparse matrix, shape=(n_samples, n_features)] The input samples. Use
dtype=np.float32 for maximum efficiency. Sparse matrices are also supported, use
sparse csc_matrix for maximum efficiency.


Returns
self [object]
Fit estimator and transform dataset.
Parameters
X [array-like or sparse matrix, shape=(n_samples, n_features)] Input data used to build
forests. Use dtype=np.float32 for maximum efficiency.
Returns
X_transformed [sparse matrix, shape=(n_samples, n_out)] Transformed dataset.
Parameters
Returns
of a nested object.
Returns
self
transform(X)
Transform dataset.
Parameters
X [array-like or sparse matrix, shape=(n_samples, n_features)] Input data to be transformed.
Use dtype=np.float32 for maximum efficiency. Sparse matrices are also supported,
use sparse csr_matrix for maximum efficiency.
Returns
X_transformed [sparse matrix, shape=(n_samples, n_out)] Transformed dataset.

Examples using sklearn.ensemble.RandomTreesEmbedding

6.12.7 sklearn.ensemble.VotingClassifier
class sklearn.ensemble.VotingClassifier(estimators, voting=’hard’, weights=None,

n_jobs=None, flatten_transform=None)
Soft Voting/Majority Rule classifier for unfitted estimators.
Parameters
estimators [list of (string, estimator) tuples] Invoking the fit method on the
VotingClassifier will fit clones of those original estimators that will be stored
in the class attribute self.estimators_. An estimator can be set to None using
set_params.
voting [str, {‘hard’, ‘soft’} (default=’hard’)] If ‘hard’, uses predicted class labels for majority
rule voting. Else if ‘soft’, predicts the class label based on the argmax of the sums of the pre-
dicted probabilities, which is recommended for an ensemble of well-calibrated classifiers.
weights [array-like, shape = [n_classifiers], optional (default=‘None‘)] Sequence of weights
(float or int) to weight the occurrences of predicted class labels (hard voting) or class prob-
abilities before averaging (soft voting). Uses uniform weights if None.
n_jobs [int or None, optional (default=None)] The number of jobs to run in parallel for fit.
flatten_transform [bool, optional (default=None)] Affects shape of transform output only
when voting=’soft’ If voting=’soft’ and flatten_transform=True, transform method returns
matrix with shape (n_samples, n_classifiers * n_classes). If flatten_transform=False, it re-
turns (n_classifiers, n_samples, n_classes).
Attributes
estimators_ [list of classifiers] The collection of fitted sub-estimators as defined in
estimators that are not None.
named_estimators_ [Bunch object, a dictionary with attribute access] Attribute to access any
fitted sub-estimators by name.
classes_ [array-like, shape = [n_predictions]] The classes labels.
Examples


>>> from sklearn.ensemble import RandomForestClassifier, VotingClassifier

... random_state=1)
>>> clf2 = RandomForestClassifier(n_estimators=50, random_state=1)
>>> X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]])
>>> y = np.array([1, 1, 1, 2, 2, 2])
>>> eclf1 = VotingClassifier(estimators=[
... ('lr', clf1), ('rf', clf2), ('gnb', clf3)], voting='hard')
>>> eclf1 = eclf1.fit(X, y)
>>> print(eclf1.predict(X))
[1 1 1 2 2 2]
>>> np.array_equal(eclf1.named_estimators_.lr.predict(X),
... eclf1.named_estimators_['lr'].predict(X))
True
... ('lr', clf1), ('rf', clf2), ('gnb', clf3)],
... voting='soft')
[1 1 1 2 2 2]
... ('lr', clf1), ('rf', clf2), ('gnb', clf3)],
... voting='soft', weights=[2,1,1],
... flatten_transform=True)
[1 1 1 2 2 2]
>>> print(eclf3.transform(X).shape)
(6, 6)
Methods
fit(X, y[, sample_weight]) Fit the estimators.

get_params([deep]) Get the parameters of the VotingClassifier
predict(X) Predict class labels for X.
labels.
set_params(**params) Setting the parameters for the voting classifier
transform(X) Return class labels or probabilities for X for each esti-
mator.
__init__(estimators, voting=’hard’, weights=None, n_jobs=None, flatten_transform=None)

Fit the estimators.
Parameters
y [array-like, shape = [n_samples]] Target values.


samples are equally weighted. Note that this is supported only if all underlying estimators
support sample weights.
Returns
self [object]
Parameters
Returns
Get the parameters of the VotingClassifier
Parameters
deep [bool] Setting it to True gets the various classifiers and the parameters of the classifiers
as well
predict(X)
Predict class labels for X.
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] The input samples.
Returns
maj [array-like, shape = [n_samples]] Predicted class labels.
predict_proba
Compute probabilities of possible outcomes for samples in X.
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] The input samples.
Returns
avg [array-like, shape = [n_samples, n_classes]] Weighted average probability for each class
per sample.
Parameters

Returns
Setting the parameters for the voting classifier
Parameters
**params [keyword arguments] Specific parameters using e.g.
set_params(parameter_name=new_value) In addition, to setting the parameters of
the VotingClassifier, the individual classifiers of the VotingClassifier can
also be set or replaced by setting them to None.
Examples
# In this example, the RandomForestClassifier is removed clf1 = LogisticRegression() clf2 = Random-

ForestClassifier() eclf = VotingClassifier(estimators=[(‘lr’, clf1), (‘rf’, clf2)] eclf.set_params(rf=None)
transform(X)
Return class labels or probabilities for X for each estimator.
Parameters
Returns
probabilities_or_labels
If voting=’soft’ and flatten_transform=True: returns array-like of shape (n_classifiers,
n_samples * n_classes), being class probabilities calculated by each classifier.
If voting=’soft’ and ‘flatten_transform=False: array-like of shape (n_classifiers,
n_samples, n_classes)
If voting=’hard’: array-like of shape (n_samples, n_classifiers), being class labels pre-
dicted by each classifier.
Examples using sklearn.ensemble.VotingClassifier

6.12.8 partial dependence
Partial dependence plots for tree ensembles.
ensemble.partial_dependence. Partial dependence of target_variables.

partial_dependence(. . . )
ensemble.partial_dependence. Partial dependence plots for features.
plot_partial_dependence(. . . )

sklearn.ensemble.partial_dependence.partial_dependence
sklearn.ensemble.partial_dependence.partial_dependence(gbrt, target_variables,
grid=None, X=None,
percentiles=(0.05, 0.95),
grid_resolution=100)
Partial dependence of target_variables.
Partial dependence plots show the dependence between the joint values of the target_variables and the
function represented by the gbrt.
Parameters
gbrt [BaseGradientBoosting] A fitted gradient boosting model.
target_variables [array-like, dtype=int] The target features for which the partial dependecy
should be computed (size should be smaller than 3 for visual renderings).
grid [array-like, shape=(n_points, len(target_variables))] The grid of target_variables
values for which the partial dependecy should be evaluated (either grid or X must be
specified).
X [array-like, shape=(n_samples, n_features)] The data on which gbrt was trained. It
is used to generate a grid for the target_variables. The grid comprises
grid_resolution equally spaced points between the two percentiles.
percentiles [(low, high), default=(0.05, 0.95)] The lower and upper percentile used create the
extreme values for the grid. Only if X is not None.
grid_resolution [int, default=100] The number of equally spaced points on the grid.
Returns
pdp [array, shape=(n_classes, n_points)] The partial dependence function evaluated on the
grid. For regression and binary classification n_classes==1.
axes [seq of ndarray or None] The axes with which the grid has been created or None if the grid
has been given.
Examples
>>> samples = [[0, 0, 2], [1, 0, 0]]

>>> labels = [0, 1]
>>> gb = GradientBoostingClassifier(random_state=0).fit(samples, labels)
>>> kwargs = dict(X=samples, percentiles=(0, 1), grid_resolution=2)
>>> partial_dependence(gb, [0], **kwargs)
(array([[-4.52..., 4.52...]]), [array([ 0., 1.])])
Examples using sklearn.ensemble.partial_dependence.partial_dependence

sklearn.ensemble.partial_dependence.plot_partial_dependence
sklearn.ensemble.partial_dependence.plot_partial_dependence(gbrt, X, features, fea-

ture_names=None,
label=None,
n_cols=3,
grid_resolution=100,
percentiles=(0.05,
0.95), n_jobs=None,
verbose=0,
ax=None,
line_kw=None,
contour_kw=None,
**fig_kw)
Partial dependence plots for features.
The len(features) plots are arranged in a grid with n_cols columns. Two-way partial dependence plots
are plotted as contour plots.
Parameters
gbrt [BaseGradientBoosting] A fitted gradient boosting model.
X [array-like, shape=(n_samples, n_features)] The data on which gbrt was trained.
features [seq of ints, strings, or tuples of ints or strings] If seq[i] is an int or a tuple with one int
value, a one-way PDP is created; if seq[i] is a tuple of two ints, a two-way PDP is created.
If feature_names is specified and seq[i] is an int, seq[i] must be < len(feature_names). If
seq[i] is a string, feature_names must be specified, and seq[i] must be in feature_names.
feature_names [seq of str] Name of each feature; feature_names[i] holds the name of the fea-
ture with index i.
label [object] The class label for which the PDPs should be computed. Only if gbrt is a multi-
class model. Must be in gbrt.classes_.
n_cols [int] The number of columns in the grid plot (default: 3).
grid_resolution [int, default=100] The number of equally spaced points on the axes.
percentiles [(low, high), default=(0.05, 0.95)] The lower and upper percentile used to create the
extreme values for the PDP axes.
n_jobs [int or None, optional (default=None)] None means 1 unless in a joblib.
parallel_backend context. -1 means using all processors. See Glossary for more
details.
verbose [int] Verbose output during PD computations. Defaults to 0.
ax [Matplotlib axis object, default None] An axis object onto which the plots will be drawn.
line_kw [dict] Dict with keywords passed to the matplotlib.pyplot.plot call. For one-
way partial dependence plots.
contour_kw [dict] Dict with keywords passed to the matplotlib.pyplot.plot call. For
two-way partial dependence plots.
**fig_kw [dict] Dict with keywords passed to the figure() call. Note that all keywords not
recognized above will be automatically included here.
Returns

fig [figure] The Matplotlib Figure object.

axs [seq of Axis objects] A seq of Axis objects, one for each subplot.
Examples

>>> X, y = make_friedman1()
>>> clf = GradientBoostingRegressor(n_estimators=10).fit(X, y)
>>> fig, axs = plot_partial_dependence(clf, X, [0, (0, 1)])
...
Examples using sklearn.ensemble.partial_dependence.plot_partial_dependence
6.13 sklearn.exceptions: Exceptions and warnings
The sklearn.exceptions module includes all custom warnings and error classes used across scikit-learn.
exceptions.ChangedBehaviorWarning Warning class used to notify the user of any change in the
behavior.
exceptions.ConvergenceWarning Custom warning to capture convergence problems
exceptions.DataConversionWarning Warning used to notify implicit data conversions happening
in the code.
exceptions.DataDimensionalityWarning Custom warning to notify potential issues with data dimen-
sionality.
exceptions.EfficiencyWarning Warning used to notify the user of inefficient computation.
exceptions.FitFailedWarning Warning class used if there is an error while fitting the es-
timator.
exceptions.NotFittedError Exception class to raise if estimator is used before fitting.
exceptions.NonBLASDotWarning Warning used when the dot operation does not use BLAS.
exceptions.UndefinedMetricWarning Warning used when the metric is invalid
6.13.1 sklearn.exceptions.ChangedBehaviorWarning
class sklearn.exceptions.ChangedBehaviorWarning
Warning class used to notify the user of any change in the behavior.
Changed in version 0.18: Moved from sklearn.base.
Attributes
args
Methods
6.13. sklearn.exceptions: Exceptions and warnings 1641

with_traceback Exception.with_traceback(tb) – set self.__traceback__

to tb and return self.
with_traceback()
Exception.with_traceback(tb) – set self.__traceback__ to tb and return self.
6.13.2 sklearn.exceptions.ConvergenceWarning
class sklearn.exceptions.ConvergenceWarning
Custom warning to capture convergence problems
Changed in version 0.18: Moved from sklearn.utils.
Attributes
args
Methods

with_traceback()
Examples using sklearn.exceptions.ConvergenceWarning

6.13.3 sklearn.exceptions.DataConversionWarning
class sklearn.exceptions.DataConversionWarning
Warning used to notify implicit data conversions happening in the code.
This warning occurs when some input data needs to be converted or interpreted in a way that may not match the
user’s expectations.
For example, this warning may occur when the user
• passes an integer array to a function which expects float input and will convert the input
• requests a non-copying operation, but a copy is required to meet the implementation’s data-type ex-
pectations;
• passes an input whose shape can be interpreted ambiguously.
Changed in version 0.18: Moved from sklearn.utils.validation.
Attributes
args

Methods

with_traceback()
6.13.4 sklearn.exceptions.DataDimensionalityWarning
class sklearn.exceptions.DataDimensionalityWarning
Custom warning to notify potential issues with data dimensionality.
For example, in random projection, this warning is raised when the number of components, which quantifies
the dimensionality of the target projection space, is higher than the number of features, which quantifies the
dimensionality of the original source space, to imply that the dimensionality of the problem will not be reduced.
Changed in version 0.18: Moved from sklearn.utils.
Attributes
args
Methods

with_traceback()
6.13.5 sklearn.exceptions.EfficiencyWarning
class sklearn.exceptions.EfficiencyWarning
Warning used to notify the user of inefficient computation.
This warning notifies the user that the efficiency may not be optimal due to some reason which may be included
as a part of the warning message. This may be subclassed into a more specific Warning class.
Attributes
args
Methods


with_traceback()
6.13.6 sklearn.exceptions.FitFailedWarning
class sklearn.exceptions.FitFailedWarning
Warning class used if there is an error while fitting the estimator.
This Warning is used in meta estimators GridSearchCV and RandomizedSearchCV and the cross-validation
helper function cross_val_score to warn when there is an error while fitting the estimator.
Attributes
args
Examples

>>> from sklearn.exceptions import FitFailedWarning
>>> import warnings
>>> warnings.simplefilter('always', FitFailedWarning)
>>> gs = GridSearchCV(LinearSVC(), {'C': [-1, -2]}, error_score=0, cv=2)
>>> X, y = [[1, 2], [3, 4], [5, 6], [7, 8]], [0, 0, 1, 1]
>>> with warnings.catch_warnings(record=True) as w:
... try:
... gs.fit(X, y) # This will raise a ValueError since C is < 0
... except ValueError:
... pass
... print(repr(w[-1].message))
...
FitFailedWarning('Estimator fit failed. The score on this train-test
partition for these parameters will be set to 0.000000.
Details: \nValueError: Penalty term must be positive; got (C=-2)\n'...)
Changed in version 0.18: Moved from sklearn.cross_validation.
Methods

with_traceback()
6.13.7 sklearn.exceptions.NotFittedError
class sklearn.exceptions.NotFittedError
Exception class to raise if estimator is used before fitting.
This class inherits from both ValueError and AttributeError to help with exception handling and backward
compatibility.

Attributes
args
Examples

>>> from sklearn.exceptions import NotFittedError
>>> try:
... LinearSVC().predict([[1, 2], [2, 3], [3, 4]])
... except NotFittedError as e:
... print(repr(e))
...
NotFittedError('This LinearSVC instance is not fitted yet'...)
Changed in version 0.18: Moved from sklearn.utils.validation.
Methods

with_traceback()
6.13.8 sklearn.exceptions.NonBLASDotWarning
class sklearn.exceptions.NonBLASDotWarning
Warning used when the dot operation does not use BLAS.
This warning is used to notify the user that BLAS was not used for dot operation and hence the efficiency may
be affected.
Changed in version 0.18: Moved from sklearn.utils.validation, extends EfficiencyWarning.
Attributes
args
Methods

with_traceback()

6.13.9 sklearn.exceptions.UndefinedMetricWarning
class sklearn.exceptions.UndefinedMetricWarning
Warning used when the metric is invalid
Changed in version 0.18: Moved from sklearn.base.
Attributes
args
Methods

with_traceback()
6.14 sklearn.feature_extraction: Feature Extraction
The sklearn.feature_extraction module deals with feature extraction from raw data. It currently includes
methods to extract features from text and images.
User guide: See the Feature extraction section for further details.
feature_extraction.DictVectorizer([dtype, Transforms lists of feature-value mappings to vectors.

. . . ])
feature_extraction.FeatureHasher([. . . ]) Implements feature hashing, aka the hashing trick.
6.14.1 sklearn.feature_extraction.DictVectorizer
class sklearn.feature_extraction.DictVectorizer(dtype=<class ‘numpy.float64’>, separa-

tor=’=’, sparse=True, sort=True)
Transforms lists of feature-value mappings to vectors.
This transformer turns lists of mappings (dict-like objects) of feature names to feature values into Numpy arrays
or scipy.sparse matrices for use with scikit-learn estimators.
When feature values are strings, this transformer will do a binary one-hot (aka one-of-K) coding: one boolean-
valued feature is constructed for each of the possible string values that the feature can take on. For instance, a
feature “f” that can take on the values “ham” and “spam” will become two features in the output, one signifying
“f=ham”, the other “f=spam”.
However, note that this transformer will only do a binary one-hot encoding when feature values are of type
string. If categorical features are represented as numeric values such as int, the DictVectorizer can be followed
by sklearn.preprocessing.OneHotEncoder to complete binary one-hot encoding.
Features that do not occur in a sample (mapping) will have a zero value in the resulting array/matrix.
Parameters

dtype [callable, optional] The type of feature values. Passed to Numpy array/scipy.sparse ma-
trix constructors as the dtype argument.
separator [string, optional] Separator string used when constructing new features for one-hot
coding.
sparse [boolean, optional.] Whether transform should produce scipy.sparse matrices. True by
default.
sort [boolean, optional.] Whether feature_names_ and vocabulary_ should be sorted
when fitting. True by default.
Attributes
vocabulary_ [dict] A dictionary mapping feature names to feature indices.
feature_names_ [list] A list of length n_features containing the feature names (e.g., “f=ham”
and “f=spam”).
See also:
FeatureHasher performs vectorization using only a hash function.

sklearn.preprocessing.OrdinalEncoder handles nominal/categorical features encoded as
columns of arbitrary data types.
Examples

>>> v = DictVectorizer(sparse=False)
>>> D = [{'foo': 1, 'bar': 2}, {'foo': 3, 'baz': 1}]
>>> X = v.fit_transform(D)
>>> X
array([[2., 0., 1.],
[0., 1., 3.]])
>>> v.inverse_transform(X) == [{'bar': 2.0, 'foo': 1.0}, {'baz': 1.0, 'foo
˓→': 3.0}]
True
>>> v.transform({'foo': 4, 'unseen_feature': 3})
array([[0., 0., 4.]])
Methods
fit(X[, y]) Learn a list of feature name -> indices mappings.

fit_transform(X[, y]) Learn a list of feature name -> indices mappings and
transform X.
get_feature_names() Returns a list of feature names, ordered by their indices.
inverse_transform(X[, dict_type]) Transform array or sparse matrix X back to feature map-
pings.
restrict(support[, indices]) Restrict the features to those in support using feature
selection.
6.14. sklearn.feature_extraction: Feature Extraction 1647


transform(X) Transform feature->value dicts to array or sparse ma-
trix.
__init__(dtype=<class ‘numpy.float64’>, separator=’=’, sparse=True, sort=True)

fit(X, y=None)
Learn a list of feature name -> indices mappings.
Parameters
X [Mapping or iterable over Mappings] Dict(s) or Mapping(s) from feature names (arbitrary
Python objects) to feature values (strings or convertible to dtype).
y [(ignored)]
Returns
self
Learn a list of feature name -> indices mappings and transform X.
Like fit(X) followed by transform(X), but does not require materializing X in memory.
Parameters
X [Mapping or iterable over Mappings] Dict(s) or Mapping(s) from feature names (arbitrary
Python objects) to feature values (strings or convertible to dtype).
y [(ignored)]
Returns
Xa [{array, sparse matrix}] Feature vectors; always 2-d.
get_feature_names()
Returns a list of feature names, ordered by their indices.
If one-of-K coding is applied to categorical features, this will include the constructed feature names but
not the original ones.
Parameters
Returns
inverse_transform(X, dict_type=<class ‘dict’>)
Transform array or sparse matrix X back to feature mappings.
X must have been produced by this DictVectorizer’s transform or fit_transform method; it may only have
passed through transformers that preserve the number of features and their order.
In the case of one-hot/one-of-K coding, the constructed feature names and values are returned rather than
the original ones.
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] Sample matrix.

dict_type [callable, optional] Constructor for feature mappings. Must conform to the col-
lections.Mapping API.
Returns
D [list of dict_type objects, length = n_samples] Feature mappings for the samples in X.
restrict(support, indices=False)
Restrict the features to those in support using feature selection.
This function modifies the estimator in-place.
Parameters
support [array-like] Boolean mask or list of indices (as returned by the get_support member
of feature selectors).
indices [boolean, optional] Whether support is a list of indices.
Returns
self
Examples

>>> from sklearn.feature_selection import SelectKBest, chi2
>>> v = DictVectorizer()
>>> D = [{'foo': 1, 'bar': 2}, {'foo': 3, 'baz': 1}]
>>> X = v.fit_transform(D)
>>> support = SelectKBest(chi2, k=2).fit(X, [0, 1])
>>> v.get_feature_names()
['bar', 'baz', 'foo']
>>> v.restrict(support.get_support())
DictVectorizer(dtype=..., separator='=', sort=True,
sparse=True)
>>> v.get_feature_names()
['bar', 'foo']
of a nested object.
Returns
self
transform(X)
Transform feature->value dicts to array or sparse matrix.
Named features not encountered during fit or fit_transform will be silently ignored.
Parameters
X [Mapping or iterable over Mappings, length = n_samples] Dict(s) or Mapping(s) from
feature names (arbitrary Python objects) to feature values (strings or convertible to dtype).
Returns
Xa [{array, sparse matrix}] Feature vectors; always 2-d.

Examples using sklearn.feature_extraction.DictVectorizer

6.14.2 sklearn.feature_extraction.FeatureHasher
class sklearn.feature_extraction.FeatureHasher(n_features=1048576, input_type=’dict’,

dtype=<class ‘numpy.float64’>, alter-
nate_sign=True, non_negative=False)
Implements feature hashing, aka the hashing trick.
This class turns sequences of symbolic feature names (strings) into scipy.sparse matrices, using a hash function
to compute the matrix column corresponding to a name. The hash function employed is the signed 32-bit version
of Murmurhash3.
Feature names of type byte string are used as-is. Unicode strings are converted to UTF-8 first, but no Unicode
normalization is done. Feature values must be (finite) numbers.
This class is a low-memory alternative to DictVectorizer and CountVectorizer, intended for large-scale (online)
learning and situations where memory is tight, e.g. when running prediction code on embedded devices.
Parameters
n_features [integer, optional] The number of features (columns) in the output matrices. Small
numbers of features are likely to cause hash collisions, but large numbers will cause larger
coefficient dimensions in linear learners.
input_type [string, optional, default “dict”] Either “dict” (the default) to accept dictionaries
over (feature_name, value); “pair” to accept pairs of (feature_name, value); or “string” to
accept single strings. feature_name should be a string, while value should be a number.
In the case of “string”, a value of 1 is implied. The feature_name is hashed to find the
appropriate column for the feature. The value’s sign might be flipped in the output (but see
non_negative, below).
dtype [numpy type, optional, default np.float64] The type of feature values. Passed to
scipy.sparse matrix constructors as the dtype argument. Do not set this to bool, np.boolean
or any unsigned integer type.
alternate_sign [boolean, optional, default True] When True, an alternating sign is added to the
features as to approximately conserve the inner product in the hashed space even for small
n_features. This approach is similar to sparse random projection.
non_negative [boolean, optional, default False] When True, an absolute value is applied to the
features matrix prior to returning it. When used in conjunction with alternate_sign=True,
this significantly reduces the inner product preservation property.
Deprecated since version 0.19: This option will be removed in 0.21.
See also:
DictVectorizer vectorizes string-valued features using a hash table.

sklearn.preprocessing.OneHotEncoder handles nominal/categorical features.

Examples
>>> from sklearn.feature_extraction import FeatureHasher

>>> h = FeatureHasher(n_features=10)
>>> D = [{'dog': 1, 'cat':2, 'elephant':4},{'dog': 2, 'run': 5}]
>>> f = h.transform(D)
>>> f.toarray()
array([[ 0., 0., -4., -1., 0., 0., 0., 0., 0., 2.],
[ 0., 0., 0., -2., -5., 0., 0., 0., 0., 0.]])
Methods
fit([X, y]) No-op.

transform(raw_X) Transform a sequence of instances to a scipy.sparse ma-
trix.
__init__(n_features=1048576, input_type=’dict’, dtype=<class ‘numpy.float64’>, alter-

nate_sign=True, non_negative=False)
fit(X=None, y=None)
No-op.
This method doesn’t do anything. It exists purely for compatibility with the scikit-learn transformer API.
Parameters
X [array-like]
Returns
self [FeatureHasher]
Parameters
Returns
Parameters
Returns


of a nested object.
Returns
self
transform(raw_X)
Transform a sequence of instances to a scipy.sparse matrix.
Parameters
raw_X [iterable over iterable over raw features, length = n_samples] Samples. Each sample
must be iterable an (e.g., a list or tuple) containing/generating feature names (and option-
ally values, see the input_type constructor argument) which will be hashed. raw_X need
not support the len function, so it can be the result of a generator; n_samples is determined
on the fly.
Returns
X [scipy.sparse matrix, shape = (n_samples, self.n_features)] Feature matrix, for use with
estimators or further transformers.
Examples using sklearn.feature_extraction.FeatureHasher
6.14.3 From images
The sklearn.feature_extraction.image submodule gathers utilities to extract features from images.
feature_extraction.image. Reshape a 2D image into a collection of patches

extract_patches_2d(. . . )
feature_extraction.image. Graph of the pixel-to-pixel connections
grid_to_graph(n_x, n_y)
feature_extraction.image. Graph of the pixel-to-pixel gradient connections
img_to_graph(img[, . . . ])
feature_extraction.image. Reconstruct the image from all of its patches.
reconstruct_from_patches_2d(. . . )
feature_extraction.image. Extracts patches from a collection of images
PatchExtractor([. . . ])
sklearn.feature_extraction.image.extract_patches_2d
sklearn.feature_extraction.image.extract_patches_2d(image, patch_size,
max_patches=None, ran-
dom_state=None)
Reshape a 2D image into a collection of patches
The resulting patches are allocated in a dedicated array.


Parameters
image [array, shape = (image_height, image_width) or] (image_height, image_width,
n_channels) The original image data. For color images, the last dimension specifies the
channel: a RGB image would have n_channels=3.
patch_size [tuple of ints (patch_height, patch_width)] the dimensions of one patch
max_patches [integer or float, optional default is None] The maximum number of patches to
extract. If max_patches is a float between 0 and 1, it is taken to be a proportion of the total
number of patches.
random_state [int, RandomState instance or None, optional (default=None)] Pseudo number
generator state used for random sampling to use if max_patches is not None. If int, ran-
dom_state is the seed used by the random number generator; If RandomState instance, ran-
dom_state is the random number generator; If None, the random number generator is the
RandomState instance used by np.random.
Returns
patches [array, shape = (n_patches, patch_height, patch_width) or] (n_patches, patch_height,
patch_width, n_channels) The collection of patches extracted from the image, where
n_patches is either max_patches or the total number of patches that can be extracted.
Examples
>>> from sklearn.feature_extraction import image

>>> one_image = np.arange(16).reshape((4, 4))
>>> one_image
array([[ 0, 1, 2, 3],
[ 4, 5, 6, 7],
[ 8, 9, 10, 11],
[12, 13, 14, 15]])
>>> patches = image.extract_patches_2d(one_image, (2, 2))
>>> print(patches.shape)
(9, 2, 2)
>>> patches[0]
array([[0, 1],
[4, 5]])
>>> patches[1]
array([[1, 2],
[5, 6]])
>>> patches[8]
array([[10, 11],
[14, 15]])
Examples using sklearn.feature_extraction.image.extract_patches_2d


sklearn.feature_extraction.image.grid_to_graph
sklearn.feature_extraction.image.grid_to_graph(n_x, n_y, n_z=1,

mask=None, return_as=<class
‘scipy.sparse.coo.coo_matrix’>,
dtype=<class ‘int’>)
Graph of the pixel-to-pixel connections
Edges exist if 2 voxels are connected.
Parameters
n_x [int] Dimension in x axis
n_y [int] Dimension in y axis
n_z [int, optional, default 1] Dimension in z axis
mask [ndarray of booleans, optional] An optional mask of the image, to consider only part of
the pixels.
return_as [np.ndarray or a sparse matrix class, optional] The class to use to build the returned
adjacency matrix.
dtype [dtype, optional, default int] The data of the returned sparse matrix. By default it is int
Notes
For scikit-learn versions 0.14.1 and prior, return_as=np.ndarray was handled by returning a dense np.matrix
instance. Going forward, np.ndarray returns an np.ndarray, as expected.
For compatibility, user code relying on this method should wrap its calls in np.asarray to avoid type issues.
sklearn.feature_extraction.image.img_to_graph
sklearn.feature_extraction.image.img_to_graph(img, mask=None, return_as=<class

‘scipy.sparse.coo.coo_matrix’>,
dtype=None)
Graph of the pixel-to-pixel gradient connections
Edges are weighted with the gradient values.
Parameters
img [ndarray, 2D or 3D] 2D or 3D image
mask [ndarray of booleans, optional] An optional mask of the image, to consider only part of
the pixels.
return_as [np.ndarray or a sparse matrix class, optional] The class to use to build the returned
adjacency matrix.
dtype [None or dtype, optional] The data of the returned sparse matrix. By default it is the
dtype of img

Notes
For scikit-learn versions 0.14.1 and prior, return_as=np.ndarray was handled by returning a dense np.matrix
instance. Going forward, np.ndarray returns an np.ndarray, as expected.
For compatibility, user code relying on this method should wrap its calls in np.asarray to avoid type issues.
sklearn.feature_extraction.image.reconstruct_from_patches_2d
sklearn.feature_extraction.image.reconstruct_from_patches_2d(patches, im-
age_size)
Reconstruct the image from all of its patches.
Patches are assumed to overlap and the image is constructed by filling in the patches from left to right, top to
bottom, averaging the overlapping regions.
Parameters
patches [array, shape = (n_patches, patch_height, patch_width) or] (n_patches, patch_height,
patch_width, n_channels) The complete set of patches. If the patches contain colour
information, channels are indexed along the last dimension: RGB patches would have
n_channels=3.
image_size [tuple of ints (image_height, image_width) or] (image_height, image_width,
n_channels) the size of the image that will be reconstructed
Returns
image [array, shape = image_size] the reconstructed image
Examples using sklearn.feature_extraction.image.reconstruct_from_patches_2d
sklearn.feature_extraction.image.PatchExtractor
class sklearn.feature_extraction.image.PatchExtractor(patch_size=None,
max_patches=None, ran-
dom_state=None)
Extracts patches from a collection of images
Parameters
patch_size [tuple of ints (patch_height, patch_width)] the dimensions of one patch
max_patches [integer or float, optional default is None] The maximum number of patches per
image to extract. If max_patches is a float in (0, 1), it is taken to mean a proportion of the
total number of patches.

Methods

transform(X) Transforms the image samples in X into a matrix of
patch data.
__init__(patch_size=None, max_patches=None, random_state=None)

fit(X, y=None)
Parameters
Parameters
Returns
of a nested object.
Returns
self
transform(X)
Transforms the image samples in X into a matrix of patch data.
Parameters
X [array, shape = (n_samples, image_height, image_width) or] (n_samples, image_height,
image_width, n_channels) Array of images from which to extract patches. For color im-
ages, the last dimension specifies the channel: a RGB image would have n_channels=3.
Returns
patches [array, shape = (n_patches, patch_height, patch_width) or] (n_patches,
patch_height, patch_width, n_channels) The collection of patches extracted from the im-
ages, where n_patches is either n_samples * max_patches or the total number of patches
that can be extracted.

6.14.4 From text
The sklearn.feature_extraction.text submodule gathers utilities to build feature vectors from text doc-
uments.
feature_extraction.text. Convert a collection of text documents to a matrix of token

CountVectorizer([. . . ]) counts
feature_extraction.text. Convert a collection of text documents to a matrix of token
HashingVectorizer([. . . ]) occurrences
feature_extraction.text. Transform a count matrix to a normalized tf or tf-idf repre-
TfidfTransformer([. . . ]) sentation
feature_extraction.text. Convert a collection of raw documents to a matrix of TF-
TfidfVectorizer([. . . ]) IDF features.
sklearn.feature_extraction.text.CountVectorizer
class sklearn.feature_extraction.text.CountVectorizer(input=’content’, encoding=’utf-

8’, decode_error=’strict’,
strip_accents=None, low-
ercase=True, preproces-
sor=None, tokenizer=None,
stop_words=None, to-
ken_pattern=’(?u)\b\w\w+\b’,
ngram_range=(1, 1), ana-
lyzer=’word’, max_df=1.0,
min_df=1, max_features=None,
vocabulary=None, binary=False,
dtype=<class ‘numpy.int64’>)
Convert a collection of text documents to a matrix of token counts
This implementation produces a sparse representation of the counts using scipy.sparse.csr_matrix.
If you do not provide an a-priori dictionary and you do not use an analyzer that does some kind of feature
selection then the number of features will be equal to the vocabulary size found by analyzing the data.
Parameters
input [string {‘filename’, ‘file’, ‘content’}] If ‘filename’, the sequence passed as an argument to
fit is expected to be a list of filenames that need reading to fetch the raw content to analyze.
If ‘file’, the sequence items must have a ‘read’ method (file-like object) that is called to fetch
the bytes in memory.
Otherwise the input is expected to be the sequence strings or bytes items are expected to be
analyzed directly.
encoding [string, ‘utf-8’ by default.] If bytes or files are given to analyze, this encoding is used
to decode.
decode_error [{‘strict’, ‘ignore’, ‘replace’}] Instruction on what to do if a byte sequence is
given to analyze that contains characters not of the given encoding. By default, it is ‘strict’,
meaning that a UnicodeDecodeError will be raised. Other values are ‘ignore’ and ‘replace’.
strip_accents [{‘ascii’, ‘unicode’, None}] Remove accents and perform other character nor-
malization during the preprocessing step. ‘ascii’ is a fast method that only works on char-

acters that have an direct ASCII mapping. ‘unicode’ is a slightly slower method that works
on any characters. None (default) does nothing.
Both ‘ascii’ and ‘unicode’ use NFKD normalization from unicodedata.normalize.
lowercase [boolean, True by default] Convert all characters to lowercase before tokenizing.
preprocessor [callable or None (default)] Override the preprocessing (string transformation)
stage while preserving the tokenizing and n-grams generation steps.
tokenizer [callable or None (default)] Override the string tokenization step while preserving the
preprocessing and n-grams generation steps. Only applies if analyzer == 'word'.
stop_words [string {‘english’}, list, or None (default)] If ‘english’, a built-in stop word list for
English is used. There are several known issues with ‘english’ and you should consider an
alternative (see Using stop words).
If a list, that list is assumed to contain stop words, all of which will be removed from the
resulting tokens. Only applies if analyzer == 'word'.
If None, no stop words will be used. max_df can be set to a value in the range [0.7, 1.0)
to automatically detect and filter stop words based on intra corpus document frequency of
terms.
token_pattern [string] Regular expression denoting what constitutes a “token”, only used if
analyzer == 'word'. The default regexp select tokens of 2 or more alphanumeric
characters (punctuation is completely ignored and always treated as a token separator).
ngram_range [tuple (min_n, max_n)] The lower and upper boundary of the range of n-values
for different n-grams to be extracted. All values of n such that min_n <= n <= max_n will
be used.
analyzer [string, {‘word’, ‘char’, ‘char_wb’} or callable] Whether the feature should be made
of word or character n-grams. Option ‘char_wb’ creates character n-grams only from text
inside word boundaries; n-grams at the edges of words are padded with space.
If a callable is passed it is used to extract the sequence of features out of the raw, unprocessed
input.
max_df [float in range [0.0, 1.0] or int, default=1.0] When building the vocabulary ignore terms
that have a document frequency strictly higher than the given threshold (corpus-specific stop
words). If float, the parameter represents a proportion of documents, integer absolute counts.
This parameter is ignored if vocabulary is not None.
min_df [float in range [0.0, 1.0] or int, default=1] When building the vocabulary ignore terms
that have a document frequency strictly lower than the given threshold. This value is also
called cut-off in the literature. If float, the parameter represents a proportion of documents,
integer absolute counts. This parameter is ignored if vocabulary is not None.
max_features [int or None, default=None] If not None, build a vocabulary that only consider
the top max_features ordered by term frequency across the corpus.
vocabulary [Mapping or iterable, optional] Either a Mapping (e.g., a dict) where keys are terms
and values are indices in the feature matrix, or an iterable over terms. If not given, a vocabu-
lary is determined from the input documents. Indices in the mapping should not be repeated
and should not have any gap between 0 and the largest index.
binary [boolean, default=False] If True, all non zero counts are set to 1. This is useful for
discrete probabilistic models that model binary events rather than integer counts.
dtype [type, optional] Type of the matrix returned by fit_transform() or transform().

Attributes
vocabulary_ [dict] A mapping of terms to feature indices.
stop_words_ [set] Terms that were ignored because they either:
• occurred in too many documents (max_df )
• occurred in too few documents (min_df )
• were cut off by feature selection (max_features).
This is only available if no vocabulary was given.
See also:
HashingVectorizer, TfidfVectorizer
Notes
The stop_words_ attribute can get large and increase the model size when pickling. This attribute is provided
only for introspection and can be safely removed using delattr or set to None before pickling.
Examples

>>> corpus = [
... 'This document is the second document.',
... 'And this is the third one.',
... ]
>>> vectorizer = CountVectorizer()
>>> print(vectorizer.get_feature_names())
['and', 'document', 'first', 'is', 'one', 'second', 'the', 'third', 'this']
>>> print(X.toarray())
[[0 1 1 1 0 0 1 0 1]
[0 2 0 1 0 1 1 0 1]
[1 0 0 1 1 0 1 1 1]
[0 1 1 1 0 0 1 0 1]]
Methods
build_analyzer() Return a callable that handles preprocessing and tok-

enization
build_preprocessor() Return a function to preprocess the text before tokeniza-
tion
build_tokenizer() Return a function that splits a string into a sequence of
tokens
decode(doc) Decode the input into a string of unicode symbols
fit(raw_documents[, y]) Learn a vocabulary dictionary of all tokens in the raw
documents.


fit_transform(raw_documents[, y]) Learn the vocabulary dictionary and return term-
document matrix.
get_feature_names() Array mapping from feature integer indices to feature
name
get_stop_words() Build or fetch the effective stop words list
inverse_transform(X) Return terms per document with nonzero entries in X.
transform(raw_documents) Transform documents to document-term matrix.
__init__(input=’content’, encoding=’utf-8’, decode_error=’strict’, strip_accents=None,

lowercase=True, preprocessor=None, tokenizer=None, stop_words=None, to-
ken_pattern=’(?u)\b\w\w+\b’, ngram_range=(1, 1), analyzer=’word’, max_df=1.0,
min_df=1, max_features=None, vocabulary=None, binary=False, dtype=<class
‘numpy.int64’>)
build_analyzer()
Return a callable that handles preprocessing and tokenization
build_preprocessor()
Return a function to preprocess the text before tokenization
build_tokenizer()
Return a function that splits a string into a sequence of tokens
decode(doc)
Decode the input into a string of unicode symbols
The decoding strategy depends on the vectorizer parameters.
Parameters
doc [string] The string to decode
fit(raw_documents, y=None)
Learn a vocabulary dictionary of all tokens in the raw documents.
Parameters
raw_documents [iterable] An iterable which yields either str, unicode or file objects.
Returns
self
fit_transform(raw_documents, y=None)
Learn the vocabulary dictionary and return term-document matrix.
This is equivalent to fit followed by transform, but more efficiently implemented.
Parameters
Returns
X [array, [n_samples, n_features]] Document-term matrix.
get_feature_names()
Array mapping from feature integer indices to feature name

Parameters
Returns
get_stop_words()
Build or fetch the effective stop words list
Return terms per document with nonzero entries in X.
Parameters
X [{array, sparse matrix}, shape = [n_samples, n_features]]
Returns
X_inv [list of arrays, len = n_samples] List of arrays of terms.
of a nested object.
Returns
self
transform(raw_documents)
Transform documents to document-term matrix.
Extract token counts out of raw text documents using the vocabulary fitted with fit or the one provided to
the constructor.
Parameters
Returns
X [sparse matrix, [n_samples, n_features]] Document-term matrix.
Examples using sklearn.feature_extraction.text.CountVectorizer

sklearn.feature_extraction.text.HashingVectorizer
class sklearn.feature_extraction.text.HashingVectorizer(input=’content’,
encoding=’utf-8’, de-
code_error=’strict’,
sor=None, tokenizer=None,
stop_words=None, to-
ken_pattern=’(?u)\b\w\w+\b’,
ngram_range=(1,
1), analyzer=’word’,
n_features=1048576, bi-
nary=False, norm=’l2’,
alternate_sign=True,
non_negative=False,
dtype=<class
‘numpy.float64’>)
Convert a collection of text documents to a matrix of token occurrences
It turns a collection of text documents into a scipy.sparse matrix holding token occurrence counts (or binary
occurrence information), possibly normalized as token frequencies if norm=’l1’ or projected on the euclidean
unit sphere if norm=’l2’.
This text vectorizer implementation uses the hashing trick to find the token string name to feature integer index
mapping.
This strategy has several advantages:
• it is very low memory scalable to large datasets as there is no need to store a vocabulary dictionary in
memory
• it is fast to pickle and un-pickle as it holds no state besides the constructor parameters
• it can be used in a streaming (partial fit) or parallel pipeline as there is no state computed during fit.
There are also a couple of cons (vs using a CountVectorizer with an in-memory vocabulary):
• there is no way to compute the inverse transform (from feature indices to string feature names) which can
be a problem when trying to introspect which features are most important to a model.
• there can be collisions: distinct tokens can be mapped to the same feature index. However in practice this
is rarely an issue if n_features is large enough (e.g. 2 ** 18 for text classification problems).
• no IDF weighting as this would render the transformer stateful.
The hash function employed is the signed 32-bit version of Murmurhash3.
Parameters
analyzed directly.
encoding [string, default=’utf-8’] If bytes or files are given to analyze, this encoding is used to
decode.

decode_error [{‘strict’, ‘ignore’, ‘replace’}] Instruction on what to do if a byte sequence is

given to analyze that contains characters not of the given encoding. By default, it is ‘strict’,
meaning that a UnicodeDecodeError will be raised. Other values are ‘ignore’ and ‘replace’.
strip_accents [{‘ascii’, ‘unicode’, None}] Remove accents and perform other character nor-
malization during the preprocessing step. ‘ascii’ is a fast method that only works on char-
acters that have an direct ASCII mapping. ‘unicode’ is a slightly slower method that works
on any characters. None (default) does nothing.
lowercase [boolean, default=True] Convert all characters to lowercase before tokenizing.
preprocessor [callable or None (default)] Override the preprocessing (string transformation)
stage while preserving the tokenizing and n-grams generation steps.
tokenizer [callable or None (default)] Override the string tokenization step while preserving the
preprocessing and n-grams generation steps. Only applies if analyzer == 'word'.
stop_words [string {‘english’}, list, or None (default)] If ‘english’, a built-in stop word list for
English is used. There are several known issues with ‘english’ and you should consider an
alternative (see Using stop words).
analyzer == 'word'. The default regexp selects tokens of 2 or more alphanumeric
ngram_range [tuple (min_n, max_n), default=(1, 1)] The lower and upper boundary of the
range of n-values for different n-grams to be extracted. All values of n such that min_n <=
n <= max_n will be used.
input.
n_features [integer, default=(2 ** 20)] The number of features (columns) in the output matri-
ces. Small numbers of features are likely to cause hash collisions, but large numbers will
cause larger coefficient dimensions in linear learners.
binary [boolean, default=False.] If True, all non zero counts are set to 1. This is useful for
discrete probabilistic models that model binary events rather than integer counts.
norm [‘l1’, ‘l2’ or None, optional] Norm used to normalize term vectors. None for no normal-
ization.
alternate_sign [boolean, optional, default True] When True, an alternating sign is added to the
features as to approximately conserve the inner product in the hashed space even for small
n_features. This approach is similar to sparse random projection.
non_negative [boolean, optional, default False] When True, an absolute value is applied to the
features matrix prior to returning it. When used in conjunction with alternate_sign=True,
this significantly reduces the inner product preservation property.
Deprecated since version 0.19: This option will be removed in 0.21.

dtype [type, optional] Type of the matrix returned by fit_transform() or transform().

See also:
CountVectorizer, TfidfVectorizer
Examples
>>> from sklearn.feature_extraction.text import HashingVectorizer

>>> corpus = [
... ]
>>> vectorizer = HashingVectorizer(n_features=2**4)
>>> print(X.shape)
(4, 16)
Methods

enization
tion
tokens
fit(X[, y]) Does nothing: this transformer is stateless.
fit_transform(X[, y]) Transform a sequence of documents to a document-term
matrix.
partial_fit(X[, y]) Does nothing: this transformer is stateless.
transform(X) Transform a sequence of documents to a document-term
matrix.
__init__(input=’content’, encoding=’utf-8’, decode_error=’strict’, strip_accents=None,

lowercase=True, preprocessor=None, tokenizer=None, stop_words=None, to-
ken_pattern=’(?u)\b\w\w+\b’, ngram_range=(1, 1), analyzer=’word’, n_features=1048576,
binary=False, norm=’l2’, alternate_sign=True, non_negative=False, dtype=<class
build_analyzer()
build_tokenizer()

decode(doc)
Parameters
fit(X, y=None)
Does nothing: this transformer is stateless.
Parameters
Transform a sequence of documents to a document-term matrix.
Parameters
X [iterable over raw text documents, length = n_samples] Samples. Each sample must be a
text document (either bytes or unicode strings, file name or file object depending on the
constructor argument) which will be tokenized and hashed.
y [any] Ignored. This parameter exists only for compatibility with sklearn.pipeline.Pipeline.
Returns
X [scipy.sparse matrix, shape = (n_samples, self.n_features)] Document-term matrix.
Parameters
Returns
get_stop_words()
Does nothing: this transformer is stateless.
This method is just there to mark the fact that this transformer can work in a streaming setup.
Parameters
of a nested object.
Returns
self

transform(X)
Transform a sequence of documents to a document-term matrix.
Parameters
X [iterable over raw text documents, length = n_samples] Samples. Each sample must be a
text document (either bytes or unicode strings, file name or file object depending on the
constructor argument) which will be tokenized and hashed.
Returns
X [scipy.sparse matrix, shape = (n_samples, self.n_features)] Document-term matrix.
Examples using sklearn.feature_extraction.text.HashingVectorizer

sklearn.feature_extraction.text.TfidfTransformer
class sklearn.feature_extraction.text.TfidfTransformer(norm=’l2’, use_idf=True,

smooth_idf=True, sublin-
ear_tf=False)
Transform a count matrix to a normalized tf or tf-idf representation
Tf means term-frequency while tf-idf means term-frequency times inverse document-frequency. This is a com-
mon term weighting scheme in information retrieval, that has also found good use in document classification.
The goal of using tf-idf instead of the raw frequencies of occurrence of a token in a given document is to scale
down the impact of tokens that occur very frequently in a given corpus and that are hence empirically less
informative than features that occur in a small fraction of the training corpus.
The formula that is used to compute the tf-idf for a term t of a document d in a document set is tf-idf(t, d) = tf(t,
d) * idf(t), and the idf is computed as idf(t) = log [ n / df(t) ] + 1 (if smooth_idf=False), where n is the total
number of documents in the document set and df(t) is the document frequency of t; the document frequency is
the number of documents in the document set that contain the term t. The effect of adding “1” to the idf in the
equation above is that terms with zero idf, i.e., terms that occur in all documents in a training set, will not be
entirely ignored. (Note that the idf formula above differs from the standard textbook notation that defines the
idf as idf(t) = log [ n / (df(t) + 1) ]).
If smooth_idf=True (the default), the constant “1” is added to the numerator and denominator of the idf
as if an extra document was seen containing every term in the collection exactly once, which prevents zero
divisions: idf(d, t) = log [ (1 + n) / (1 + df(d, t)) ] + 1.
Furthermore, the formulas used to compute tf and idf depend on parameter settings that correspond to the
SMART notation used in IR as follows:
Tf is “n” (natural) by default, “l” (logarithmic) when sublinear_tf=True. Idf is “t” when use_idf is given,
“n” (none) otherwise. Normalization is “c” (cosine) when norm='l2', “n” (none) when norm=None.
Parameters
norm [‘l1’, ‘l2’ or None, optional (default=’l2’)] Each output row will have unit norm, either:
* ‘l2’: Sum of squares of vector elements is 1. The cosine similarity between two vectors is

their dot product when l2 norm has been applied. * ‘l1’: Sum of absolute values of vector
elements is 1. See preprocessing.normalize
use_idf [boolean (default=True)] Enable inverse-document-frequency reweighting.
smooth_idf [boolean (default=True)] Smooth idf weights by adding one to document frequen-
cies, as if an extra document was seen containing every term in the collection exactly once.
Prevents zero divisions.
sublinear_tf [boolean (default=False)] Apply sublinear tf scaling, i.e. replace tf with 1 +
log(tf).
Attributes
idf_ [array, shape (n_features)] The inverse document frequency (IDF) vector; only defined if
use_idf is True.
References
[Yates2011], [MRS2008]
Methods
fit(X[, y]) Learn the idf vector (global term weights)

transform(X[, copy]) Transform a count matrix to a tf or tf-idf representation
__init__(norm=’l2’, use_idf=True, smooth_idf=True, sublinear_tf=False)

fit(X, y=None)
Learn the idf vector (global term weights)
Parameters
X [sparse matrix, [n_samples, n_features]] a matrix of term/token counts
Parameters
Returns
Parameters

Returns
of a nested object.
Returns
self
transform(X, copy=True)
Transform a count matrix to a tf or tf-idf representation
Parameters
X [sparse matrix, [n_samples, n_features]] a matrix of term/token counts
copy [boolean, default True] Whether to copy X and operate on the copy or perform in-place
operations.
Returns
vectors [sparse matrix, [n_samples, n_features]]
Examples using sklearn.feature_extraction.text.TfidfTransformer

sklearn.feature_extraction.text.TfidfVectorizer
class sklearn.feature_extraction.text.TfidfVectorizer(input=’content’, encoding=’utf-

8’, decode_error=’strict’,
sor=None, tokenizer=None, ana-
lyzer=’word’, stop_words=None,
token_pattern=’(?u)\b\w\w+\b’,
ngram_range=(1, 1),
max_df=1.0, min_df=1,
max_features=None, vocab-
ulary=None, binary=False,
dtype=<class ‘numpy.float64’>,
norm=’l2’, use_idf=True,
smooth_idf=True, sublin-
ear_tf=False)
Convert a collection of raw documents to a matrix of TF-IDF features.
Equivalent to CountVectorizer followed by TfidfTransformer.
Parameters

analyzed directly.
encoding [string, ‘utf-8’ by default.] If bytes or files are given to analyze, this encoding is used
to decode.
decode_error [{‘strict’, ‘ignore’, ‘replace’} (default=’strict’)] Instruction on what to do if a
byte sequence is given to analyze that contains characters not of the given encoding. By
default, it is ‘strict’, meaning that a UnicodeDecodeError will be raised. Other values are
‘ignore’ and ‘replace’.
strip_accents [{‘ascii’, ‘unicode’, None} (default=None)] Remove accents and perform other
character normalization during the preprocessing step. ‘ascii’ is a fast method that only
works on characters that have an direct ASCII mapping. ‘unicode’ is a slightly slower
method that works on any characters. None (default) does nothing.
lowercase [boolean (default=True)] Convert all characters to lowercase before tokenizing.
preprocessor [callable or None (default=None)] Override the preprocessing (string transforma-
tion) stage while preserving the tokenizing and n-grams generation steps.
tokenizer [callable or None (default=None)] Override the string tokenization step while pre-
serving the preprocessing and n-grams generation steps. Only applies if analyzer ==
'word'.
input.
stop_words [string {‘english’}, list, or None (default=None)] If a string, it is passed to
_check_stop_list and the appropriate stop list is returned. ‘english’ is currently the only
supported string value. There are several known issues with ‘english’ and you should con-
sider an alternative (see Using stop words).
If None, no stop words will be used. max_df can be set to a value in the range [0.7, 1.0)
to automatically detect and filter stop words based on intra corpus document frequency of
terms.
analyzer == 'word'. The default regexp selects tokens of 2 or more alphanumeric
ngram_range [tuple (min_n, max_n) (default=(1, 1))] The lower and upper boundary of the
range of n-values for different n-grams to be extracted. All values of n such that min_n <=
n <= max_n will be used.
max_df [float in range [0.0, 1.0] or int (default=1.0)] When building the vocabulary ignore
terms that have a document frequency strictly higher than the given threshold (corpus-

specific stop words). If float, the parameter represents a proportion of documents, integer
absolute counts. This parameter is ignored if vocabulary is not None.
min_df [float in range [0.0, 1.0] or int (default=1)] When building the vocabulary ignore terms
that have a document frequency strictly lower than the given threshold. This value is also
called cut-off in the literature. If float, the parameter represents a proportion of documents,
integer absolute counts. This parameter is ignored if vocabulary is not None.
max_features [int or None (default=None)] If not None, build a vocabulary that only consider
the top max_features ordered by term frequency across the corpus.
vocabulary [Mapping or iterable, optional (default=None)] Either a Mapping (e.g., a dict)
where keys are terms and values are indices in the feature matrix, or an iterable over terms.
If not given, a vocabulary is determined from the input documents.
binary [boolean (default=False)] If True, all non-zero term counts are set to 1. This does not
mean outputs will have only 0/1 values, only that the tf term in tf-idf is binary. (Set idf and
normalization to False to get 0/1 outputs.)
dtype [type, optional (default=float64)] Type of the matrix returned by fit_transform() or trans-
form().
norm [‘l1’, ‘l2’ or None, optional (default=’l2’)] Each output row will have unit norm, either:
* ‘l2’: Sum of squares of vector elements is 1. The cosine similarity between two vectors is
their dot product when l2 norm has been applied. * ‘l1’: Sum of absolute values of vector
elements is 1. See preprocessing.normalize
use_idf [boolean (default=True)] Enable inverse-document-frequency reweighting.
smooth_idf [boolean (default=True)] Smooth idf weights by adding one to document frequen-
cies, as if an extra document was seen containing every term in the collection exactly once.
Prevents zero divisions.
sublinear_tf [boolean (default=False)] Apply sublinear tf scaling, i.e. replace tf with 1 +
log(tf).
Attributes
vocabulary_ [dict] A mapping of terms to feature indices.
idf_ [array, shape (n_features)] The inverse document frequency (IDF) vector; only defined if
use_idf is True.
stop_words_ [set] Terms that were ignored because they either:
• occurred in too many documents (max_df )
• occurred in too few documents (min_df )
• were cut off by feature selection (max_features).
This is only available if no vocabulary was given.
See also:
CountVectorizer Transforms text into a sparse matrix of n-gram counts.

TfidfTransformer Performs the TF-IDF transformation from a provided matrix of counts.

Notes
The stop_words_ attribute can get large and increase the model size when pickling. This attribute is provided
only for introspection and can be safely removed using delattr or set to None before pickling.
Examples

>>> corpus = [
... ]
>>> print(vectorizer.get_feature_names())
['and', 'document', 'first', 'is', 'one', 'second', 'the', 'third', 'this']
>>> print(X.shape)
(4, 9)
Methods

enization
tion
tokens
fit(raw_documents[, y]) Learn vocabulary and idf from training set.
fit_transform(raw_documents[, y]) Learn vocabulary and idf, return term-document matrix.
get_feature_names() Array mapping from feature integer indices to feature
name
inverse_transform(X) Return terms per document with nonzero entries in X.
transform(raw_documents[, copy]) Transform documents to document-term matrix.
__init__(input=’content’, encoding=’utf-8’, decode_error=’strict’, strip_accents=None, lower-

case=True, preprocessor=None, tokenizer=None, analyzer=’word’, stop_words=None,
token_pattern=’(?u)\b\w\w+\b’, ngram_range=(1, 1), max_df=1.0, min_df=1,
max_features=None, vocabulary=None, binary=False, dtype=<class ‘numpy.float64’>,
norm=’l2’, use_idf=True, smooth_idf=True, sublinear_tf=False)
build_analyzer()

build_tokenizer()
decode(doc)
Parameters
fit(raw_documents, y=None)
Learn vocabulary and idf from training set.
Parameters
raw_documents [iterable] an iterable which yields either str, unicode or file objects
Returns
self [TfidfVectorizer]
fit_transform(raw_documents, y=None)
Learn vocabulary and idf, return term-document matrix.
This is equivalent to fit followed by transform, but more efficiently implemented.
Parameters
Returns
X [sparse matrix, [n_samples, n_features]] Tf-idf-weighted document-term matrix.
get_feature_names()
Array mapping from feature integer indices to feature name
Parameters
Returns
get_stop_words()
Return terms per document with nonzero entries in X.
Parameters
X [{array, sparse matrix}, shape = [n_samples, n_features]]
Returns
X_inv [list of arrays, len = n_samples] List of arrays of terms.

of a nested object.
Returns
self
transform(raw_documents, copy=True)
Transform documents to document-term matrix.
Uses the vocabulary and document frequencies (df) learned by fit (or fit_transform).
Parameters
copy [boolean, default True] Whether to copy X and operate on the copy or perform in-place
operations.
Returns
X [sparse matrix, [n_samples, n_features]] Tf-idf-weighted document-term matrix.
Examples using sklearn.feature_extraction.text.TfidfVectorizer
6.15 sklearn.feature_selection: Feature Selection
The sklearn.feature_selection module implements feature selection algorithms. It currently includes uni-
variate filter selection methods and the recursive feature elimination algorithm.
User guide: See the Feature selection section for further details.
feature_selection. Univariate feature selector with configurable strategy.

GenericUnivariateSelect([. . . ])
feature_selection.SelectPercentile([. . . ]) Select features according to a percentile of the highest
scores.
feature_selection.SelectKBest([score_func, Select features according to the k highest scores.
k])
feature_selection.SelectFpr([score_func, al- Filter: Select the pvalues below alpha based on a FPR test.
pha])
feature_selection.SelectFdr([score_func, al- Filter: Select the p-values for an estimated false discovery
pha]) rate
feature_selection. Meta-transformer for selecting features based on impor-
SelectFromModel(estimator) tance weights.
feature_selection.SelectFwe([score_func, al- Filter: Select the p-values corresponding to Family-wise
pha]) error rate
6.15. sklearn.feature_selection: Feature Selection 1673


feature_selection.RFE(estimator[, . . . ]) Feature ranking with recursive feature elimination.
feature_selection.RFECV (estimator[, step, . . . ]) Feature ranking with recursive feature elimination and
cross-validated selection of the best number of features.
feature_selection. Feature selector that removes all low-variance features.
VarianceThreshold([threshold])
6.15.1 sklearn.feature_selection.GenericUnivariateSelect
class sklearn.feature_selection.GenericUnivariateSelect(score_func=<function
f_classif>, mode=’percentile’,
param=1e-05)
Univariate feature selector with configurable strategy.
Parameters
score_func [callable] Function taking two arrays X and y, and returning a pair of arrays (scores,
pvalues). For modes ‘percentile’ or ‘kbest’ it can return a single array scores.
mode [{‘percentile’, ‘k_best’, ‘fpr’, ‘fdr’, ‘fwe’}] Feature selection mode.
param [float or int depending on the feature selection mode] Parameter of the corresponding
mode.
Attributes
scores_ [array-like, shape=(n_features,)] Scores of features.
pvalues_ [array-like, shape=(n_features,)] p-values of feature scores, None if score_func re-
turned scores only.
See also:
f_classif ANOVA F-value between label/feature for classification tasks.

mutual_info_classif Mutual information for a discrete target.
chi2 Chi-squared stats of non-negative features for classification tasks.
f_regression F-value between label/feature for regression tasks.
mutual_info_regression Mutual information for a continuous target.
SelectPercentile Select features based on percentile of the highest scores.
SelectKBest Select features based on the k highest scores.
SelectFpr Select features based on a false positive rate test.
SelectFdr Select features based on an estimated false discovery rate.
SelectFwe Select features based on family-wise error rate.
Examples

>>> from sklearn.feature_selection import GenericUnivariateSelect, chi2
>>> X.shape

(569, 30)
>>> transformer = GenericUnivariateSelect(chi2, 'k_best', param=20)
>>> X_new = transformer.fit_transform(X, y)
>>> X_new.shape
(569, 20)
Methods
fit(X, y) Run score function on (X, y) and get the appropriate

features.
get_support([indices]) Get a mask, or integer index, of the features selected
inverse_transform(X) Reverse the transformation operation
transform(X) Reduce X to the selected features.
__init__(score_func=<function f_classif>, mode=’percentile’, param=1e-05)

fit(X, y)
Run score function on (X, y) and get the appropriate features.
Parameters
X [array-like, shape = [n_samples, n_features]] The training input samples.
Returns
self [object]
Parameters
Returns
Parameters
Returns

get_support(indices=False)
Get a mask, or integer index, of the features selected
Parameters
indices [boolean (default False)] If True, the return value will be an array of integers, rather
than a boolean mask.
Returns
support [array] An index that selects the retained features from a feature vector. If indices
is False, this is a boolean array of shape [# input features], in which an element is True
iff its corresponding feature is selected for retention. If indices is True, this is an integer
array of shape [# output features] whose values are indices into the input feature vector.
Reverse the transformation operation
Parameters
X [array of shape [n_samples, n_selected_features]] The input samples.
Returns
X_r [array of shape [n_samples, n_original_features]] X with columns of zeros inserted
where features would have been removed by transform.
of a nested object.
Returns
self
transform(X)
Reduce X to the selected features.
Parameters
Returns
X_r [array of shape [n_samples, n_selected_features]] The input samples with only the se-
lected features.
6.15.2 sklearn.feature_selection.SelectPercentile
class sklearn.feature_selection.SelectPercentile(score_func=<function f_classif>, per-

centile=10)
Select features according to a percentile of the highest scores.
Parameters
pvalues) or a single array with scores. Default is f_classif (see below “See also”). The
default function only works with classification tasks.

percentile [int, optional, default=10] Percent of features to keep.

Attributes
turned only scores.
See also:

GenericUnivariateSelect Univariate feature selector with configurable mode.
Notes
Ties between features with equal scores will be broken in an unspecified way.
Examples

>>> from sklearn.feature_selection import SelectPercentile, chi2
>>> X, y = load_digits(return_X_y=True)
>>> X.shape
(1797, 64)
>>> X_new = SelectPercentile(chi2, percentile=10).fit_transform(X, y)
>>> X_new.shape
(1797, 7)
Methods

features.

__init__(score_func=<function f_classif>, percentile=10)

fit(X, y)
Parameters
Returns
self [object]
Parameters
Returns
Parameters
Returns
Parameters
Returns
Parameters
Returns


of a nested object.
Returns
self
transform(X)
Parameters
Returns
lected features.
Examples using sklearn.feature_selection.SelectPercentile

6.15.3 sklearn.feature_selection.SelectKBest
class sklearn.feature_selection.SelectKBest(score_func=<function f_classif>, k=10)

Select features according to the k highest scores.
Parameters
pvalues) or a single array with scores. Default is f_classif (see below “See also”). The
default function only works with classification tasks.
k [int or “all”, optional, default=10] Number of top features to select. The “all” option bypasses
selection, for use in a parameter search.
Attributes
turned only scores.
See also:



Notes
Ties between features with equal scores will be broken in an unspecified way.
Examples

>>> from sklearn.feature_selection import SelectKBest, chi2
>>> X.shape
(1797, 64)
>>> X_new = SelectKBest(chi2, k=20).fit_transform(X, y)
>>> X_new.shape
(1797, 20)
Methods

features.
__init__(score_func=<function f_classif>, k=10)

fit(X, y)
Parameters
Returns

self [object]
Parameters
Returns
Parameters
Returns
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self

transform(X)
Parameters
Returns
lected features.
Examples using sklearn.feature_selection.SelectKBest

6.15.4 sklearn.feature_selection.SelectFpr
class sklearn.feature_selection.SelectFpr(score_func=<function f_classif>, alpha=0.05)

Filter: Select the pvalues below alpha based on a FPR test.
FPR test stands for False Positive Rate test. It controls the total amount of false detections.
Parameters
pvalues). Default is f_classif (see below “See also”). The default function only works with
classification tasks.
alpha [float, optional] The highest p-value for features to be kept.
Attributes
pvalues_ [array-like, shape=(n_features,)] p-values of feature scores.
See also:

mutual_info_classif
mutual_info_regression Mutual information between features and the target.

Examples

>>> from sklearn.feature_selection import SelectFpr, chi2
>>> X.shape
(569, 30)
>>> X_new = SelectFpr(chi2, alpha=0.01).fit_transform(X, y)
>>> X_new.shape
(569, 16)
Methods

features.
__init__(score_func=<function f_classif>, alpha=0.05)

fit(X, y)
Parameters
Returns
self [object]
Parameters
Returns

Parameters
Returns
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
transform(X)
Parameters
Returns
lected features.
6.15.5 sklearn.feature_selection.SelectFdr
class sklearn.feature_selection.SelectFdr(score_func=<function f_classif>, alpha=0.05)

Filter: Select the p-values for an estimated false discovery rate
This uses the Benjamini-Hochberg procedure. alpha is an upper bound on the expected false discovery rate.


Parameters
alpha [float, optional] The highest uncorrected p-value for features to keep.
Attributes
See also:

mutual_info_regression Mutual information for a contnuous target.
References
https://en.wikipedia.org/wiki/False_discovery_rate
Examples

>>> from sklearn.feature_selection import SelectFdr, chi2
>>> X.shape
(569, 30)
>>> X_new = SelectFdr(chi2, alpha=0.01).fit_transform(X, y)
>>> X_new.shape
(569, 16)
Methods

features.



fit(X, y)
Parameters
Returns
self [object]
Parameters
Returns
Parameters
Returns
Parameters
Returns

Parameters
Returns
of a nested object.
Returns
self
transform(X)
Parameters
Returns
lected features.
6.15.6 sklearn.feature_selection.SelectFromModel
class sklearn.feature_selection.SelectFromModel(estimator, threshold=None, prefit=False,

norm_order=1, max_features=None)
Meta-transformer for selecting features based on importance weights.
Parameters
estimator [object] The base estimator from which the transformer is built. This can be both a
fitted (if prefit is set to True) or a non-fitted estimator. The estimator must have either a
feature_importances_ or coef_ attribute after fitting.
threshold [string, float, optional default None] The threshold value to use for feature selection.
Features whose importance is greater or equal are kept while the others are discarded. If
“median” (resp. “mean”), then the threshold value is the median (resp. the mean) of the
feature importances. A scaling factor (e.g., “1.25*mean”) may also be used. If None and
if the estimator has a parameter penalty set to l1, either explicitly or implicitly (e.g, Lasso),
the threshold used is 1e-5. Otherwise, “mean” is used by default.
prefit [bool, default False] Whether a prefit model is expected to be passed into the constructor
directly or not. If True, transform must be called directly and SelectFromModel can-
not be used with cross_val_score, GridSearchCV and similar utilities that clone
the estimator. Otherwise train the model using fit and then transform to do feature
selection.

norm_order [non-zero int, inf, -inf, default 1] Order of the norm used to filter the vectors of
coefficients below threshold in the case where the coef_ attribute of the estimator is
of dimension 2.
max_features [int or None, optional] The maximum number of features selected scoring above
threshold. To disable threshold and only select based on max_features, set
threshold=-np.inf.
Attributes
estimator_ [an estimator] The base estimator from which the transformer is built. This is stored
only when a non-fitted estimator is passed to the SelectFromModel, i.e when prefit is
False.
threshold_ [float] The threshold value used for feature selection.
Methods
fit(X[, y]) Fit the SelectFromModel meta-transformer.

partial_fit(X[, y]) Fit the SelectFromModel meta-transformer only once.
__init__(estimator, threshold=None, prefit=False, norm_order=1, max_features=None)

fit(X, y=None, **fit_params)
Fit the SelectFromModel meta-transformer.
Parameters
X [array-like of shape (n_samples, n_features)] The training input samples.
y [array-like, shape (n_samples,)] The target values (integers that correspond to classes in
classification, real numbers in regression).
**fit_params [Other estimator specific parameters]
Returns
self [object]
Parameters
Returns

Parameters
Returns
Parameters
Returns
Parameters
Returns
partial_fit(X, y=None, **fit_params)
Fit the SelectFromModel meta-transformer only once.
Parameters
X [array-like of shape (n_samples, n_features)] The training input samples.
y [array-like, shape (n_samples,)] The target values (integers that correspond to classes in
classification, real numbers in regression).
**fit_params [Other estimator specific parameters]
Returns
self [object]
of a nested object.
Returns
self

transform(X)
Parameters
Returns
lected features.
Examples using sklearn.feature_selection.SelectFromModel

6.15.7 sklearn.feature_selection.SelectFwe
class sklearn.feature_selection.SelectFwe(score_func=<function f_classif>, alpha=0.05)

Filter: Select the p-values corresponding to Family-wise error rate
Parameters
alpha [float, optional] The highest uncorrected p-value for features to keep.
Attributes
See also:

Examples


>>> from sklearn.feature_selection import SelectFwe, chi2
>>> X.shape
(569, 30)
>>> X_new = SelectFwe(chi2, alpha=0.01).fit_transform(X, y)
>>> X_new.shape
(569, 15)
Methods

features.

fit(X, y)
Parameters
Returns
self [object]
Parameters
Returns
Parameters
Returns


Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
transform(X)
Parameters
Returns
lected features.
6.15.8 sklearn.feature_selection.RFE
class sklearn.feature_selection.RFE(estimator, n_features_to_select=None, step=1, verbose=0)

Feature ranking with recursive feature elimination.
Given an external estimator that assigns weights to features (e.g., the coefficients of a linear model), the goal of
recursive feature elimination (RFE) is to select features by recursively considering smaller and smaller sets of
features. First, the estimator is trained on the initial set of features and the importance of each feature is obtained
either through a coef_ attribute or through a feature_importances_ attribute. Then, the least important
features are pruned from current set of features. That procedure is recursively repeated on the pruned set until
the desired number of features to select is eventually reached.


Parameters
estimator [object] A supervised learning estimator with a fit method that provides in-
formation about feature importance either through a coef_ attribute or through a
feature_importances_ attribute.
n_features_to_select [int or None (default=None)] The number of features to select. If None,
half of the features are selected.
step [int or float, optional (default=1)] If greater than or equal to 1, then step corresponds to
the (integer) number of features to remove at each iteration. If within (0.0, 1.0), then step
corresponds to the percentage (rounded down) of features to remove at each iteration.
verbose [int, (default=0)] Controls verbosity of output.
Attributes
n_features_ [int] The number of selected features.
support_ [array of shape [n_features]] The mask of selected features.
ranking_ [array of shape [n_features]] The feature ranking, such that ranking_[i] corre-
sponds to the ranking position of the i-th feature. Selected (i.e., estimated best) features are
assigned rank 1.
estimator_ [object] The external estimator fit on the reduced dataset.
See also:
RFECV Recursive feature elimination with built-in cross-validated selection of the best number of features
References
[1]
Examples
The following example shows how to retrieve the 5 right informative features in the Friedman #1 dataset.

>>> from sklearn.feature_selection import RFE
>>> from sklearn.svm import SVR
>>> X, y = make_friedman1(n_samples=50, n_features=10, random_state=0)
>>> estimator = SVR(kernel="linear")
>>> selector = RFE(estimator, 5, step=1)
>>> selector = selector.fit(X, y)
>>> selector.support_
array([ True, True, True, True, True, False, False, False, False,
False])
>>> selector.ranking_
array([1, 1, 1, 1, 1, 6, 4, 3, 2, 5])
Methods


fit(X, y) Fit the RFE model and then the underlying estimator on
the selected features.
predict(X) Reduce X to the selected features and then predict using
the underlying estimator.
score(X, y) Reduce X to the selected features and then return the
score of the underlying estimator.
__init__(estimator, n_features_to_select=None, step=1, verbose=0)

Parameters
Returns
score [array, shape = [n_samples, n_classes] or [n_samples]] The decision function of the
fit(X, y)
Fit the RFE model and then the underlying estimator on the selected features.
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] The training input samples.
y [array-like, shape = [n_samples]] The target values.

Parameters
Returns

Parameters
Returns
Parameters
Returns
Parameters
Returns
predict(X)
Reduce X to the selected features and then predict using the underlying estimator.
Parameters
Returns
y [array of shape [n_samples]] The predicted target values.
Parameters
Returns
predict_proba(X)
Parameters

Returns
score(X, y)
Reduce X to the selected features and then return the score of the underlying estimator.
Parameters
y [array of shape [n_samples]] The target values.
of a nested object.
Returns
self
transform(X)
Parameters
Returns
lected features.
Examples using sklearn.feature_selection.RFE
6.15.9 sklearn.feature_selection.RFECV
class sklearn.feature_selection.RFECV(estimator, step=1, min_features_to_select=1, cv=’warn’,

scoring=None, verbose=0, n_jobs=None)
Feature ranking with recursive feature elimination and cross-validated selection of the best number of features.
Parameters
estimator [object] A supervised learning estimator with a fit method that provides in-
formation about feature importance either through a coef_ attribute or through a
feature_importances_ attribute.

step [int or float, optional (default=1)] If greater than or equal to 1, then step corresponds
to the (integer) number of features to remove at each iteration. If within (0.0, 1.0), then
step corresponds to the percentage (rounded down) of features to remove at each itera-
tion. Note that the last iteration may remove fewer than step features in order to reach
min_features_to_select.
min_features_to_select [int, (default=1)] The minimum number of features to be selected.
This number of features will always be scored, even if the difference between the origi-
nal feature count and min_features_to_select isn’t divisible by step.
• CV splitter,
StratifiedKFold is used. If the estimator is a classifier or if y is neither binary nor
multiclass, sklearn.model_selection.KFold is used.
Changed in version 0.20: cv default value of None will change from 3-fold to 5-fold in
v0.22.
scoring [string, callable or None, optional, (default=None)] A string (see model evaluation doc-
X, y).
verbose [int, (default=0)] Controls verbosity of output.
n_jobs [int or None, optional (default=None)] Number of cores to run in parallel while fitting
across folds. None means 1 unless in a joblib.parallel_backend context. -1
Attributes
n_features_ [int] The number of selected features with cross-validation.
support_ [array of shape [n_features]] The mask of selected features.
ranking_ [array of shape [n_features]] The feature ranking, such that ranking_[i] corresponds
to the ranking position of the i-th feature. Selected (i.e., estimated best) features are assigned
rank 1.
grid_scores_ [array of shape [n_subsets_of_features]] The cross-validation scores such that
grid_scores_[i] corresponds to the CV score of the i-th subset of features.
estimator_ [object] The external estimator fit on the reduced dataset.
See also:
RFE Recursive feature elimination

Notes
The size of grid_scores_ is equal to ceil((n_features - min_features_to_select) /

step) + 1, where step is the number of features removed at each iteration.
References
[1]
Examples
The following example shows how to retrieve the a-priori not known 5 informative features in the Friedman #1
dataset.

>>> from sklearn.feature_selection import RFECV
>>> X, y = make_friedman1(n_samples=50, n_features=10, random_state=0)
>>> estimator = SVR(kernel="linear")
>>> selector = RFECV(estimator, step=1, cv=5)
>>> selector = selector.fit(X, y)
>>> selector.support_
array([ True, True, True, True, True, False, False, False, False,
False])
>>> selector.ranking_
array([1, 1, 1, 1, 1, 6, 4, 3, 2, 5])
Methods

fit(X, y[, groups]) Fit the RFE model and automatically tune the number
of selected features.
predict(X) Reduce X to the selected features and then predict using
the underlying estimator.
score(X, y) Reduce X to the selected features and then return the
score of the underlying estimator.
__init__(estimator, step=1, min_features_to_select=1, cv=’warn’, scoring=None, verbose=0,

n_jobs=None)

Parameters
Returns
score [array, shape = [n_samples, n_classes] or [n_samples]] The decision function of the
fit(X, y, groups=None)
Fit the RFE model and automatically tune the number of selected features.
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] Training vector, where
n_samples is the number of samples and n_features is the total number of features.
y [array-like, shape = [n_samples]] Target values (integers for classification, real numbers
for regression).
groups [array-like, shape = [n_samples], optional] Group labels for the samples used while
splitting the dataset into train/test set.

Parameters
Returns
Parameters
Returns
Parameters
Returns

Parameters
Returns
predict(X)
Reduce X to the selected features and then predict using the underlying estimator.
Parameters
Returns
y [array of shape [n_samples]] The predicted target values.
Parameters
Returns
predict_proba(X)
Parameters
Returns
score(X, y)
Reduce X to the selected features and then return the score of the underlying estimator.
Parameters
y [array of shape [n_samples]] The target values.

of a nested object.
Returns
self
transform(X)
Parameters
Returns
lected features.
Examples using sklearn.feature_selection.RFECV
6.15.10 sklearn.feature_selection.VarianceThreshold
class sklearn.feature_selection.VarianceThreshold(threshold=0.0)
Feature selector that removes all low-variance features.
This feature selection algorithm looks only at the features (X), not the desired outputs (y), and can thus be used
for unsupervised learning.
Parameters
threshold [float, optional] Features with a training-set variance lower than this threshold will be
removed. The default is to keep all features with non-zero variance, i.e. remove the features
that have the same value in all samples.
Attributes
variances_ [array, shape (n_features,)] Variances of individual features.
Examples
The following dataset has integer features, two of which are the same in every sample. These are removed with
the default setting for threshold:
>>> X = [[0, 2, 0, 3], [0, 1, 4, 3], [0, 1, 1, 3]]

>>> selector = VarianceThreshold()
>>> selector.fit_transform(X)
array([[2, 0],
[1, 4],
[1, 1]])

Methods
fit(X[, y]) Learn empirical variances from X.

__init__(threshold=0.0)
fit(X, y=None)
Learn empirical variances from X.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Sample vectors from which
to compute variances.
y [any] Ignored. This parameter exists only for compatibility with sklearn.pipeline.Pipeline.
Returns
self
Parameters
Returns
Parameters
Returns
Parameters
Returns

Parameters
Returns
of a nested object.
Returns
self
transform(X)
Parameters
Returns
lected features.
feature_selection.chi2(X, y) Compute chi-squared stats between each non-negative fea-

ture and class.
feature_selection.f_classif(X, y) Compute the ANOVA F-value for the provided sample.
feature_selection.f_regression(X, y[, cen- Univariate linear regression tests.
ter])
feature_selection. Estimate mutual information for a discrete target variable.
mutual_info_classif(X, y)
feature_selection. Estimate mutual information for a continuous target vari-
mutual_info_regression(X, y) able.
6.15.11 sklearn.feature_selection.chi2
sklearn.feature_selection.chi2(X, y)
Compute chi-squared stats between each non-negative feature and class.
This score can be used to select the n_features features with the highest values for the test chi-squared statistic
from X, which must contain only non-negative features such as booleans or frequencies (e.g., term counts in
document classification), relative to the classes.
Recall that the chi-square test measures dependence between stochastic variables, so using this function “weeds

out” the features that are the most likely to be independent of class and therefore irrelevant for classification.
Parameters
X [{array-like, sparse matrix}, shape = (n_samples, n_features_in)] Sample vectors.
y [array-like, shape = (n_samples,)] Target vector (class labels).
Returns
chi2 [array, shape = (n_features,)] chi2 statistics of each feature.
pval [array, shape = (n_features,)] p-values of each feature.
See also:

Notes
Complexity of this algorithm is O(n_classes * n_features).
Examples using sklearn.feature_selection.chi2

6.15.12 sklearn.feature_selection.f_classif
sklearn.feature_selection.f_classif(X, y)
Compute the ANOVA F-value for the provided sample.
Parameters
X [{array-like, sparse matrix} shape = [n_samples, n_features]] The set of regressors that will
be tested sequentially.
y [array of shape(n_samples)] The data matrix.
Returns
F [array, shape = [n_features,]] The set of F values.
pval [array, shape = [n_features,]] The set of p-values.
See also:


Examples using sklearn.feature_selection.f_classif
6.15.13 sklearn.feature_selection.f_regression
sklearn.feature_selection.f_regression(X, y, center=True)
Univariate linear regression tests.
Linear model for testing the individual effect of each of many regressors. This is a scoring function to be used
in a feature selection procedure, not a free standing feature selection procedure.
This is done in 2 steps:
1. The correlation between each regressor and the target is computed, that is, ((X[:, i] - mean(X[:, i])) * (y -
mean_y)) / (std(X[:, i]) * std(y)).
2. It is converted to an F score then to a p-value.
For more on usage see the User Guide.
Parameters
X [{array-like, sparse matrix} shape = (n_samples, n_features)] The set of regressors that will
be tested sequentially.
y [array of shape(n_samples).] The data matrix
center [True, bool,] If true, X and y will be centered.
Returns
F [array, shape=(n_features,)] F values of features.
pval [array, shape=(n_features,)] p-values of F-scores.
See also:

Examples using sklearn.feature_selection.f_regression


6.15.14 sklearn.feature_selection.mutual_info_classif
sklearn.feature_selection.mutual_info_classif(X, y, discrete_features=’auto’,
n_neighbors=3, copy=True, ran-
dom_state=None)
Estimate mutual information for a discrete target variable.
Mutual information (MI) [1] between two random variables is a non-negative value, which measures the depen-
dency between the variables. It is equal to zero if and only if two random variables are independent, and higher
values mean higher dependency.
The function relies on nonparametric methods based on entropy estimation from k-nearest neighbors distances
as described in [2] and [3]. Both methods are based on the idea originally proposed in [4].
It can be used for univariate features selection, read more in the User Guide.
Parameters
X [array_like or sparse matrix, shape (n_samples, n_features)] Feature matrix.
y [array_like, shape (n_samples,)] Target vector.
discrete_features [{‘auto’, bool, array_like}, default ‘auto’] If bool, then determines whether
to consider all features discrete or continuous. If array, then it should be either a boolean
mask with shape (n_features,) or array with indices of discrete features. If ‘auto’, it is
assigned to False for dense X and to True for sparse X.
n_neighbors [int, default 3] Number of neighbors to use for MI estimation for continuous vari-
ables, see [2] and [3]. Higher values reduce variance of the estimation, but could introduce
a bias.
copy [bool, default True] Whether to make a copy of the given data. If set to False, the initial
data will be overwritten.
pseudo random number generator for adding small noise to continuous variables in order
to remove repeated values. If int, random_state is the seed used by the random number
generator; If RandomState instance, random_state is the random number generator; If None,
the random number generator is the RandomState instance used by np.random.
Returns
mi [ndarray, shape (n_features,)] Estimated mutual information between each feature and the
target.
Notes
1. The term “discrete features” is used instead of naming them “categorical”, because it describes the essence
more accurately. For example, pixel intensities of an image are discrete features (but hardly categorical)
and you will get better results if mark them as such. Also note, that treating a continuous variable as
discrete and vice versa will usually give incorrect results, so be attentive about that.
2. True mutual information can’t be negative. If its estimate turns out to be negative, it is replaced by zero.
References
[1], [2], [3], [4]

6.15.15 sklearn.feature_selection.mutual_info_regression
sklearn.feature_selection.mutual_info_regression(X, y, discrete_features=’auto’,
n_neighbors=3, copy=True, ran-
dom_state=None)
Estimate mutual information for a continuous target variable.
Mutual information (MI) [1] between two random variables is a non-negative value, which measures the depen-
dency between the variables. It is equal to zero if and only if two random variables are independent, and higher
values mean higher dependency.
The function relies on nonparametric methods based on entropy estimation from k-nearest neighbors distances
as described in [2] and [3]. Both methods are based on the idea originally proposed in [4].
It can be used for univariate features selection, read more in the User Guide.
Parameters
X [array_like or sparse matrix, shape (n_samples, n_features)] Feature matrix.
y [array_like, shape (n_samples,)] Target vector.
discrete_features [{‘auto’, bool, array_like}, default ‘auto’] If bool, then determines whether
to consider all features discrete or continuous. If array, then it should be either a boolean
mask with shape (n_features,) or array with indices of discrete features. If ‘auto’, it is
assigned to False for dense X and to True for sparse X.
n_neighbors [int, default 3] Number of neighbors to use for MI estimation for continuous vari-
ables, see [2] and [3]. Higher values reduce variance of the estimation, but could introduce
a bias.
copy [bool, default True] Whether to make a copy of the given data. If set to False, the initial
data will be overwritten.
pseudo random number generator for adding small noise to continuous variables in order
to remove repeated values. If int, random_state is the seed used by the random number
Returns
mi [ndarray, shape (n_features,)] Estimated mutual information between each feature and the
target.
Notes
1. The term “discrete features” is used instead of naming them “categorical”, because it describes the essence
more accurately. For example, pixel intensities of an image are discrete features (but hardly categorical)
and you will get better results if mark them as such. Also note, that treating a continuous variable as
discrete and vice versa will usually give incorrect results, so be attentive about that.
2. True mutual information can’t be negative. If its estimate turns out to be negative, it is replaced by zero.
References
[1], [2], [3], [4]

Examples using sklearn.feature_selection.mutual_info_regression
6.16 sklearn.gaussian_process: Gaussian Processes
The sklearn.gaussian_process module implements Gaussian Process based regression and classification.
User guide: See the Gaussian Processes section for further details.
Gaussian
gaussian_process.GaussianProcessClassifier([. . . ]) process classification (GPC) based on Laplace
approximation.
Gaussian
gaussian_process.GaussianProcessRegressor([. . . ]) process regression (GPR).
6.16.1 sklearn.gaussian_process.GaussianProcessClassifier
class sklearn.gaussian_process.GaussianProcessClassifier(kernel=None, opti-

mizer=’fmin_l_bfgs_b’,
n_restarts_optimizer=0,
max_iter_predict=100,
warm_start=False,
copy_X_train=True,
random_state=None,
multi_class=’one_vs_rest’,
n_jobs=None)
Gaussian process classification (GPC) based on Laplace approximation.
The implementation is based on Algorithm 3.1, 3.2, and 5.1 of Gaussian Processes for Machine Learning
(GPML) by Rasmussen and Williams.
Internally, the Laplace approximation is used for approximating the non-Gaussian posterior by a Gaussian.
Currently, the implementation is restricted to using the logistic link function. For multi-class classification,
several binary one-versus rest classifiers are fitted. Note that this class thus does not implement a true multi-
class Laplace approximation.
Parameters
kernel [kernel object] The kernel specifying the covariance function of the GP. If None is
passed, the kernel “1.0 * RBF(1.0)” is used as default. Note that the kernel’s hyperpa-
rameters are optimized during fitting.
optimizer [string or callable, optional (default: “fmin_l_bfgs_b”)] Can either be one of the
internally supported optimizers for optimizing the kernel’s parameters, specified by a string,
or an externally defined optimizer passed as a callable. If a callable is passed, it must have
the signature:
def optimizer(obj_func, initial_theta, bounds):

# * 'obj_func' is the objective function to be maximized, which
# takes the hyperparameters theta as parameter and an
# optional flag eval_gradient, which determines if the
# gradient is returned additionally to the function value
# * 'initial_theta': the initial value for theta, which can be
# used by local optimizers
# * 'bounds': the bounds on the values of theta

....
# Returned are the best found hyperparameters theta and
# the corresponding value of the target function.
return theta_opt, func_min
Per default, the ‘fmin_l_bfgs_b’ algorithm from scipy.optimize is used. If None is passed,
the kernel’s parameters are kept fixed. Available internal optimizers are:
'fmin_l_bfgs_b'
n_restarts_optimizer [int, optional (default: 0)] The number of restarts of the optimizer for
finding the kernel’s parameters which maximize the log-marginal likelihood. The first run
of the optimizer is performed from the kernel’s initial parameters, the remaining ones (if
any) from thetas sampled log-uniform randomly from the space of allowed theta-values. If
greater than 0, all bounds must be finite. Note that n_restarts_optimizer=0 implies that one
run is performed.
max_iter_predict [int, optional (default: 100)] The maximum number of iterations in New-
ton’s method for approximating the posterior during predict. Smaller values will reduce
computation time at the cost of worse results.
warm_start [bool, optional (default: False)] If warm-starts are enabled, the solution of the last
Newton iteration on the Laplace approximation of the posterior mode is used as initializa-
tion for the next call of _posterior_mode(). This can speed up convergence when _poste-
rior_mode is called several times on similar problems as in hyperparameter optimization.
See the Glossary.
copy_X_train [bool, optional (default: True)] If True, a persistent copy of the training data is
stored in the object. Otherwise, just a reference to the training data is stored, which might
cause predictions to change if the data is modified externally.
random_state [int, RandomState instance or None, optional (default: None)] The generator
used to initialize the centers. If int, random_state is the seed used by the random number
multi_class [string, default] Specifies how multi-class classification problems are handled.
Supported are “one_vs_rest” and “one_vs_one”. In “one_vs_rest”, one binary Gaussian
process classifier is fitted for each class, which is trained to separate this class from the rest.
In “one_vs_one”, one binary Gaussian process classifier is fitted for each pair of classes,
which is trained to separate these two classes. The predictions of these binary predictors are
combined into multi-class predictions. Note that “one_vs_one” does not support predicting
probability estimates.
Attributes
kernel_ [kernel object] The kernel used for prediction. In case of binary classification, the
structure of the kernel is the same as the one passed as parameter but with optimized hy-
perparameters. In case of multi-class classification, a CompoundKernel is returned which
consists of the different kernels used in the one-versus-rest classifiers.
log_marginal_likelihood_value_ [float] The log-marginal-likelihood of self.kernel_.
theta
classes_ [array-like, shape = (n_classes,)] Unique class labels.
6.16. sklearn.gaussian_process: Gaussian Processes 1709

n_classes_ [int] The number of classes in the training data
Examples

>>> from sklearn.gaussian_process import GaussianProcessClassifier
>>> from sklearn.gaussian_process.kernels import RBF
>>> kernel = 1.0 * RBF(1.0)
>>> gpc = GaussianProcessClassifier(kernel=kernel,
... random_state=0).fit(X, y)
>>> gpc.score(X, y)
0.9866...
>>> gpc.predict_proba(X[:2,:])
array([[0.83548752, 0.03228706, 0.13222543],
[0.79064206, 0.06525643, 0.14410151]])
Methods
fit(X, y) Fit Gaussian process classification model

log_marginal_likelihood([theta, Returns log-marginal likelihood of theta for training
eval_gradient]) data.
predict_proba(X) Return probability estimates for the test vector X.
labels.
__init__(kernel=None, optimizer=’fmin_l_bfgs_b’, n_restarts_optimizer=0, max_iter_predict=100,

warm_start=False, copy_X_train=True, random_state=None, multi_class=’one_vs_rest’,
n_jobs=None)
fit(X, y)
Fit Gaussian process classification model
Parameters
X [array-like, shape = (n_samples, n_features)] Training data
y [array-like, shape = (n_samples,)] Target values, must be binary
Returns
self [returns an instance of self.]
Parameters
Returns


log_marginal_likelihood(theta=None, eval_gradient=False)
Returns log-marginal likelihood of theta for training data.
In the case of multi-class classification, the mean log-marginal likelihood of the one-versus-rest classifiers
are returned.
Parameters
theta [array-like, shape = (n_kernel_params,) or none] Kernel hyperparameters for which
the log-marginal likelihood is evaluated. In the case of multi-class classification, theta may
be the hyperparameters of the compound kernel or of an individual kernel. In the latter
case, all individual kernel get assigned the same theta values. If None, the precomputed
log_marginal_likelihood of self.kernel_.theta is returned.
eval_gradient [bool, default: False] If True, the gradient of the log-marginal likelihood with
respect to the kernel hyperparameters at position theta is returned additionally. Note that
gradient computation is not supported for non-binary classification. If True, theta must not
be None.
Returns
log_likelihood [float] Log-marginal likelihood of theta for training data.
log_likelihood_gradient [array, shape = (n_kernel_params,), optional] Gradient of the log-
marginal likelihood with respect to the kernel hyperparameters at position theta. Only
returned when eval_gradient is True.
predict(X)
Parameters
Returns
C [array, shape = (n_samples,)] Predicted target values for X, values are from classes_
predict_proba(X)
Return probability estimates for the test vector X.
Parameters
Returns
C [array-like, shape = (n_samples, n_classes)] Returns the probability of the samples for
each class in the model. The columns correspond to the classes in sorted order, as they
appear in the attribute classes_.
Parameters

Returns
of a nested object.
Returns
self
Examples using sklearn.gaussian_process.GaussianProcessClassifier

• Illustration of Gaussian process classification (GPC) on the XOR dataset
• Iso-probability lines for Gaussian Processes classification (GPC)
• Probabilistic predictions with Gaussian process classification (GPC)
6.16.2 sklearn.gaussian_process.GaussianProcessRegressor
class sklearn.gaussian_process.GaussianProcessRegressor(kernel=None, alpha=1e-10,

optimizer=’fmin_l_bfgs_b’,
n_restarts_optimizer=0,
normalize_y=False,
copy_X_train=True, ran-
dom_state=None)
Gaussian process regression (GPR).
The implementation is based on Algorithm 2.1 of Gaussian Processes for Machine Learning (GPML) by Ras-
mussen and Williams.
In addition to standard scikit-learn estimator API, GaussianProcessRegressor:
• allows prediction without prior fitting (based on the GP prior)
• provides an additional method sample_y(X), which evaluates samples drawn from the GPR (prior or pos-
terior) at given inputs
• exposes a method log_marginal_likelihood(theta), which can be used externally for other ways of selecting
hyperparameters, e.g., via Markov chain Monte Carlo.
Parameters
kernel [kernel object] The kernel specifying the covariance function of the GP. If None is
passed, the kernel “1.0 * RBF(1.0)” is used as default. Note that the kernel’s hyperpa-
rameters are optimized during fitting.

alpha [float or array-like, optional (default: 1e-10)] Value added to the diagonal of the kernel
matrix during fitting. Larger values correspond to increased noise level in the observations.
This can also prevent a potential numerical issue during fitting, by ensuring that the calcu-
lated values form a positive definite matrix. If an array is passed, it must have the same
number of entries as the data used for fitting and is used as datapoint-dependent noise level.
Note that this is equivalent to adding a WhiteKernel with c=alpha. Allowing to specify the
noise level directly as a parameter is mainly for convenience and for consistency with Ridge.
optimizer [string or callable, optional (default: “fmin_l_bfgs_b”)] Can either be one of the
internally supported optimizers for optimizing the kernel’s parameters, specified by a string,
or an externally defined optimizer passed as a callable. If a callable is passed, it must have
the signature:
def optimizer(obj_func, initial_theta, bounds):

# * 'obj_func' is the objective function to be minimized, which
# takes the hyperparameters theta as parameter and an
# optional flag eval_gradient, which determines if the
# gradient is returned additionally to the function value
# * 'initial_theta': the initial value for theta, which can be
# used by local optimizers
# * 'bounds': the bounds on the values of theta
....
# Returned are the best found hyperparameters theta and
# the corresponding value of the target function.
return theta_opt, func_min
Per default, the ‘fmin_l_bfgs_b’ algorithm from scipy.optimize is used. If None is passed,
the kernel’s parameters are kept fixed. Available internal optimizers are:
'fmin_l_bfgs_b'
n_restarts_optimizer [int, optional (default: 0)] The number of restarts of the optimizer for
finding the kernel’s parameters which maximize the log-marginal likelihood. The first run
of the optimizer is performed from the kernel’s initial parameters, the remaining ones (if
any) from thetas sampled log-uniform randomly from the space of allowed theta-values. If
greater than 0, all bounds must be finite. Note that n_restarts_optimizer == 0 implies that
one run is performed.
normalize_y [boolean, optional (default: False)] Whether the target values y are normalized,
i.e., the mean of the observed target values become zero. This parameter should be set to
True if the target values’ mean is expected to differ considerable from zero. When enabled,
the normalization effectively modifies the GP’s prior based on the data, which contradicts
the likelihood principle; normalization is thus disabled per default.
copy_X_train [bool, optional (default: True)] If True, a persistent copy of the training data is
stored in the object. Otherwise, just a reference to the training data is stored, which might
cause predictions to change if the data is modified externally.
random_state [int, RandomState instance or None, optional (default: None)] The generator
Attributes
X_train_ [array-like, shape = (n_samples, n_features)] Feature values in training data (also
required for prediction)

y_train_ [array-like, shape = (n_samples, [n_output_dims])] Target values in training data (also
required for prediction)
kernel_ [kernel object] The kernel used for prediction. The structure of the kernel is the same
as the one passed as parameter but with optimized hyperparameters
L_ [array-like, shape = (n_samples, n_samples)] Lower-triangular Cholesky decomposition of
the kernel in X_train_
alpha_ [array-like, shape = (n_samples,)] Dual coefficients of training data points in kernel
space
log_marginal_likelihood_value_ [float] The log-marginal-likelihood of self.kernel_.
theta
Examples

>>> from sklearn.gaussian_process import GaussianProcessRegressor
>>> from sklearn.gaussian_process.kernels import DotProduct, WhiteKernel
>>> X, y = make_friedman2(n_samples=500, noise=0, random_state=0)
>>> kernel = DotProduct() + WhiteKernel()
>>> gpr = GaussianProcessRegressor(kernel=kernel,
... random_state=0).fit(X, y)
>>> gpr.score(X, y)
0.3680...
>>> gpr.predict(X[:2,:], return_std=True)
(array([653.0..., 592.1...]), array([316.6..., 316.6...]))
Methods
fit(X, y) Fit Gaussian process regression model.

log_marginal_likelihood([theta, Returns log-marginal likelihood of theta for training
eval_gradient]) data.
predict(X[, return_std, return_cov]) Predict using the Gaussian process regression model
sample_y(X[, n_samples, random_state]) Draw samples from Gaussian process and evaluate at X.
diction.
__init__(kernel=None, alpha=1e-10, optimizer=’fmin_l_bfgs_b’, n_restarts_optimizer=0, normal-

ize_y=False, copy_X_train=True, random_state=None)
fit(X, y)
Fit Gaussian process regression model.
Parameters
X [array-like, shape = (n_samples, n_features)] Training data
y [array-like, shape = (n_samples, [n_output_dims])] Target values
Returns

Parameters
Returns
log_marginal_likelihood(theta=None, eval_gradient=False)
Returns log-marginal likelihood of theta for training data.
Parameters
theta [array-like, shape = (n_kernel_params,) or None] Kernel hyperparameters for
which the log-marginal likelihood is evaluated. If None, the precomputed
log_marginal_likelihood of self.kernel_.theta is returned.
eval_gradient [bool, default: False] If True, the gradient of the log-marginal likelihood
with respect to the kernel hyperparameters at position theta is returned additionally. If
True, theta must not be None.
Returns
log_likelihood [float] Log-marginal likelihood of theta for training data.
log_likelihood_gradient [array, shape = (n_kernel_params,), optional] Gradient of the log-
marginal likelihood with respect to the kernel hyperparameters at position theta. Only
returned when eval_gradient is True.
predict(X, return_std=False, return_cov=False)
Predict using the Gaussian process regression model
We can also predict based on an unfitted model by using the GP prior. In addition to the mean of the pre-
dictive distribution, also its standard deviation (return_std=True) or covariance (return_cov=True). Note
that at most one of the two can be requested.
Parameters
X [array-like, shape = (n_samples, n_features)] Query points where the GP is evaluated
return_std [bool, default: False] If True, the standard-deviation of the predictive distribu-
tion at the query points is returned along with the mean.
return_cov [bool, default: False] If True, the covariance of the joint predictive distribution
at the query points is returned along with the mean
Returns
y_mean [array, shape = (n_samples, [n_output_dims])] Mean of predictive distribution a
query points
y_std [array, shape = (n_samples,), optional] Standard deviation of predictive distribution at
query points. Only returned when return_std is True.
y_cov [array, shape = (n_samples, n_samples), optional] Covariance of joint predictive dis-
tribution a query points. Only returned when return_cov is True.
rng
DEPRECATED: Attribute rng was deprecated in version 0.19 and will be removed in 0.21.

sample_y(X, n_samples=1, random_state=0)

Draw samples from Gaussian process and evaluate at X.
Parameters
X [array-like, shape = (n_samples_X, n_features)] Query points where the GP samples are
evaluated
n_samples [int, default: 1] The number of samples drawn from the Gaussian process
random_state [int, RandomState instance or None, optional (default=0)] If int, ran-
Returns
y_samples [array, shape = (n_samples_X, [n_output_dims], n_samples)] Values of
n_samples samples drawn from Gaussian process and evaluated at query points.
Parameters
Returns
of a nested object.
Returns
self
y_train_mean
DEPRECATED: Attribute y_train_mean was deprecated in version 0.19 and will be removed in 0.21.
Examples using sklearn.gaussian_process.GaussianProcessRegressor
• Comparison of kernel ridge and Gaussian process regression

• Illustration of prior and posterior Gaussian process for different kernels
• Gaussian process regression (GPR) with noise-level estimation

• Gaussian Processes regression: basic introductory example

Kernels:
gaussian_process.kernels. Kernel which is composed of a set of other kernels.

CompoundKernel(kernels)
gaussian_process.kernels. Constant kernel.
ConstantKernel([. . . ])
gaussian_process.kernels. Dot-Product kernel.
DotProduct([. . . ])
gaussian_process.kernels. Exp-Sine-Squared kernel.
ExpSineSquared([. . . ])
gaussian_process.kernels. Exponentiate kernel by given exponent.
Exponentiation(. . . )
gaussian_process.kernels.Hyperparameter A kernel hyperparameter’s specification in form of a
namedtuple.
gaussian_process.kernels.Kernel Base class for all kernels.
gaussian_process.kernels.Matern([. . . ]) Matern kernel.
gaussian_process.kernels. Wrapper for kernels in sklearn.metrics.pairwise.
PairwiseKernel([. . . ])
gaussian_process.kernels.Product(k1, k2) Product-kernel k1 * k2 of two kernels k1 and k2.
gaussian_process.kernels.RBF([length_scale, Radial-basis function kernel (aka squared-exponential ker-
. . . ]) nel).
gaussian_process.kernels. Rational Quadratic kernel.
RationalQuadratic([. . . ])
gaussian_process.kernels.Sum(k1, k2) Sum-kernel k1 + k2 of two kernels k1 and k2.
gaussian_process.kernels. White kernel.
WhiteKernel([. . . ])
6.16.3 sklearn.gaussian_process.kernels.CompoundKernel
class sklearn.gaussian_process.kernels.CompoundKernel(kernels)
Kernel which is composed of a set of other kernels.
Parameters
kernels [list of Kernel objects] The other kernels
Attributes
bounds Returns the log-transformed bounds on the theta.
hyperparameters Returns a list of all hyperparameter specifications.
n_dims Returns the number of non-fixed hyperparameters of the kernel.
theta Returns the (flattened, log-transformed) non-fixed hyperparameters.
Methods

__call__(X[, Y, eval_gradient]) Return the kernel k(X, Y) and optionally its gradient.
clone_with_theta(theta) Returns a clone of self with given hyperparameters
theta.
diag(X) Returns the diagonal of the kernel k(X, X).
get_params([deep]) Get parameters of this kernel.
is_stationary() Returns whether the kernel is stationary.
set_params(**params) Set the parameters of this kernel.
__init__(kernels)
__call__(X, Y=None, eval_gradient=False)
Return the kernel k(X, Y) and optionally its gradient.
Note that this compound kernel returns the results of all simple kernel stacked along an additional axis.
Parameters
X [array, shape (n_samples_X, n_features)] Left argument of the returned kernel k(X, Y)
Y [array, shape (n_samples_Y, n_features), (optional, default=None)] Right argument of the
returned kernel k(X, Y). If None, k(X, X) if evaluated instead.
eval_gradient [bool (optional, default=False)] Determines whether the gradient with re-
spect to the kernel hyperparameter is determined.
Returns
K [array, shape (n_samples_X, n_samples_Y, n_kernels)] Kernel k(X, Y)
K_gradient [array, shape (n_samples_X, n_samples_X, n_dims, n_kernels)] The gradient
of the kernel k(X, X) with respect to the hyperparameter of the kernel. Only returned when
eval_gradient is True.
bounds
Returns the log-transformed bounds on the theta.
Returns
bounds [array, shape (n_dims, 2)] The log-transformed bounds on the kernel’s hyperparam-
eters theta
clone_with_theta(theta)
Returns a clone of self with given hyperparameters theta.
Parameters
theta [array, shape (n_dims,)] The hyperparameters
diag(X)
Returns the diagonal of the kernel k(X, X).
The result of this method is identical to np.diag(self(X)); however, it can be evaluated more efficiently
since only the diagonal is evaluated.
Parameters
Returns
K_diag [array, shape (n_samples_X, n_kernels)] Diagonal of kernel k(X, X)
Get parameters of this kernel.

Parameters
Returns
hyperparameters
Returns a list of all hyperparameter specifications.
is_stationary()
Returns whether the kernel is stationary.
n_dims
Returns the number of non-fixed hyperparameters of the kernel.
Set the parameters of this kernel.
The method works on simple kernels as well as on nested kernels. The latter have parameters of the form
<component>__<parameter> so that it’s possible to update each component of a nested object.
Returns
self
theta
Returns the (flattened, log-transformed) non-fixed hyperparameters.
Note that theta are typically the log-transformed values of the kernel’s hyperparameters as this representa-
tion of the search space is more amenable for hyperparameter search, as hyperparameters like length-scales
naturally live on a log-scale.
Returns
theta [array, shape (n_dims,)] The non-fixed, log-transformed hyperparameters of the kernel
6.16.4 sklearn.gaussian_process.kernels.ConstantKernel
class sklearn.gaussian_process.kernels.ConstantKernel(constant_value=1.0,
constant_value_bounds=(1e-
05, 100000.0))
Constant kernel.
Can be used as part of a product-kernel where it scales the magnitude of the other factor (kernel) or as part of a
sum-kernel, where it modifies the mean of the Gaussian process.
k(x_1, x_2) = constant_value for all x_1, x_2
Parameters
constant_value [float, default: 1.0] The constant value which defines the covariance: k(x_1,
x_2) = constant_value
constant_value_bounds [pair of floats >= 0, default: (1e-5, 1e5)] The lower and upper bound
on constant_value
Attributes

hyperparameter_constant_value
Methods
theta.
__init__(constant_value=1.0, constant_value_bounds=(1e-05, 100000.0))

Parameters
spect to the kernel hyperparameter is determined. Only supported when Y is None.
Returns
K [array, shape (n_samples_X, n_samples_Y)] Kernel k(X, Y)
K_gradient [array (opt.), shape (n_samples_X, n_samples_X, n_dims)] The gradient of the
kernel k(X, X) with respect to the hyperparameter of the kernel. Only returned when
bounds
Returns
eters theta
Parameters
diag(X)

Parameters
Returns
K_diag [array, shape (n_samples_X,)] Diagonal of kernel k(X, X)
Parameters
Returns
hyperparameters
is_stationary()
n_dims
Returns
self
theta
Returns
Examples using sklearn.gaussian_process.kernels.ConstantKernel

6.16.5 sklearn.gaussian_process.kernels.DotProduct
class sklearn.gaussian_process.kernels.DotProduct(sigma_0=1.0, sigma_0_bounds=(1e-

05, 100000.0))
Dot-Product kernel.

The DotProduct kernel is non-stationary and can be obtained from linear regression by putting N(0, 1) priors
on the coefficients of x_d (d = 1, . . . , D) and a prior of N(0, sigma_0^2) on the bias. The DotProduct
kernel is invariant to a rotation of the coordinates about the origin, but not translations. It is parameterized by
a parameter sigma_0^2. For sigma_0^2 =0, the kernel is called the homogeneous linear kernel, otherwise it is
inhomogeneous. The kernel is given by
k(x_i, x_j) = sigma_0 ^ 2 + x_i cdot x_j
The DotProduct kernel is commonly combined with exponentiation.
Parameters
sigma_0 [float >= 0, default: 1.0] Parameter controlling the inhomogenity of the kernel. If
sigma_0=0, the kernel is homogenous.
sigma_0_bounds [pair of floats >= 0, default: (1e-5, 1e5)] The lower and upper bound on l
Attributes
hyperparameter_sigma_0
Methods
theta.
__init__(sigma_0=1.0, sigma_0_bounds=(1e-05, 100000.0))

Parameters
Returns

bounds
Returns
eters theta
Parameters
diag(X)
Parameters
Returns
Parameters
Returns
hyperparameters
is_stationary()
n_dims
Returns
self
theta

Returns
Examples using sklearn.gaussian_process.kernels.DotProduct

6.16.6 sklearn.gaussian_process.kernels.ExpSineSquared
class sklearn.gaussian_process.kernels.ExpSineSquared(length_scale=1.0,
periodicity=1.0,
length_scale_bounds=(1e-
05, 100000.0),
periodicity_bounds=(1e-05,
100000.0))
Exp-Sine-Squared kernel.
The ExpSineSquared kernel allows modeling periodic functions. It is parameterized by a length-scale parameter
length_scale>0 and a periodicity parameter periodicity>0. Only the isotropic variant where l is a scalar is
supported at the moment. The kernel given by:
k(x_i, x_j) = exp(-2 (sin(pi / periodicity * d(x_i, x_j)) / length_scale) ^ 2)
Parameters
length_scale [float > 0, default: 1.0] The length scale of the kernel.
periodicity [float > 0, default: 1.0] The periodicity of the kernel.
length_scale_bounds [pair of floats >= 0, default: (1e-5, 1e5)] The lower and upper bound on
length_scale
periodicity_bounds [pair of floats >= 0, default: (1e-5, 1e5)] The lower and upper bound on
periodicity
Attributes
hyperparameter_length_scale
hyperparameter_periodicity

Methods
theta.
__init__(length_scale=1.0, periodicity=1.0, length_scale_bounds=(1e-05, 100000.0),

periodicity_bounds=(1e-05, 100000.0))
Parameters
Returns
bounds
Returns
eters theta
Parameters
diag(X)
Parameters
Returns

Parameters
Returns
hyperparameters
is_stationary()
n_dims
Returns
self
theta
Returns
Examples using sklearn.gaussian_process.kernels.ExpSineSquared

6.16.7 sklearn.gaussian_process.kernels.Exponentiation
class sklearn.gaussian_process.kernels.Exponentiation(kernel, exponent)

Exponentiate kernel by given exponent.
The resulting kernel is defined as k_exp(X, Y) = k(X, Y) ** exponent
Parameters
kernel [Kernel object] The base kernel

exponent [float] The exponent for the base kernel

Attributes
hyperparameters Returns a list of all hyperparameter.
Methods
theta.
__init__(kernel, exponent)
Parameters
Returns
bounds
Returns
eters theta
Parameters
diag(X)

Parameters
Returns
Parameters
Returns
hyperparameters
Returns a list of all hyperparameter.
is_stationary()
n_dims
Returns
self
theta
Returns
6.16.8 sklearn.gaussian_process.kernels.Hyperparameter
class sklearn.gaussian_process.kernels.Hyperparameter
A kernel hyperparameter’s specification in form of a namedtuple.
Attributes
name [string] Alias for field number 0
value_type [string] Alias for field number 1

bounds [pair of floats >= 0 or “fixed”] Alias for field number 2

n_elements [int, default=1] Alias for field number 3
fixed [bool, default: None] Alias for field number 4
Methods
count(. . . )
index((value, [start, . . . ) Raises ValueError if the value is not present.

__call__($self, /, *args, **kwargs)
Call self as a function.
bounds
Alias for field number 2
count(value) → integer – return number of occurrences of value
fixed
index(value[, start[, stop ]]) → integer – return first index of value.
Raises ValueError if the value is not present.
n_elements
name
value_type
6.16.9 sklearn.gaussian_process.kernels.Kernel
class sklearn.gaussian_process.kernels.Kernel
Base class for all kernels.
Attributes
Methods

__call__(X[, Y, eval_gradient]) Evaluate the kernel.

theta.

Evaluate the kernel.
bounds
Returns
eters theta
Parameters
diag(X)
Parameters
Returns
Parameters
Returns
hyperparameters
is_stationary()
n_dims

Returns
self
theta
Returns
6.16.10 sklearn.gaussian_process.kernels.Matern
class sklearn.gaussian_process.kernels.Matern(length_scale=1.0, length_scale_bounds=(1e-

05, 100000.0), nu=1.5)
Matern kernel.
The class of Matern kernels is a generalization of the RBF and the absolute exponential kernel parameterized by
an additional parameter nu. The smaller nu, the less smooth the approximated function is. For nu=inf, the kernel
becomes equivalent to the RBF kernel and for nu=0.5 to the absolute exponential kernel. Important intermediate
values are nu=1.5 (once differentiable functions) and nu=2.5 (twice differentiable functions).
See Rasmussen and Williams 2006, pp84 for details regarding the different variants of the Matern kernel.
Parameters
length_scale [float or array with shape (n_features,), default: 1.0] The length scale of the kernel.
If a float, an isotropic kernel is used. If an array, an anisotropic kernel is used where each
dimension of l defines the length-scale of the respective feature dimension.
length_scale
nu [float, default: 1.5] The parameter nu controlling the smoothness of the learned function.
The smaller nu, the less smooth the approximated function is. For nu=inf, the kernel be-
comes equivalent to the RBF kernel and for nu=0.5 to the absolute exponential kernel. Im-
portant intermediate values are nu=1.5 (once differentiable functions) and nu=2.5 (twice
differentiable functions). Note that values of nu not in [0.5, 1.5, 2.5, inf] incur a consid-
erably higher computational cost (appr. 10 times higher) since they require to evaluate the
modified Bessel function. Furthermore, in contrast to l, nu is kept fixed to its initial value
and not optimized.
Attributes
anisotropic


Methods
theta.
__init__(length_scale=1.0, length_scale_bounds=(1e-05, 100000.0), nu=1.5)

Parameters
Returns
bounds
Returns
eters theta
Parameters
diag(X)
Parameters

Returns
Parameters
Returns
hyperparameters
is_stationary()
n_dims
Returns
self
theta
Returns
Examples using sklearn.gaussian_process.kernels.Matern
6.16.11 sklearn.gaussian_process.kernels.PairwiseKernel
class sklearn.gaussian_process.kernels.PairwiseKernel(gamma=1.0,
gamma_bounds=(1e-05,
100000.0), metric=’linear’, pair-
wise_kernels_kwargs=None)
Wrapper for kernels in sklearn.metrics.pairwise.
A thin wrapper around the functionality of the kernels in sklearn.metrics.pairwise.

Note: Evaluation of eval_gradient is not analytic but numeric and all kernels support only isotropic dis-
tances. The parameter gamma is considered to be a hyperparameter and may be optimized. The other
kernel parameters are set directly at initialization and are kept fixed.
Parameters
gamma [float >= 0, default: 1.0] Parameter gamma of the pairwise kernel specified by metric
gamma_bounds [pair of floats >= 0, default: (1e-5, 1e5)] The lower and upper bound on
gamma
metric [string, or callable, default: “linear”] The metric to use when calculating kernel between
instances in a feature array. If metric is a string, it must be one of the metrics in pair-
wise.PAIRWISE_KERNEL_FUNCTIONS. If metric is “precomputed”, X is assumed to be
a kernel matrix. Alternatively, if metric is a callable function, it is called on each pair of
instances (rows) and the resulting value recorded. The callable should take two arrays from
X as input and return a value indicating the distance between them.
pairwise_kernels_kwargs [dict, default: None] All entries of this dict (if any) are passed as
keyword arguments to the pairwise kernel function.
Attributes
hyperparameter_gamma
Methods
theta.
__init__(gamma=1.0, gamma_bounds=(1e-05, 100000.0), metric=’linear’, pair-

wise_kernels_kwargs=None)
Parameters

Returns
bounds
Returns
eters theta
Parameters
diag(X)
Parameters
Returns
Parameters
Returns
hyperparameters
is_stationary()
n_dims
Returns
self

theta
Returns
6.16.12 sklearn.gaussian_process.kernels.Product
class sklearn.gaussian_process.kernels.Product(k1, k2)

Product-kernel k1 * k2 of two kernels k1 and k2.
The resulting kernel is defined as k_prod(X, Y) = k1(X, Y) * k2(X, Y)
Parameters
k1 [Kernel object] The first base-kernel of the product-kernel
k2 [Kernel object] The second base-kernel of the product-kernel
Attributes
Methods
theta.
__init__(k1, k2)
Parameters

Returns
bounds
Returns
eters theta
Parameters
diag(X)
Parameters
Returns
Parameters
Returns
hyperparameters
is_stationary()
n_dims
Returns
self

theta
Returns
6.16.13 sklearn.gaussian_process.kernels.RBF
class sklearn.gaussian_process.kernels.RBF(length_scale=1.0, length_scale_bounds=(1e-05,

100000.0))
Radial-basis function kernel (aka squared-exponential kernel).
The RBF kernel is a stationary kernel. It is also known as the “squared exponential” kernel. It is parameterized
by a length-scale parameter length_scale>0, which can either be a scalar (isotropic variant of the kernel) or a
vector with the same number of dimensions as the inputs X (anisotropic variant of the kernel). The kernel is
given by:
k(x_i, x_j) = exp(-1 / 2 d(x_i / length_scale, x_j / length_scale)^2)
This kernel is infinitely differentiable, which implies that GPs with this kernel as covariance function have mean
square derivatives of all orders, and are thus very smooth.
Parameters
length_scale [float or array with shape (n_features,), default: 1.0] The length scale of the kernel.
If a float, an isotropic kernel is used. If an array, an anisotropic kernel is used where each
dimension of l defines the length-scale of the respective feature dimension.
length_scale
Attributes
anisotropic
Methods
theta.


__init__(length_scale=1.0, length_scale_bounds=(1e-05, 100000.0))

Parameters
Returns
bounds
Returns
eters theta
Parameters
diag(X)
Parameters
Returns
Parameters
Returns


hyperparameters
is_stationary()
n_dims
Returns
self
theta
Returns
Examples using sklearn.gaussian_process.kernels.RBF

6.16.14 sklearn.gaussian_process.kernels.RationalQuadratic
class sklearn.gaussian_process.kernels.RationalQuadratic(length_scale=1.0,
alpha=1.0,
length_scale_bounds=(1e-
05, 100000.0),
alpha_bounds=(1e-05,
100000.0))
Rational Quadratic kernel.

The RationalQuadratic kernel can be seen as a scale mixture (an infinite sum) of RBF kernels with different
characteristic length-scales. It is parameterized by a length-scale parameter length_scale>0 and a scale mixture
parameter alpha>0. Only the isotropic variant where length_scale is a scalar is supported at the moment. The
kernel given by:
k(x_i, x_j) = (1 + d(x_i, x_j)^2 / (2*alpha * length_scale^2))^-alpha
Parameters
length_scale [float > 0, default: 1.0] The length scale of the kernel.
alpha [float > 0, default: 1.0] Scale mixture parameter
length_scale
alpha_bounds [pair of floats >= 0, default: (1e-5, 1e5)] The lower and upper bound on alpha
Attributes
hyperparameter_alpha
Methods
theta.
__init__(length_scale=1.0, alpha=1.0, length_scale_bounds=(1e-05, 100000.0), alpha_bounds=(1e-

05, 100000.0))
Parameters
Returns


bounds
Returns
eters theta
Parameters
diag(X)
Parameters
Returns
Parameters
Returns
hyperparameters
is_stationary()
n_dims
Returns
self

theta
Returns
Examples using sklearn.gaussian_process.kernels.RationalQuadratic

6.16.15 sklearn.gaussian_process.kernels.Sum
class sklearn.gaussian_process.kernels.Sum(k1, k2)

Sum-kernel k1 + k2 of two kernels k1 and k2.
The resulting kernel is defined as k_sum(X, Y) = k1(X, Y) + k2(X, Y)
Parameters
k1 [Kernel object] The first base-kernel of the sum-kernel
k2 [Kernel object] The second base-kernel of the sum-kernel
Attributes
Methods
theta.
__init__(k1, k2)
Parameters

Returns
bounds
Returns
eters theta
Parameters
diag(X)
Parameters
Returns
Parameters
Returns
hyperparameters
is_stationary()
n_dims

Returns
self
theta
Returns
6.16.16 sklearn.gaussian_process.kernels.WhiteKernel
class sklearn.gaussian_process.kernels.WhiteKernel(noise_level=1.0,
noise_level_bounds=(1e-05,
100000.0))
White kernel.
The main use-case of this kernel is as part of a sum-kernel where it explains the noise-component of the signal.
Tuning its parameter corresponds to estimating the noise-level.
k(x_1, x_2) = noise_level if x_1 == x_2 else 0
Parameters
noise_level [float, default: 1.0] Parameter controlling the noise level
noise_level_bounds [pair of floats >= 0, default: (1e-5, 1e5)] The lower and upper bound on
noise_level
Attributes
hyperparameter_noise_level
Methods
theta.


__init__(noise_level=1.0, noise_level_bounds=(1e-05, 100000.0))

Parameters
Returns
bounds
Returns
eters theta
Parameters
diag(X)
Parameters
Returns
Parameters
Returns


hyperparameters
is_stationary()
n_dims
Returns
self
theta
Returns
Examples using sklearn.gaussian_process.kernels.WhiteKernel

6.17 sklearn.isotonic: Isotonic regression
User guide: See the Isotonic regression section for further details.
isotonic.IsotonicRegression([y_min, y_max, Isotonic regression model.

. . . ])
6.17.1 sklearn.isotonic.IsotonicRegression
class sklearn.isotonic.IsotonicRegression(y_min=None, y_max=None, increasing=True,

out_of_bounds=’nan’)
Isotonic regression model.
The isotonic regression optimization problem is defined by:
min sum w_i (y[i] - y_[i]) ** 2
6.17. sklearn.isotonic: Isotonic regression 1747

subject to y_[i] <= y_[j] whenever X[i] <= X[j]

and min(y_) = y_min, max(y_) = y_max
where:
• y[i] are inputs (real numbers)
• y_[i] are fitted
• X specifies the order. If X is non-decreasing then y_ is non-decreasing.
• w[i] are optional strictly positive weights (default to 1.0)

Parameters
y_min [optional, default: None] If not None, set the lowest value of the fit to y_min.
y_max [optional, default: None] If not None, set the highest value of the fit to y_max.
increasing [boolean or string, optional, default: True] If boolean, whether or not to fit the
isotonic regression with y increasing or decreasing.
The string value “auto” determines whether y should increase or decrease based on the
Spearman correlation estimate’s sign.
out_of_bounds [string, optional, default: “nan”] The out_of_bounds parameter handles
how x-values outside of the training domain are handled. When set to “nan”, predicted
y-values will be NaN. When set to “clip”, predicted y-values will be set to the value corre-
sponding to the nearest train interval endpoint. When set to “raise”, allow interp1d to
throw ValueError.
Attributes
X_min_ [float] Minimum value of input array X_ for left bound.
X_max_ [float] Maximum value of input array X_ for right bound.
f_ [function] The stepwise interpolating function that covers the input domain X.
Notes
Ties are broken using the secondary method from Leeuw, 1977.
References
Isotonic Median Regression: A Linear Programming Approach Nilotpal Chakravarti Mathematics of Operations
Research Vol. 14, No. 2 (May, 1989), pp. 303-308
Isotone Optimization in R : Pool-Adjacent-Violators Algorithm (PAVA) and Active Set Methods Leeuw, Hornik,
Mair Journal of Statistical Software 2009
Correctness of Kruskal’s algorithms for monotone regression with ties Leeuw, Psychometrica, 1977
Methods

fit(X, y[, sample_weight]) Fit the model using X, y as training data.

predict(T) Predict new data by linear interpolation.
diction.
transform(T) Transform new data by linear interpolation
__init__(y_min=None, y_max=None, increasing=True, out_of_bounds=’nan’)

Parameters
X [array-like, shape=(n_samples,)] Training data.
y [array-like, shape=(n_samples,)] Training target.
sample_weight [array-like, shape=(n_samples,), optional, default: None] Weights. If set to
None, all weights will be set to 1 (equal weights).
Returns
self [object] Returns an instance of self.
Notes
X is stored for future use, as transform needs X to interpolate new input data.
Parameters
Returns
Parameters
Returns
predict(T)
Predict new data by linear interpolation.
Parameters

T [array-like, shape=(n_samples,)] Data to transform.

Returns
T_ [array, shape=(n_samples,)] Transformed data.
Parameters
Returns
of a nested object.
Returns
self
transform(T)
Transform new data by linear interpolation
Parameters
T [array-like, shape=(n_samples,)] Data to transform.
Returns
T_ [array, shape=(n_samples,)] The transformed data
Examples using sklearn.isotonic.IsotonicRegression
• Isotonic Regression
isotonic.check_increasing(x, y) Determine whether y is monotonically correlated with x.

isotonic.isotonic_regression(y[, . . . ]) Solve the isotonic regression model:
6.17.2 sklearn.isotonic.check_increasing
sklearn.isotonic.check_increasing(x, y)
Determine whether y is monotonically correlated with x.

y is found increasing or decreasing with respect to x based on a Spearman correlation test.

Parameters
x [array-like, shape=(n_samples,)] Training data.
y [array-like, shape=(n_samples,)] Training target.
Returns
increasing_bool [boolean] Whether the relationship is increasing or decreasing.
Notes
The Spearman correlation coefficient is estimated from the data, and the sign of the resulting estimate is used as
the result.
In the event that the 95% confidence interval based on Fisher transform spans zero, a warning is raised.
References
Fisher transformation. Wikipedia. https://en.wikipedia.org/wiki/Fisher_transformation
6.17.3 sklearn.isotonic.isotonic_regression
sklearn.isotonic.isotonic_regression(y, sample_weight=None, y_min=None, y_max=None,

increasing=True)
Solve the isotonic regression model:
min sum w[i] (y[i] - y_[i]) ** 2
subject to y_min = y_[1] <= y_[2] ... <= y_[n] = y_max
where:
• y[i] are inputs (real numbers)
• y_[i] are fitted
• w[i] are optional strictly positive weights (default to 1.0)

Parameters
y [iterable of floats] The data.
sample_weight [iterable of floats, optional, default: None] Weights on each point of the regres-
sion. If None, weight is set to 1 (equal weights).
y_min [optional, default: None] If not None, set the lowest value of the fit to y_min.
y_max [optional, default: None] If not None, set the highest value of the fit to y_max.
increasing [boolean, optional, default: True] Whether to compute y_ is increasing (if set to
True) or decreasing (if set to False)
Returns
y_ [list of floats] Isotonic fit of y.

References
“Active set algorithms for isotonic regression; A unifying framework” by Michael J. Best and Nilotpal
Chakravarti, section 3.
6.18 sklearn.impute: Impute
Transformers for missing value imputation

User guide: See the Imputation of missing values section for further details.
impute.SimpleImputer([missing_values, . . . ]) Imputation transformer for completing missing values.

impute.MissingIndicator([missing_values, . . . ]) Binary indicators for missing values.
6.18.1 sklearn.impute.SimpleImputer
class sklearn.impute.SimpleImputer(missing_values=nan, strategy=’mean’, fill_value=None, ver-

bose=0, copy=True)
Imputation transformer for completing missing values.
Parameters
missing_values [number, string, np.nan (default) or None] The placeholder for the missing
values. All occurrences of missing_values will be imputed.
strategy [string, optional (default=”mean”)] The imputation strategy.
• If “mean”, then replace missing values using the mean along each column. Can only be
used with numeric data.
• If “median”, then replace missing values using the median along each column. Can only
be used with numeric data.
• If “most_frequent”, then replace missing using the most frequent value along each col-
umn. Can be used with strings or numeric data.
• If “constant”, then replace missing values with fill_value. Can be used with strings or
numeric data.
New in version 0.20: strategy=”constant” for fixed value imputation.
fill_value [string or numerical value, optional (default=None)] When strategy == “constant”,
fill_value is used to replace all occurrences of missing_values. If left to the default, fill_value
will be 0 when imputing numerical data and “missing_value” for strings or object data types.
verbose [integer, optional (default=0)] Controls the verbosity of the imputer.
copy [boolean, optional (default=True)] If True, a copy of X will be created. If False, imputation
will be done in-place whenever possible. Note that, in the following cases, a new copy will
always be made, even if copy=False:
• If X is not an array of floating values;
• If X is encoded as a CSR matrix.
Attributes

statistics_ [array of shape (n_features,)] The imputation fill value for each feature.
Notes
Columns which only contained missing values at fit are discarded upon transform if strategy is not “constant”.
Examples

>>> from sklearn.impute import SimpleImputer
>>> imp_mean = SimpleImputer(missing_values=np.nan, strategy='mean')
>>> imp_mean.fit([[7, 2, 3], [4, np.nan, 6], [10, 5, 9]])
...
SimpleImputer(copy=True, fill_value=None, missing_values=nan,
strategy='mean', verbose=0)
>>> X = [[np.nan, 2, 3], [4, np.nan, 6], [10, np.nan, 9]]
>>> print(imp_mean.transform(X))
...
[[ 7. 2. 3. ]
[ 4. 3.5 6. ]
[10. 3.5 9. ]]
Methods
fit(X[, y]) Fit the imputer on X.

transform(X) Impute all missing values in X.
__init__(missing_values=nan, strategy=’mean’, fill_value=None, verbose=0, copy=True)

fit(X, y=None)
Fit the imputer on X.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Input data, where
Returns
self [SimpleImputer]
Parameters
Returns
6.18. sklearn.impute: Impute 1753


Parameters
Returns
of a nested object.
Returns
self
transform(X)
Impute all missing values in X.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] The input data to complete.
Examples using sklearn.impute.SimpleImputer

6.18.2 sklearn.impute.MissingIndicator
class sklearn.impute.MissingIndicator(missing_values=nan, features=’missing-only’,

sparse=’auto’, error_on_new=True)
Binary indicators for missing values.
Note that this component typically should not not be used in a vanilla Pipeline consisting of transformers
and a classifier, but rather could be added using a FeatureUnion or ColumnTransformer.
Parameters
missing_values [number, string, np.nan (default) or None] The placeholder for the missing
values. All occurrences of missing_values will be indicated (True in the output array), the
other values will be marked as False.
features [str, optional] Whether the imputer mask should represent all or a subset of features.
• If “missing-only” (default), the imputer mask will only represent features containing
missing values during fit time.
• If “all”, the imputer mask will represent all features.

sparse [boolean or “auto”, optional] Whether the imputer mask format should be sparse or
dense.
• If “auto” (default), the imputer mask will be of same type as input.
• If True, the imputer mask will be a sparse matrix.
• If False, the imputer mask will be a numpy array.
error_on_new [boolean, optional] If True (default), transform will raise an error when there
are features with missing values in transform that have no missing values in fit. This is
applicable only when features="missing-only".
Attributes
features_ [ndarray, shape (n_missing_features,) or (n_features,)] The features indices which
will be returned when calling transform. They are computed during fit. For
features='all', it is to range(n_features).
Examples

>>> from sklearn.impute import MissingIndicator
>>> X1 = np.array([[np.nan, 1, 3],
... [4, 0, np.nan],
... [8, 1, 0]])
>>> X2 = np.array([[5, 1, np.nan],
... [np.nan, 2, 3],
... [2, 4, 0]])
>>> indicator = MissingIndicator()
>>> indicator.fit(X1)
MissingIndicator(error_on_new=True, features='missing-only',
missing_values=nan, sparse='auto')
>>> X2_tr = indicator.transform(X2)
>>> X2_tr
array([[False, True],
[ True, False],
[False, False]])
Methods
fit(X[, y]) Fit the transformer on X.

fit_transform(X[, y]) Generate missing values indicator for X.
transform(X) Generate missing values indicator for X.
__init__(missing_values=nan, features=’missing-only’, sparse=’auto’, error_on_new=True)

fit(X, y=None)
Fit the transformer on X.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Input data, where
6.18. sklearn.impute: Impute 1755

Returns
self [object] Returns self.
Generate missing values indicator for X.
Parameters
Returns
Xt [{ndarray or sparse matrix}, shape (n_samples, n_features)] The missing indicator for
input data. The data type of Xt will be boolean.
Parameters
Returns
of a nested object.
Returns
self
transform(X)
Generate missing values indicator for X.
Parameters
Returns
Xt [{ndarray or sparse matrix}, shape (n_samples, n_features)] The missing indicator for
input data. The data type of Xt will be boolean.
Examples using sklearn.impute.MissingIndicator
6.19 sklearn.kernel_approximation Kernel Approximation
The sklearn.kernel_approximation module implements several approximate kernel feature maps base on
Fourier transforms.
User guide: See the Kernel Approximation section for further details.

kernel_approximation. Approximate feature map for additive chi2 kernel.

AdditiveChi2Sampler([. . . ])
kernel_approximation.Nystroem([kernel, . . . ]) Approximate a kernel map using a subset of the training
data.
kernel_approximation.RBFSampler([gamma, Approximates feature map of an RBF kernel by Monte
. . . ]) Carlo approximation of its Fourier transform.
kernel_approximation. Approximates feature map of the “skewed chi-squared”
SkewedChi2Sampler([. . . ]) kernel by Monte Carlo approximation of its Fourier trans-
form.
6.19.1 sklearn.kernel_approximation.AdditiveChi2Sampler
class sklearn.kernel_approximation.AdditiveChi2Sampler(sample_steps=2, sam-

ple_interval=None)
Approximate feature map for additive chi2 kernel.
Uses sampling the fourier transform of the kernel characteristic at regular intervals.
Since the kernel that is to be approximated is additive, the components of the input vectors can be treated
separately. Each entry in the original space is transformed into 2*sample_steps+1 features, where sample_steps
is a parameter of the method. Typical values of sample_steps include 1, 2 and 3.
Optimal choices for the sampling interval for certain data ranges can be computed (see the reference). The
default values should be reasonable.
Parameters
sample_steps [int, optional] Gives the number of (complex) sampling points.
sample_interval [float, optional] Sampling interval. Must be specified when sample_steps not
in {1,2,3}.
See also:
SkewedChi2Sampler A Fourier-approximation to a non-additive variant of the chi squared kernel.

sklearn.metrics.pairwise.chi2_kernel The exact chi squared kernel.
sklearn.metrics.pairwise.additive_chi2_kernel The exact additive chi squared kernel.
Notes
This estimator approximates a slightly different version of the additive chi squared kernel then metric.
additive_chi2 computes.
References
See “Efficient additive kernels via explicit feature maps” A. Vedaldi and A. Zisserman, Pattern Analysis and
Machine Intelligence, 2011
6.19. sklearn.kernel_approximation Kernel Approximation 1757

Examples

>>> from sklearn.kernel_approximation import AdditiveChi2Sampler
>>> chi2sampler = AdditiveChi2Sampler(sample_steps=2)
>>> X_transformed = chi2sampler.fit_transform(X, y)
>>> clf = SGDClassifier(max_iter=5, random_state=0, tol=1e-3)
>>> clf.fit(X_transformed, y)
power_t=0.5, random_state=0, shuffle=True, tol=0.001,
>>> clf.score(X_transformed, y)
0.9543...
Methods
fit(X[, y]) Set the parameters

transform(X) Apply approximate feature map to X.
__init__(sample_steps=2, sample_interval=None)
fit(X, y=None)
Set the parameters
Parameters
X [array-like, shape (n_samples, n_features)] Training data, where n_samples in the number
Returns
self [object] Returns the transformer.
Parameters
Returns

Parameters
Returns
of a nested object.
Returns
self
transform(X)
Apply approximate feature map to X.
Parameters
X [{array-like, sparse matrix}, shape = (n_samples, n_features)]
Returns
X_new [{array, sparse matrix}, shape = (n_samples, n_features * (2*sample_steps + 1))]
Whether the return value is an array of sparse matrix depends on the type of the input X.
6.19.2 sklearn.kernel_approximation.Nystroem
class sklearn.kernel_approximation.Nystroem(kernel=’rbf’, gamma=None, coef0=None,

degree=None, kernel_params=None,
n_components=100, random_state=None)
Approximate a kernel map using a subset of the training data.
Constructs an approximate feature map for an arbitrary kernel using a subset of the data as basis.
Parameters
kernel [string or callable, default=”rbf”] Kernel map to be approximated. A callable should
accept two arguments and the keyword arguments passed to this object as kernel_params,
and should return a floating point number.
gamma [float, default=None] Gamma parameter for the RBF, laplacian, polynomial, exponen-
tial chi2 and sigmoid kernels. Interpretation of the default value is left to the kernel; see the
documentation for sklearn.metrics.pairwise. Ignored by other kernels.
coef0 [float, default=None] Zero coefficient for polynomial and sigmoid kernels. Ignored by
other kernels.
degree [float, default=None] Degree of the polynomial kernel. Ignored by other kernels.
kernel_params [mapping of string to any, optional] Additional parameters (keyword argu-
ments) for kernel function passed as callable object.
n_components [int] Number of features to construct. How many data points will be used to
construct the mapping.


Attributes
components_ [array, shape (n_components, n_features)] Subset of training points used to con-
struct the feature map.
component_indices_ [array, shape (n_components)] Indices of components_ in the training
set.
normalization_ [array, shape (n_components, n_components)] Normalization matrix needed
for embedding. Square root of the kernel matrix on components_.
See also:
RBFSampler An approximation to the RBF kernel using random Fourier features.

sklearn.metrics.pairwise.kernel_metrics List of built-in kernels.
References
• Williams, C.K.I. and Seeger, M. “Using the Nystroem method to speed up kernel machines”, Advances in
neural information processing systems 2001
• T. Yang, Y. Li, M. Mahdavi, R. Jin and Z. Zhou “Nystroem Method vs Random Fourier Features: A
Theoretical and Empirical Comparison”, Advances in Neural Information Processing Systems 2012
Examples
>>> from sklearn import datasets, svm

>>> from sklearn.kernel_approximation import Nystroem
>>> digits = datasets.load_digits(n_class=9)
>>> data = digits.data / 16.
>>> clf = svm.LinearSVC()
>>> feature_map_nystroem = Nystroem(gamma=.2,
... random_state=1,
... n_components=300)
>>> data_transformed = feature_map_nystroem.fit_transform(data)
>>> clf.fit(data_transformed, digits.target)
...
verbose=0)
>>> clf.score(data_transformed, digits.target)
0.9987...
Methods
fit(X[, y]) Fit estimator to data.



transform(X) Apply feature map to X.
__init__(kernel=’rbf’, gamma=None, coef0=None, degree=None, kernel_params=None,

n_components=100, random_state=None)
fit(X, y=None)
Fit estimator to data.
Samples a subset of training points, computes kernel on these and computes normalization matrix.
Parameters
X [array-like, shape=(n_samples, n_feature)] Training data.
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
transform(X)
Apply feature map to X.
Computes an approximate feature map using the kernel between some training points and X.
Parameters
X [array-like, shape=(n_samples, n_features)] Data to transform.
Returns

X_transformed [array, shape=(n_samples, n_components)] Transformed data.
Examples using sklearn.kernel_approximation.Nystroem
6.19.3 sklearn.kernel_approximation.RBFSampler
class sklearn.kernel_approximation.RBFSampler(gamma=1.0, n_components=100, ran-

dom_state=None)
Approximates feature map of an RBF kernel by Monte Carlo approximation of its Fourier transform.
It implements a variant of Random Kitchen Sinks.[1]
Parameters
gamma [float] Parameter of RBF kernel: exp(-gamma * x^2)
n_components [int] Number of Monte Carlo samples per original feature. Equals the dimen-
sionality of the computed feature space.
Notes
See “Random Features for Large-Scale Kernel Machines” by A. Rahimi and Benjamin Recht.
[1] “Weighted Sums of Random Kitchen Sinks: Replacing minimization with randomization in learning” by A.
Rahimi and Benjamin Recht. (http://people.eecs.berkeley.edu/~brecht/papers/08.rah.rec.nips.pdf)
Examples
>>> from sklearn.kernel_approximation import RBFSampler

>>> X = [[0, 0], [1, 1], [1, 0], [0, 1]]
>>> y = [0, 0, 1, 1]
>>> rbf_feature = RBFSampler(gamma=1, random_state=1)
>>> X_features = rbf_feature.fit_transform(X)
>>> clf = SGDClassifier(max_iter=5, tol=1e-3)
...
power_t=0.5, random_state=None, shuffle=True, tol=0.001,
1.0

Methods

transform(X) Apply the approximate feature map to X.
__init__(gamma=1.0, n_components=100, random_state=None)

fit(X, y=None)
Samples random projection according to n_features.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Training data, where
n_samples in the number of samples and n_features is the number of features.
Returns
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
transform(X)
Apply the approximate feature map to X.

Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] New data, where n_samples
in the number of samples and n_features is the number of features.
Returns
Examples using sklearn.kernel_approximation.RBFSampler
6.19.4 sklearn.kernel_approximation.SkewedChi2Sampler
class sklearn.kernel_approximation.SkewedChi2Sampler(skewedness=1.0,
n_components=100, ran-
dom_state=None)
Approximates feature map of the “skewed chi-squared” kernel by Monte Carlo approximation of its Fourier
transform.
Parameters
skewedness [float] “skewedness” parameter of the kernel. Needs to be cross-validated.
n_components [int] number of Monte Carlo samples per original feature. Equals the dimen-
sionality of the computed feature space.
See also:
AdditiveChi2Sampler A different approach for approximating an additive variant of the chi squared ker-
nel.
sklearn.metrics.pairwise.chi2_kernel The exact chi squared kernel.
References
See “Random Fourier Approximations for Skewed Multiplicative Histogram Kernels” by Fuxin Li, Catalin
Ionescu and Cristian Sminchisescu.
Examples
>>> from sklearn.kernel_approximation import SkewedChi2Sampler

>>> X = [[0, 0], [1, 1], [1, 0], [0, 1]]
>>> y = [0, 0, 1, 1]
>>> chi2_feature = SkewedChi2Sampler(skewedness=.01,
... n_components=10,
... random_state=0)

>>> X_features = chi2_feature.fit_transform(X, y)

>>> clf = SGDClassifier(max_iter=10, tol=1e-3)
1.0
Methods

transform(X) Apply the approximate feature map to X.
__init__(skewedness=1.0, n_components=100, random_state=None)

fit(X, y=None)
Samples random projection according to n_features.
Parameters
X [array-like, shape (n_samples, n_features)] Training data, where n_samples in the number
Returns
Parameters
Returns
Parameters
Returns


of a nested object.
Returns
self
transform(X)
Apply the approximate feature map to X.
Parameters
X [array-like, shape (n_samples, n_features)] New data, where n_samples in the number of
samples and n_features is the number of features. All values of X must be strictly greater
than “-skewedness”.
Returns
6.20 sklearn.kernel_ridge Kernel Ridge Regression
Module sklearn.kernel_ridge implements kernel ridge regression.

User guide: See the Kernel ridge regression section for further details.
kernel_ridge.KernelRidge([alpha, kernel, . . . ]) Kernel ridge regression.
6.20.1 sklearn.kernel_ridge.KernelRidge
class sklearn.kernel_ridge.KernelRidge(alpha=1, kernel=’linear’, gamma=None, degree=3,

coef0=1, kernel_params=None)
Kernel ridge regression.
Kernel ridge regression (KRR) combines ridge regression (linear least squares with l2-norm regularization) with
the kernel trick. It thus learns a linear function in the space induced by the respective kernel and the data. For
non-linear kernels, this corresponds to a non-linear function in the original space.
The form of the model learned by KRR is identical to support vector regression (SVR). However, different loss
functions are used: KRR uses squared error loss while support vector regression uses epsilon-insensitive loss,
both combined with l2 regularization. In contrast to SVR, fitting a KRR model can be done in closed-form and
is typically faster for medium-sized datasets. On the other hand, the learned model is non-sparse and thus slower
than SVR, which learns a sparse model for epsilon > 0, at prediction-time.
This estimator has built-in support for multi-variate regression (i.e., when y is a 2d-array of shape [n_samples,
n_targets]).
Parameters
alpha [{float, array-like}, shape = [n_targets]] Small positive values of alpha improve the con-
ditioning of the problem and reduce the variance of the estimates. Alpha corresponds to

(2*C)^-1 in other linear models such as LogisticRegression or LinearSVC. If an array is

passed, penalties are assumed to be specific to the targets. Hence they must correspond in
number.
kernel [string or callable, default=”linear”] Kernel mapping used internally. A callable should
accept two arguments and the keyword arguments passed to this object as kernel_params,
and should return a floating point number. Set to “precomputed” in order to pass a precom-
puted kernel matrix to the estimator methods instead of samples.
gamma [float, default=None] Gamma parameter for the RBF, laplacian, polynomial, exponen-
tial chi2 and sigmoid kernels. Interpretation of the default value is left to the kernel; see the
documentation for sklearn.metrics.pairwise. Ignored by other kernels.
degree [float, default=3] Degree of the polynomial kernel. Ignored by other kernels.
coef0 [float, default=1] Zero coefficient for polynomial and sigmoid kernels. Ignored by other
kernels.
kernel_params [mapping of string to any, optional] Additional parameters (keyword argu-
ments) for kernel function passed as callable object.
Attributes
dual_coef_ [array, shape = [n_samples] or [n_samples, n_targets]] Representation of weight
vector(s) in kernel space
X_fit_ [{array-like, sparse matrix}, shape = [n_samples, n_features]] Training data, which is
also required for prediction. If kernel == “precomputed” this is instead the precomputed
training matrix, shape = [n_samples, n_samples].
See also:
sklearn.linear_model.Ridge Linear ridge regression.

sklearn.svm.SVR Support Vector Regression implemented using libsvm.
References
• Kevin P. Murphy “Machine Learning: A Probabilistic Perspective”, The MIT Press chapter 14.4.3, pp.
492-493
Examples
>>> from sklearn.kernel_ridge import KernelRidge

>>> n_samples, n_features = 10, 5
>>> y = rng.randn(n_samples)
>>> X = rng.randn(n_samples, n_features)
>>> clf = KernelRidge(alpha=1.0)
>>> clf.fit(X, y)
KernelRidge(alpha=1.0, coef0=1, degree=3, gamma=None, kernel='linear',
kernel_params=None)
6.20. sklearn.kernel_ridge Kernel Ridge Regression 1767

Methods
fit(X[, y, sample_weight]) Fit Kernel Ridge regression model

predict(X) Predict using the kernel ridge model
diction.
__init__(alpha=1, kernel=’linear’, gamma=None, degree=3, coef0=1, kernel_params=None)

Fit Kernel Ridge regression model
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] Training data. If ker-
nel == “precomputed” this is instead a precomputed kernel matrix, shape = [n_samples,
n_samples].
y [array-like, shape = [n_samples] or [n_samples, n_targets]] Target values
sample_weight [float or array-like of shape [n_samples]] Individual weights for each sam-
ple, ignored if None is passed.
Returns
Parameters
Returns
predict(X)
Predict using the kernel ridge model
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] Samples. If kernel
== “precomputed” this is instead a precomputed kernel matrix, shape = [n_samples,
n_samples_fitted], where n_samples_fitted is the number of samples used in the fitting
for this estimator.
Returns
C [array, shape = [n_samples] or [n_samples, n_targets]] Returns predicted values.

Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.kernel_ridge.KernelRidge
• Comparison of kernel ridge regression and SVR

6.21 sklearn.linear_model: Generalized Linear Models
The sklearn.linear_model module implements generalized linear models. It includes Ridge regression,
Bayesian Regression, Lasso and Elastic Net estimators computed with Least Angle Regression and coordinate de-
scent. It also implements Stochastic Gradient Descent related algorithms.
User guide: See the Generalized Linear Models section for further details.
linear_model.ARDRegression([n_iter, tol, . . . ]) Bayesian ARD regression.

linear_model.BayesianRidge([n_iter, tol, . . . ]) Bayesian ridge regression
linear_model.ElasticNet([alpha, l1_ratio, . . . ]) Linear regression with combined L1 and L2 priors as regu-
larizer.
linear_model.ElasticNetCV ([l1_ratio, eps, . . . ]) Elastic Net model with iterative fitting along a regulariza-
tion path.
linear_model.HuberRegressor([epsilon, . . . ]) Linear regression model that is robust to outliers.
linear_model.Lars([fit_intercept, verbose, . . . ]) Least Angle Regression model a.k.a.
linear_model.LarsCV ([fit_intercept, . . . ]) Cross-validated Least Angle Regression model.
linear_model.Lasso([alpha, fit_intercept, . . . ]) Linear Model trained with L1 prior as regularizer (aka the
Lasso)
linear_model.LassoCV ([eps, n_alphas, . . . ]) Lasso linear model with iterative fitting along a regulariza-
tion path.
linear_model.LassoLars([alpha, . . . ]) Lasso model fit with Least Angle Regression a.k.a.
linear_model.LassoLarsCV ([fit_intercept, . . . ]) Cross-validated Lasso, using the LARS algorithm.
6.21. sklearn.linear_model: Generalized Linear Models 1769


linear_model.LassoLarsIC([criterion, . . . ]) Lasso model fit with Lars using BIC or AIC for model se-
lection
linear_model.LinearRegression([. . . ]) Ordinary least squares Linear Regression.
linear_model.LogisticRegression([penalty, Logistic Regression (aka logit, MaxEnt) classifier.
. . . ])
linear_model.LogisticRegressionCV ([Cs, Logistic Regression CV (aka logit, MaxEnt) classifier.
. . . ])
linear_model.MultiTaskLasso([alpha, . . . ]) Multi-task Lasso model trained with L1/L2 mixed-norm as
regularizer.
linear_model.MultiTaskElasticNet([alpha, Multi-task ElasticNet model trained with L1/L2 mixed-
. . . ]) norm as regularizer
linear_model.MultiTaskLassoCV ([eps, . . . ]) Multi-task Lasso model trained with L1/L2 mixed-norm as
regularizer.
linear_model.MultiTaskElasticNetCV ([. . . ]) Multi-task L1/L2 ElasticNet with built-in cross-validation.
linear_model.OrthogonalMatchingPursuit([. . . ]) Orthogonal Matching Pursuit model (OMP)
linear_model.OrthogonalMatchingPursuitCV ([.Cross-validated
. . ]) Orthogonal Matching Pursuit model
(OMP).
linear_model.PassiveAggressiveClassifier([.Passive . . ]) Aggressive Classifier
linear_model.PassiveAggressiveRegressor([C,Passive Aggressive Regressor
. . . ])
linear_model.Perceptron([penalty, alpha, . . . ]) Read more in the User Guide.
linear_model.RANSACRegressor([. . . ]) RANSAC (RANdom SAmple Consensus) algorithm.
linear_model.Ridge([alpha, fit_intercept, . . . ]) Linear least squares with l2 regularization.
linear_model.RidgeClassifier([alpha, . . . ]) Classifier using Ridge regression.
linear_model.RidgeClassifierCV ([alphas, Ridge classifier with built-in cross-validation.
. . . ])
linear_model.RidgeCV ([alphas, . . . ]) Ridge regression with built-in cross-validation.
linear_model.SGDClassifier([loss, penalty, Linear classifiers (SVM, logistic regression, a.o.) with
. . . ]) SGD training.
linear_model.SGDRegressor([loss, penalty, . . . ]) Linear model fitted by minimizing a regularized empirical
loss with SGD
linear_model.TheilSenRegressor([. . . ]) Theil-Sen Estimator: robust multivariate regression model.
6.21.1 sklearn.linear_model.ARDRegression
class sklearn.linear_model.ARDRegression(n_iter=300, tol=0.001, alpha_1=1e-

06, alpha_2=1e-06, lambda_1=1e-06,
lambda_2=1e-06, compute_score=False, thresh-
old_lambda=10000.0, fit_intercept=True, normal-
ize=False, copy_X=True, verbose=False)
Bayesian ARD regression.
Fit the weights of a regression model, using an ARD prior. The weights of the regression model are assumed
to be in Gaussian distributions. Also estimate the parameters lambda (precisions of the distributions of the
weights) and alpha (precision of the distribution of the noise). The estimation is done by an iterative procedures
(Evidence Maximization)
Parameters
n_iter [int, optional] Maximum number of iterations. Default is 300
tol [float, optional] Stop the algorithm if w has converged. Default is 1.e-3.

alpha_1 [float, optional] Hyper-parameter : shape parameter for the Gamma distribution prior
over the alpha parameter. Default is 1.e-6.
alpha_2 [float, optional] Hyper-parameter : inverse scale parameter (rate parameter) for the
Gamma distribution prior over the alpha parameter. Default is 1.e-6.
lambda_1 [float, optional] Hyper-parameter : shape parameter for the Gamma distribution prior
over the lambda parameter. Default is 1.e-6.
lambda_2 [float, optional] Hyper-parameter : inverse scale parameter (rate parameter) for the
Gamma distribution prior over the lambda parameter. Default is 1.e-6.
compute_score [boolean, optional] If True, compute the objective function at each step of the
model. Default is False.
threshold_lambda [float, optional] threshold for removing (pruning) weights with high preci-
sion from the computation. Default is 1.e+4.
fit_intercept [boolean, optional] whether to calculate the intercept for this model. If set to
Default is True.
copy_X [boolean, optional, default True.] If True, X will be copied; else, it may be overwritten.
verbose [boolean, optional, default False] Verbose mode when fitting the model.
Attributes
coef_ [array, shape = (n_features)] Coefficients of the regression model (mean of distribution)
alpha_ [float] estimated precision of the noise.
lambda_ [array, shape = (n_features)] estimated precisions of the weights.
sigma_ [array, shape = (n_features, n_features)] estimated variance-covariance matrix of the
weights
scores_ [float] if computed, value of the objective function (to be maximized)
Notes
For an example, see examples/linear_model/plot_ard.py.
References
D. J. C. MacKay, Bayesian nonlinear modeling for the prediction competition, ASHRAE Transactions, 1994.
R. Salakhutdinov, Lecture notes on Statistical Machine Learning, http://www.utstat.toronto.edu/~rsalakhu/
sta4273/notes/Lecture2.pdf#page=15 Their beta is our self.alpha_ Their alpha is our self.lambda_
ARD is a little different than the slide: only dimensions/features for which self.lambda_ < self.
threshold_lambda are kept and the rest are discarded.

Examples

>>> clf = linear_model.ARDRegression()
>>> clf.fit([[0,0], [1, 1], [2, 2]], [0, 1, 2])
...
ARDRegression(alpha_1=1e-06, alpha_2=1e-06, compute_score=False,
copy_X=True, fit_intercept=True, lambda_1=1e-06, lambda_2=1e-06,
n_iter=300, normalize=False, threshold_lambda=10000.0, tol=0.001,
verbose=False)
array([1.])
Methods
fit(X, y) Fit the ARDRegression model according to the given

training data and parameters.
predict(X[, return_std]) Predict using the linear model.
diction.
__init__(n_iter=300, tol=0.001, alpha_1=1e-06, alpha_2=1e-06, lambda_1=1e-06, lambda_2=1e-

06, compute_score=False, threshold_lambda=10000.0, fit_intercept=True, normal-
ize=False, copy_X=True, verbose=False)
fit(X, y)
Fit the ARDRegression model according to the given training data and parameters.
Iterative procedure to maximize the evidence
Parameters
X [array-like, shape = [n_samples, n_features]] Training vector, where n_samples in the
y [array, shape = [n_samples]] Target values (integers). Will be cast to X’s dtype if necessary
Returns
Parameters
Returns
Predict using the linear model.
In addition to the mean of the predictive distribution, also its standard deviation can be returned.

Parameters
tion.
Returns
y_mean [array, shape = (n_samples,)] Mean of predictive distribution of query points.
y_std [array, shape = (n_samples,)] Standard deviation of predictive distribution of query
points.
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.linear_model.ARDRegression
6.21.2 sklearn.linear_model.BayesianRidge
class sklearn.linear_model.BayesianRidge(n_iter=300, tol=0.001, alpha_1=1e-06,

alpha_2=1e-06, lambda_1=1e-06, lambda_2=1e-
06, compute_score=False, fit_intercept=True,
normalize=False, copy_X=True, verbose=False)
Bayesian ridge regression
Fit a Bayesian ridge model and optimize the regularization parameters lambda (precision of the weights) and
alpha (precision of the noise).


Parameters
n_iter [int, optional] Maximum number of iterations. Default is 300.
tol [float, optional] Stop the algorithm if w has converged. Default is 1.e-3.
alpha_1 [float, optional] Hyper-parameter : shape parameter for the Gamma distribution prior
over the alpha parameter. Default is 1.e-6
alpha_2 [float, optional] Hyper-parameter : inverse scale parameter (rate parameter) for the
Gamma distribution prior over the alpha parameter. Default is 1.e-6.
lambda_1 [float, optional] Hyper-parameter : shape parameter for the Gamma distribution prior
over the lambda parameter. Default is 1.e-6.
lambda_2 [float, optional] Hyper-parameter : inverse scale parameter (rate parameter) for the
Gamma distribution prior over the lambda parameter. Default is 1.e-6
compute_score [boolean, optional] If True, compute the objective function at each step of the
model. Default is False
fit_intercept [boolean, optional] whether to calculate the intercept for this model. If set to
Default is True.
Attributes
coef_ [array, shape = (n_features)] Coefficients of the regression model (mean of distribution)
alpha_ [float] estimated precision of the noise.
lambda_ [float] estimated precision of the weights.
sigma_ [array, shape = (n_features, n_features)] estimated variance-covariance matrix of the
weights
scores_ [float] if computed, value of the objective function (to be maximized)
Notes
For an example, see examples/linear_model/plot_bayesian_ridge.py.
References
D. J. C. MacKay, Bayesian Interpolation, Computation and Neural Systems, Vol. 4, No. 3, 1992.
R. Salakhutdinov, Lecture notes on Statistical Machine Learning, http://www.utstat.toronto.edu/~rsalakhu/
sta4273/notes/Lecture2.pdf#page=15 Their beta is our self.alpha_ Their alpha is our self.lambda_

Examples

>>> clf = linear_model.BayesianRidge()
>>> clf.fit([[0,0], [1, 1], [2, 2]], [0, 1, 2])
...
BayesianRidge(alpha_1=1e-06, alpha_2=1e-06, compute_score=False,
copy_X=True, fit_intercept=True, lambda_1=1e-06, lambda_2=1e-06,
n_iter=300, normalize=False, tol=0.001, verbose=False)
array([1.])
Methods
fit(X, y[, sample_weight]) Fit the model

predict(X[, return_std]) Predict using the linear model.
diction.
__init__(n_iter=300, tol=0.001, alpha_1=1e-06, alpha_2=1e-06, lambda_1=1e-06, lambda_2=1e-

06, compute_score=False, fit_intercept=True, normalize=False, copy_X=True, ver-
bose=False)
Fit the model
Parameters
X [numpy array of shape [n_samples,n_features]] Training data
y [numpy array of shape [n_samples]] Target values. Will be cast to X’s dtype if necessary
sample_weight [numpy array of shape [n_samples]] Individual weights for each sample
New in version 0.20: parameter sample_weight support to BayesianRidge.
Returns
Parameters
Returns
Predict using the linear model.
In addition to the mean of the predictive distribution, also its standard deviation can be returned.
Parameters


tion.
Returns
y_mean [array, shape = (n_samples,)] Mean of predictive distribution of query points.
y_std [array, shape = (n_samples,)] Standard deviation of predictive distribution of query
points.
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.linear_model.BayesianRidge

6.21.3 sklearn.linear_model.ElasticNet
class sklearn.linear_model.ElasticNet(alpha=1.0, l1_ratio=0.5, fit_intercept=True, nor-

malize=False, precompute=False, max_iter=1000,
copy_X=True, tol=0.0001, warm_start=False, posi-
tive=False, random_state=None, selection=’cyclic’)
Linear regression with combined L1 and L2 priors as regularizer.
Minimizes the objective function:

1 / (2 * n_samples) * ||y - Xw||^2_2

+ 0.5 * alpha * (1 - l1_ratio) * ||w||^2_2
If you are interested in controlling the L1 and L2 penalty separately, keep in mind that this is equivalent to:
a * L1 + b * L2
where:
alpha = a + b and l1_ratio = a / (a + b)
The parameter l1_ratio corresponds to alpha in the glmnet R package while alpha corresponds to the lambda
parameter in glmnet. Specifically, l1_ratio = 1 is the lasso penalty. Currently, l1_ratio <= 0.01 is not reliable,
unless you supply your own sequence of alpha.
Parameters
alpha [float, optional] Constant that multiplies the penalty terms. Defaults to 1.0. See the
notes for the exact mathematical meaning of this parameter.‘‘alpha = 0‘‘ is equivalent to an
ordinary least square, solved by the LinearRegression object. For numerical reasons,
using alpha = 0 with the Lasso object is not advised. Given this, you should use the
LinearRegression object.
l1_ratio [float] The ElasticNet mixing parameter, with 0 <= l1_ratio <= 1. For
l1_ratio = 0 the penalty is an L2 penalty. For l1_ratio = 1 it is an L1 penalty.
For 0 < l1_ratio < 1, the penalty is a combination of L1 and L2.
fit_intercept [bool] Whether the intercept should be estimated or not. If False, the data is
assumed to be already centered.
precompute [True | False | array-like] Whether to use a precomputed Gram matrix to speed up
calculations. The Gram matrix can also be passed as argument. For sparse input this option
is always True to preserve sparsity.
tol.
warm_start [bool, optional] When set to True, reuse the solution of the previous call to fit as
initialization, otherwise, just erase the previous solution. See the Glossary.

Attributes
tion formula)
sparse_coef_ [scipy.sparse matrix, shape (n_features, 1) | (n_targets, n_features)] sparse
representation of the fitted coef_
n_iter_ [array-like, shape (n_targets,)] number of iterations run by the coordinate descent solver
to reach the specified tolerance.
See also:
ElasticNetCV Elastic net model with best model selection by cross-validation.

SGDRegressor implements elastic net regression with incremental training.
SGDClassifier implements logistic regression with elastic net penalty
(SGDClassifier(loss="log", penalty="elasticnet")).
Notes
Examples
>>> from sklearn.linear_model import ElasticNet


>>> regr = ElasticNet(random_state=0)
>>> regr.fit(X, y)
ElasticNet(alpha=1.0, copy_X=True, fit_intercept=True, l1_ratio=0.5,
random_state=0, selection='cyclic', tol=0.0001, warm_start=False)
[18.83816048 64.55968825]
1.451...
>>> print(regr.predict([[0, 0]]))
[1.451...]
Methods
fit(X, y[, check_input]) Fit model with coordinate descent.



diction.
__init__(alpha=1.0, l1_ratio=0.5, fit_intercept=True, normalize=False, precompute=False,

max_iter=1000, copy_X=True, tol=0.0001, warm_start=False, positive=False, ran-
dom_state=None, selection=’cyclic’)
fit(X, y, check_input=True)
Fit model with coordinate descent.
Parameters
X [ndarray or scipy.sparse matrix, (n_samples, n_features)] Data
y [ndarray, shape (n_samples,) or (n_samples, n_targets)] Target. Will be cast to X’s dtype
if necessary
check_input [boolean, (default=True)] Allow to bypass several input checking. Don’t use
this parameter unless you know what you do.
Notes
Coordinate descent is an algorithm that considers each column of data at a time hence it will automatically
convert the X input as a Fortran-contiguous numpy array if necessary.
To avoid memory re-allocation it is advised to allocate the initial data in memory directly using that format.
Parameters
Returns
1 / (2 * n_samples) * ||y - Xw||^2_2
+ 0.5 * alpha * (1 - l1_ratio) * ||w||^2_2

(1 / (2 * n_samples)) * ||Y - XW||^Fro_2

Where:

Parameters
can be sparse.
1e-3
set automatically
passed as argument.
written.
when y.ndim == 1).
Returns
along the path.
alpha.

See also:
Notes

predict(X)
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
sparse_coef_
sparse representation of the fitted coef_
Examples using sklearn.linear_model.ElasticNet

• Train error vs Test error

6.21.4 sklearn.linear_model.HuberRegressor
class sklearn.linear_model.HuberRegressor(epsilon=1.35, max_iter=100, alpha=0.0001,

warm_start=False, fit_intercept=True, tol=1e-05)
Linear regression model that is robust to outliers.
The Huber Regressor optimizes the squared loss for the samples where |(y - X'w) / sigma| <
epsilon and the absolute loss for the samples where |(y - X'w) / sigma| > epsilon, where w
and sigma are parameters to be optimized. The parameter sigma makes sure that if y is scaled up or down by a
certain factor, one does not need to rescale epsilon to achieve the same robustness. Note that this does not take
into account the fact that the different features of X may be of different scales.
This makes sure that the loss function is not heavily influenced by the outliers while not completely ignoring
their effect.
Read more in the User Guide
Parameters
epsilon [float, greater than 1.0, default 1.35] The parameter epsilon controls the number of
samples that should be classified as outliers. The smaller the epsilon, the more robust it is
to outliers.
max_iter [int, default 100] Maximum number of iterations that scipy.optimize.fmin_l_bfgs_b
should run for.
alpha [float, default 0.0001] Regularization parameter.
warm_start [bool, default False] This is useful if the stored attributes of a previously used
model has to be reused. If set to False, then the coefficients will be rewritten for every call
to fit. See the Glossary.
fit_intercept [bool, default True] Whether or not to fit the intercept. This can be set to False if
the data is already centered around the origin.
tol [float, default 1e-5] The iteration will stop when max{|proj g_i | i = 1, ...,
n} <= tol where pg_i is the i-th component of the projected gradient.
Attributes
coef_ [array, shape (n_features,)] Features got by optimizing the Huber loss.
intercept_ [float] Bias.
scale_ [float] The value by which |y - X'w - c| is scaled down.
n_iter_ [int] Number of iterations that fmin_l_bfgs_b has run for.
Changed in version 0.20: In SciPy <= 1.0.0 the number of lbfgs iterations may exceed
max_iter. n_iter_ will now report at most max_iter.
outliers_ [array, shape (n_samples,)] A boolean mask which is set to True where the samples
are identified as outliers.
References
[1], [2]

Examples

>>> from sklearn.linear_model import HuberRegressor, LinearRegression
>>> X, y, coef = make_regression(
... n_samples=200, n_features=2, noise=4.0, coef=True, random_state=0)
>>> X[:4] = np.random.uniform(10, 20, (4, 2))
>>> y[:4] = np.random.uniform(10, 20, 4)
>>> huber = HuberRegressor().fit(X, y)
>>> huber.score(X, y)
-7.284608623514573
>>> huber.predict(X[:1,])
array([806.7200...])
>>> linear = LinearRegression().fit(X, y)
>>> print("True coefficients:", coef)
True coefficients: [20.4923... 34.1698...]
>>> print("Huber coefficients:", huber.coef_)
Huber coefficients: [17.7906... 31.0106...]
>>> print("Linear Regression coefficients:", linear.coef_)
Linear Regression coefficients: [-1.9221... 7.0226...]
Methods
diction.
__init__(epsilon=1.35, max_iter=100, alpha=0.0001, warm_start=False, fit_intercept=True, tol=1e-

05)
Parameters
sample_weight [array-like, shape (n_samples,)] Weight given to each sample.
Returns
self [object]
Parameters

Returns
predict(X)
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.linear_model.HuberRegressor

6.21.5 sklearn.linear_model.Lars
class sklearn.linear_model.Lars(fit_intercept=True, verbose=False, normalize=True, precom-

pute=’auto’, n_nonzero_coefs=500, eps=2.220446049250313e-
16, copy_X=True, fit_path=True, positive=False)
Least Angle Regression model a.k.a. LAR


Parameters
as argument.
n_nonzero_coefs [int, optional] Target number of non-zero coefficients. Use np.inf for no
limit.
fit_path [boolean] If True the full path is stored in the coef_path_ attribute. If you compute
the solution for a large problem or many targets, setting fit_path to False will lead to
a speedup, especially with a small alpha.
want to remove fit_intercept which is set True by default.
Deprecated since version 0.20: The option is broken and deprecated. It will be removed in
v0.22.
Attributes
alphas_ [array, shape (n_alphas + 1,) | list of n_targets such arrays] Maximum of covari-
ances (in absolute value) at each iteration. n_alphas is either n_nonzero_coefs or
n_features, whichever is smaller.
active_ [list, length = n_alphas | list of n_targets such lists] Indices of active variables at the end
of the path.
coef_path_ [array, shape (n_features, n_alphas + 1) | list of n_targets such arrays] The varying
values of the coefficients along the path. It is not present if the fit_path parameter is
False.
coef_ [array, shape (n_features,) or (n_targets, n_features)] Parameter vector (w in the formula-
tion formula).
intercept_ [float | array, shape (n_targets,)] Independent term in decision function.
n_iter_ [array-like or int] The number of iterations taken by lars_path to find the grid of alphas
for each target.
See also:
lars_path, LarsCV , sklearn.decomposition.sparse_encode

Examples

>>> reg = linear_model.Lars(n_nonzero_coefs=1)
>>> reg.fit([[-1, 1], [0, 0], [1, 1]], [-1.1111, 0, -1.1111])
...
Lars(copy_X=True, eps=..., fit_intercept=True, fit_path=True,
n_nonzero_coefs=1, normalize=True, positive=False, precompute='auto',
verbose=False)
[ 0. -1.11...]
Methods
fit(X, y[, Xy]) Fit the model using X, y as training data.

diction.
__init__(fit_intercept=True, verbose=False, normalize=True, precompute=’auto’,

n_nonzero_coefs=500, eps=2.220446049250313e-16, copy_X=True, fit_path=True,
positive=False)
fit(X, y, Xy=None)
Parameters
y [array-like, shape (n_samples,) or (n_samples, n_targets)] Target values.
Xy [array-like, shape (n_samples,) or (n_samples, n_targets), optional] Xy = np.dot(X.T, y)
that can be precomputed. It is useful only when the Gram matrix is precomputed.
Returns
Parameters
Returns
predict(X)
Parameters

Returns
Parameters
Returns
of a nested object.
Returns
self
6.21.6 sklearn.linear_model.Lasso
class sklearn.linear_model.Lasso(alpha=1.0, fit_intercept=True, normalize=False, precom-

pute=False, copy_X=True, max_iter=1000, tol=0.0001,
warm_start=False, positive=False, random_state=None, selec-
tion=’cyclic’)
Linear Model trained with L1 prior as regularizer (aka the Lasso)
Technically the Lasso model is optimizing the same objective function as the Elastic Net with l1_ratio=1.0
(no L2 penalty).
Parameters
alpha [float, optional] Constant that multiplies the L1 term. Defaults to 1.0. alpha = 0 is
equivalent to an ordinary least square, solved by the LinearRegression object. For
numerical reasons, using alpha = 0 with the Lasso object is not advised. Given this,
you should use the LinearRegression object.

fit_intercept [boolean, optional, default True] Whether to calculate the intercept for this model.
If set to False, no intercept will be used in calculations (e.g. data is expected to be already
centered).
precompute [True | False | array-like, default=False] Whether to use a precomputed Gram ma-
trix to speed up calculations. If set to 'auto' let us decide. The Gram matrix can also be
passed as argument. For sparse input this option is always True to preserve sparsity.
tol.
Attributes
tion formula)
sparse_coef_ [scipy.sparse matrix, shape (n_features, 1) | (n_targets, n_features)] sparse
representation of the fitted coef_
n_iter_ [int | array-like, shape (n_targets,)] number of iterations run by the coordinate descent
solver to reach the specified tolerance.
See also:
lars_path, lasso_path, LassoLars, LassoCV , LassoLarsCV , sklearn.decomposition.
sparse_encode
Notes


Examples

>>> clf = linear_model.Lasso(alpha=0.1)
>>> clf.fit([[0,0], [1, 1], [2, 2]], [0, 1, 2])
[0.85 0. ]
0.15...
Methods
fit(X, y[, check_input]) Fit model with coordinate descent.

diction.
__init__(alpha=1.0, fit_intercept=True, normalize=False, precompute=False, copy_X=True,

max_iter=1000, tol=0.0001, warm_start=False, positive=False, random_state=None,
selection=’cyclic’)
fit(X, y, check_input=True)
Fit model with coordinate descent.
Parameters
X [ndarray or scipy.sparse matrix, (n_samples, n_features)] Data
y [ndarray, shape (n_samples,) or (n_samples, n_targets)] Target. Will be cast to X’s dtype
if necessary
Notes
Parameters
Returns


1 / (2 * n_samples) * ||y - Xw||^2_2

+ 0.5 * alpha * (1 - l1_ratio) * ||w||^2_2
(1 / (2 * n_samples)) * ||Y - XW||^Fro_2

Where:

Parameters
can be sparse.
1e-3
set automatically
passed as argument.
written.

when y.ndim == 1).
Returns
along the path.
alpha.
See also:
Notes

predict(X)
Parameters
Returns
Parameters
Returns

of a nested object.
Returns
self
sparse_coef_
Examples using sklearn.linear_model.Lasso

• Lasso on dense and sparse data
6.21.7 sklearn.linear_model.LassoLars
class sklearn.linear_model.LassoLars(alpha=1.0, fit_intercept=True, verbose=False, nor-

malize=True, precompute=’auto’, max_iter=500,
eps=2.220446049250313e-16, copy_X=True,
fit_path=True, positive=False)
Lasso model fit with Least Angle Regression a.k.a. Lars
It is a Linear Model trained with an L1 prior as regularizer.

Parameters
alpha [float] Constant that multiplies the penalty term. Defaults to 1.0. alpha = 0 is equiv-
alent to an ordinary least square, solved by LinearRegression. For numerical reasons,
using alpha = 0 with the LassoLars object is not advised and you should prefer the Lin-
earRegression object.

as argument.
fit_path [boolean] If True the full path is stored in the coef_path_ attribute. If you compute
the solution for a large problem or many targets, setting fit_path to False will lead to
a speedup, especially with a small alpha.
model coefficients will not converge to the ordinary-least-squares solution for small values
congruence with the solution of the coordinate descent Lasso estimator.
Attributes
alphas_ [array, shape (n_alphas + 1,) | list of n_targets such arrays] Maximum of covariances
(in absolute value) at each iteration. n_alphas is either max_iter, n_features, or
the number of nodes in the path with correlation greater than alpha, whichever is smaller.
active_ [list, length = n_alphas | list of n_targets such lists] Indices of active variables at the end
of the path.
coef_path_ [array, shape (n_features, n_alphas + 1) or list] If a list is passed it’s expected to be
one of n_targets such arrays. The varying values of the coefficients along the path. It is not
present if the fit_path parameter is False.
coef_ [array, shape (n_features,) or (n_targets, n_features)] Parameter vector (w in the formula-
tion formula).
intercept_ [float | array, shape (n_targets,)] Independent term in decision function.
n_iter_ [array-like or int.] The number of iterations taken by lars_path to find the grid of alphas
for each target.
See also:
lars_path, lasso_path, Lasso, LassoCV , LassoLarsCV , LassoLarsIC, sklearn.
decomposition.sparse_encode
Examples

>>> reg = linear_model.LassoLars(alpha=0.01)
>>> reg.fit([[-1, 1], [0, 0], [1, 1]], [-1, 0, -1])
...
LassoLars(alpha=0.01, copy_X=True, eps=..., fit_intercept=True,
fit_path=True, max_iter=500, normalize=True, positive=False,
precompute='auto', verbose=False)

[ 0. -0.963257...]
Methods
fit(X, y[, Xy]) Fit the model using X, y as training data.

diction.
__init__(alpha=1.0, fit_intercept=True, verbose=False, normalize=True, precompute=’auto’,

max_iter=500, eps=2.220446049250313e-16, copy_X=True, fit_path=True, posi-
tive=False)
fit(X, y, Xy=None)
Parameters
y [array-like, shape (n_samples,) or (n_samples, n_targets)] Target values.
Xy [array-like, shape (n_samples,) or (n_samples, n_targets), optional] Xy = np.dot(X.T, y)
that can be precomputed. It is useful only when the Gram matrix is precomputed.
Returns
Parameters
Returns
predict(X)
Parameters
Returns

Parameters
Returns
of a nested object.
Returns
self
6.21.8 sklearn.linear_model.LinearRegression
class sklearn.linear_model.LinearRegression(fit_intercept=True, normalize=False,

copy_X=True, n_jobs=None)
Ordinary least squares Linear Regression.
Parameters
fit_intercept [boolean, optional, default True] whether to calculate the intercept for this model.
If set to False, no intercept will be used in calculations (e.g. data is expected to be already
centered).
This will only provide speedup for n_targets > 1 and sufficient large problems. None means
Attributes
coef_ [array, shape (n_features, ) or (n_targets, n_features)] Estimated coefficients for the linear
regression problem. If multiple targets are passed during the fit (y 2D), this is a 2D array of
shape (n_targets, n_features), while if only one target is passed, this is a 1D array of length
n_features.
intercept_ [array] Independent term in the linear model.

Notes
From the implementation point of view, this is just plain Ordinary Least Squares (scipy.linalg.lstsq) wrapped as
a predictor object.
Examples

>>> X = np.array([[1, 1], [1, 2], [2, 2], [2, 3]])
>>> # y = 1 * x_0 + 2 * x_1 + 3
>>> y = np.dot(X, np.array([1, 2])) + 3
>>> reg = LinearRegression().fit(X, y)
>>> reg.score(X, y)
1.0
>>> reg.coef_
array([1., 2.])
>>> reg.intercept_
3.0000...
>>> reg.predict(np.array([[3, 5]]))
array([16.])
Methods
fit(X, y[, sample_weight]) Fit linear model.

diction.
__init__(fit_intercept=True, normalize=False, copy_X=True, n_jobs=None)

Fit linear model.
Parameters
X [array-like or sparse matrix, shape (n_samples, n_features)] Training data
y [array_like, shape (n_samples, n_targets)] Target values. Will be cast to X’s dtype if nec-
essary
sample_weight [numpy array of shape [n_samples]] Individual weights for each sample
New in version 0.17: parameter sample_weight support to LinearRegression.
Returns
Parameters

Returns
predict(X)
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.linear_model.LinearRegression
• Isotonic Regression
• Ordinary Least Squares and Ridge Regression Variance
• Logistic function
• Sparsity Example: Fitting only features 1 and 2

6.21.9 sklearn.linear_model.LogisticRegression
class sklearn.linear_model.LogisticRegression(penalty=’l2’, dual=False, tol=0.0001,

C=1.0, fit_intercept=True, inter-
cept_scaling=1, class_weight=None,
random_state=None, solver=’warn’,
max_iter=100, multi_class=’warn’, ver-
bose=0, warm_start=False, n_jobs=None)
Logistic Regression (aka logit, MaxEnt) classifier.
In the multiclass case, the training algorithm uses the one-vs-rest (OvR) scheme if the ‘multi_class’ option is
set to ‘ovr’, and uses the cross- entropy loss if the ‘multi_class’ option is set to ‘multinomial’. (Currently the
‘multinomial’ option is supported only by the ‘lbfgs’, ‘sag’ and ‘newton-cg’ solvers.)
This class implements regularized logistic regression using the ‘liblinear’ library, ‘newton-cg’, ‘sag’ and ‘lbfgs’
solvers. It can handle both dense and sparse input. Use C-ordered arrays or CSR matrices containing 64-bit
floats for optimal performance; any other input format will be converted (and copied).
The ‘newton-cg’, ‘sag’, and ‘lbfgs’ solvers support only L2 regularization with primal formulation. The ‘liblin-
ear’ solver supports both L1 and L2 regularization, with a dual formulation only for the L2 penalty.
Parameters
penalty [str, ‘l1’ or ‘l2’, default: ‘l2’] Used to specify the norm used in the penalization. The
‘newton-cg’, ‘sag’ and ‘lbfgs’ solvers support only l2 penalties.
New in version 0.19: l1 penalty with SAGA solver (allowing ‘multinomial’ + L1)
dual [bool, default: False] Dual or primal formulation. Dual formulation is only implemented
for l2 penalty with liblinear solver. Prefer dual=False when n_samples > n_features.
tol [float, default: 1e-4] Tolerance for stopping criteria.
C [float, default: 1.0] Inverse of regularization strength; must be a positive float. Like in support
vector machines, smaller values specify stronger regularization.
fit_intercept [bool, default: True] Specifies if a constant (a.k.a. bias or intercept) should be
added to the decision function.

class_weight [dict or ‘balanced’, default: None] Weights associated with classes in the form
bincount(y)).
New in version 0.17: class_weight=’balanced’
random_state [int, RandomState instance or None, optional, default: None] The seed of the
instance used by np.random. Used when solver == ‘sag’ or ‘liblinear’.
solver [str, {‘newton-cg’, ‘lbfgs’, ‘liblinear’, ‘sag’, ‘saga’}, default: ‘liblinear’.] Algorithm to
use in the optimization problem.
• For small datasets, ‘liblinear’ is a good choice, whereas ‘sag’ and ‘saga’ are faster for
large ones.
• For multiclass problems, only ‘newton-cg’, ‘sag’, ‘saga’ and ‘lbfgs’ handle multinomial
loss; ‘liblinear’ is limited to one-versus-rest schemes.
• ‘newton-cg’, ‘lbfgs’ and ‘sag’ only handle L2 penalty, whereas ‘liblinear’ and ‘saga’
handle L1 penalty.
Note that ‘sag’ and ‘saga’ fast convergence is only guaranteed on features with approxi-
mately the same scale. You can preprocess the data with a scaler from sklearn.preprocessing.
Changed in version 0.20: Default will change from ‘liblinear’ to ‘lbfgs’ in 0.22.
max_iter [int, default: 100] Useful only for the newton-cg, sag and lbfgs solvers. Maximum
number of iterations taken for the solvers to converge.
verbose [int, default: 0] For the liblinear and lbfgs solvers set verbose to any positive number
for verbosity.
fit as initialization, otherwise, just erase the previous solution. Useless for liblinear solver.
See the Glossary.
New in version 0.17: warm_start to support lbfgs, newton-cg, sag, saga solvers.

n_jobs [int or None, optional (default=None)] Number of CPU cores used when parallelizing
over classes if multi_class=’ovr’”. This parameter is ignored when the solver is set to
‘liblinear’ regardless of whether ‘multi_class’ is specified or not. None means 1 unless in
a joblib.parallel_backend context. -1 means using all processors. See Glossary
for more details.
Attributes
classes_ [array, shape (n_classes, )] A list of class labels known to the classifier.
coef_ [array, shape (1, n_features) or (n_classes, n_features)] Coefficient of the features in the
decision function.
coef_ is of shape (1, n_features) when the given problem is binary. In particular, when
multi_class=’multinomial’, coef_ corresponds to outcome 1 (True) and -coef_ corresponds
to outcome 0 (False).
intercept_ [array, shape (1,) or (n_classes,)] Intercept (a.k.a. bias) added to the decision func-
tion.
If fit_intercept is set to False, the intercept is set to zero. intercept_ is of shape (1,) when
the given problem is binary. In particular, when multi_class=’multinomial’, intercept_ cor-
responds to outcome 1 (True) and -intercept_ corresponds to outcome 0 (False).
n_iter_ [array, shape (n_classes,) or (1, )] Actual number of iterations for all classes. If binary
or multinomial, it returns only 1 element. For liblinear solver, only the maximum number
of iteration across all classes is given.
Changed in version 0.20: In SciPy <= 1.0.0 the number of lbfgs iterations may exceed
max_iter. n_iter_ will now report at most max_iter.
See also:
SGDClassifier incrementally trained logistic regression (when given the parameter loss="log").
LogisticRegressionCV Logistic regression with built-in cross validation
Notes
The underlying C implementation uses a random number generator to select features when fitting the model.
It is thus not uncommon, to have slightly different results for the same input data. If that happens, try with a
smaller tol parameter.
Predict output may not match that of standalone liblinear in certain cases. See differences from liblinear in the
narrative documentation.
References
LIBLINEAR – A Library for Large Linear Classification https://www.csie.ntu.edu.tw/~cjlin/liblinear/

SAG – Mark Schmidt, Nicolas Le Roux, and Francis Bach Minimizing Finite Sums with the Stochastic Av-
erage Gradient https://hal.inria.fr/hal-00860051/document
SAGA – Defazio, A., Bach F. & Lacoste-Julien S. (2014). SAGA: A Fast Incremental Gradient Method With
Support for Non-Strongly Convex Composite Objectives https://arxiv.org/abs/1407.0202
Hsiang-Fu Yu, Fang-Lan Huang, Chih-Jen Lin (2011). Dual coordinate descent methods for logistic re-
gression and maximum entropy models. Machine Learning 85(1-2):41-75. https://www.csie.ntu.edu.tw/
~cjlin/papers/maxent_dual.pdf

Examples

>>> clf = LogisticRegression(random_state=0, solver='lbfgs',
... multi_class='multinomial').fit(X, y)
>>> clf.predict(X[:2, :])
array([0, 0])
>>> clf.predict_proba(X[:2, :])
array([[9.8...e-01, 1.8...e-02, 1.4...e-08],
[9.7...e-01, 2.8...e-02, ...e-08]])
>>> clf.score(X, y)
0.97...
Methods

predict_log_proba(X) Log of probability estimates.
labels.
__init__(penalty=’l2’, dual=False, tol=0.0001, C=1.0, fit_intercept=True, intercept_scaling=1,

class_weight=None, random_state=None, solver=’warn’, max_iter=100,
multi_class=’warn’, verbose=0, warm_start=False, n_jobs=None)
Parameters
Returns
densify()
Returns

self [estimator]
Parameters
New in version 0.17: sample_weight support to LogisticRegression.
Returns
self [object]
Parameters
Returns
predict(X)
Parameters
Returns
Parameters
Returns
T [array-like, shape = [n_samples, n_classes]] Returns the log-probability of the sample for
predict_proba(X)
For a multi_class problem, if multi_class is set to be “multinomial” the softmax function is used to find
the predicted probability of each class. Else use a one-vs-rest approach, i.e calculate the probability of
each class assuming it to be positive using the logistic function. and normalize these values across all the
classes.

Parameters
Returns
T [array-like, shape = [n_samples, n_classes]] Returns the probability of the sample for each
class in the model, where classes are ordered as they are in self.classes_.
Parameters
Returns
of a nested object.
Returns
self
sparsify()
Returns
self [estimator]
Notes
densify.
Examples using sklearn.linear_model.LogisticRegression


• Logistic function
• Logistic Regression 3-class Classifier
6.21.10 sklearn.linear_model.MultiTaskLasso
class sklearn.linear_model.MultiTaskLasso(alpha=1.0, fit_intercept=True, normalize=False,

copy_X=True, max_iter=1000, tol=0.0001,
warm_start=False, random_state=None, selec-
tion=’cyclic’)
Multi-task Lasso model trained with L1/L2 mixed-norm as regularizer.
Where:

Parameters
alpha [float, optional] Constant that multiplies the L1/L2 term. Defaults to 1.0


tol.
leads to significantly faster convergence especially when tol is higher than 1e-4
Attributes
intercept_ [array, shape (n_tasks,)] independent term in decision function.
tolerance.
See also:
MultiTaskLasso Multi-task L1/L2 Lasso with built-in cross-validation
Lasso, MultiTaskElasticNet
Notes

Examples

>>> clf = linear_model.MultiTaskLasso(alpha=0.1)
>>> clf.fit([[0,0], [1, 1], [2, 2]], [[0, 0], [1, 1], [2, 2]])
MultiTaskLasso(alpha=0.1, copy_X=True, fit_intercept=True, max_iter=1000,
normalize=False, random_state=None, selection='cyclic', tol=0.0001,
warm_start=False)
[[0.89393398 0. ]
[0.89393398 0. ]]

[0.10606602 0.10606602]
Methods
fit(X, y) Fit MultiTaskElasticNet model with coordinate descent

diction.
__init__(alpha=1.0, fit_intercept=True, normalize=False, copy_X=True, max_iter=1000,

tol=0.0001, warm_start=False, random_state=None, selection=’cyclic’)
fit(X, y)
Fit MultiTaskElasticNet model with coordinate descent
Parameters
X [ndarray, shape (n_samples, n_features)] Data
y [ndarray, shape (n_samples, n_tasks)] Target. Will be cast to X’s dtype if necessary
Notes
Parameters
Returns
path(X, y, l1_ratio=0.5, eps=0.001, n_alphas=100, alphas=None, precompute=’auto’, Xy=None,
check_input=True, **params)
1 / (2 * n_samples) * ||y - Xw||^2_2

+ 0.5 * alpha * (1 - l1_ratio) * ||w||^2_2

(1 / (2 * n_samples)) * ||Y - XW||^Fro_2

Where:

Parameters
can be sparse.
1e-3
set automatically
passed as argument.
written.
when y.ndim == 1).
Returns
along the path.

alpha.
See also:
Notes

predict(X)
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
sparse_coef_

Examples using sklearn.linear_model.MultiTaskLasso
6.21.11 sklearn.linear_model.MultiTaskElasticNet
class sklearn.linear_model.MultiTaskElasticNet(alpha=1.0, l1_ratio=0.5,

fit_intercept=True, normalize=False,
copy_X=True, max_iter=1000, tol=0.0001,
warm_start=False, random_state=None,
selection=’cyclic’)
Multi-task ElasticNet model trained with L1/L2 mixed-norm as regularizer
The optimization objective for MultiTaskElasticNet is:
(1 / (2 * n_samples)) * ||Y - XW||_Fro^2

Where:
||W||_21 = sum_i sqrt(sum_j w_ij ^ 2)

Parameters
alpha [float, optional] Constant that multiplies the L1/L2 term. Defaults to 1.0
l1_ratio [float] The ElasticNet mixing parameter, with 0 < l1_ratio <= 1. For l1_ratio = 1 the
penalty is an L1/L2 penalty. For l1_ratio = 0 it is an L2 penalty. For 0 < l1_ratio <
1, the penalty is a combination of L1/L2 and L2.
tol.

Attributes
coef_ [array, shape (n_tasks, n_features)] Parameter vector (W in the cost function formula). If
a 1D y is passed in at fit (non multi-task usage), coef_ is then a 1D array. Note that coef_
stores the transpose of W, W.T.
tolerance.
See also:
MultiTaskElasticNet Multi-task L1/L2 ElasticNet with built-in cross-validation.
ElasticNet, MultiTaskLasso
Notes

Examples

>>> clf = linear_model.MultiTaskElasticNet(alpha=0.1)
>>> clf.fit([[0,0], [1, 1], [2, 2]], [[0, 0], [1, 1], [2, 2]])
...
MultiTaskElasticNet(alpha=0.1, copy_X=True, fit_intercept=True,
l1_ratio=0.5, max_iter=1000, normalize=False, random_state=None,
[[0.45663524 0.45612256]
[0.45663524 0.45612256]]
[0.0872422 0.0872422]
Methods
fit(X, y) Fit MultiTaskElasticNet model with coordinate descent

diction.

__init__(alpha=1.0, l1_ratio=0.5, fit_intercept=True, normalize=False, copy_X=True,

max_iter=1000, tol=0.0001, warm_start=False, random_state=None, selection=’cyclic’)
fit(X, y)
Fit MultiTaskElasticNet model with coordinate descent
Parameters
X [ndarray, shape (n_samples, n_features)] Data
y [ndarray, shape (n_samples, n_tasks)] Target. Will be cast to X’s dtype if necessary
Notes
Parameters
Returns
path(X, y, l1_ratio=0.5, eps=0.001, n_alphas=100, alphas=None, precompute=’auto’, Xy=None,
1 / (2 * n_samples) * ||y - Xw||^2_2
+ 0.5 * alpha * (1 - l1_ratio) * ||w||^2_2

(1 / (2 * n_samples)) * ||Y - XW||^Fro_2
Where:

Parameters
can be sparse.


1e-3
set automatically
passed as argument.
written.
when y.ndim == 1).
Returns
along the path.
alpha.
See also:
Notes

predict(X)
Parameters


Returns
Parameters
Returns
of a nested object.
Returns
self
sparse_coef_
6.21.12 sklearn.linear_model.OrthogonalMatchingPursuit
class sklearn.linear_model.OrthogonalMatchingPursuit(n_nonzero_coefs=None,
tol=None, fit_intercept=True, nor-
malize=True, precompute=’auto’)
Orthogonal Matching Pursuit model (OMP)
Parameters
n_nonzero_coefs [int, optional] Desired number of non-zero entries in the solution. If None
(by default) this value is set to 10% of n_features.
tol [float, optional] Maximum norm of the residual. If not None, overrides n_nonzero_coefs.
fit_intercept [boolean, optional] whether to calculate the intercept for this model. If set to false,
no intercept will be used in calculations (e.g. data is expected to be already centered).


precompute [{True, False, ‘auto’}, default ‘auto’] Whether to use a precomputed Gram and
Xy matrix to speed up calculations. Improves performance when n_targets or n_samples is
very large. Note that if you already have such matrices, you can pass them directly to the fit
method.
Attributes
coef_ [array, shape (n_features,) or (n_targets, n_features)] parameter vector (w in the formula)
intercept_ [float or array, shape (n_targets,)] independent term in decision function.
n_iter_ [int or array-like] Number of active features across every target.
See also:
orthogonal_mp, orthogonal_mp_gram, lars_path, Lars, LassoLars, decomposition.
sparse_encode, OrthogonalMatchingPursuitCV
Notes
Orthogonal matching pursuit was introduced in G. Mallat, Z. Zhang, Matching pursuits with time-frequency
dictionaries, IEEE Transactions on Signal Processing, Vol. 41, No. 12. (December 1993), pp. 3397-3415.
(http://blanche.polytechnique.fr/~mallat/papiers/MallatPursuit93.pdf)
This implementation is based on Rubinstein, R., Zibulevsky, M. and Elad, M., Efficient Implementation of
the K-SVD Algorithm using Batch Orthogonal Matching Pursuit Technical Report - CS Technion, April 2008.
http://www.cs.technion.ac.il/~ronrubin/Publications/KSVD-OMP-v2.pdf
Examples
>>> from sklearn.linear_model import OrthogonalMatchingPursuit

>>> X, y = make_regression(noise=4, random_state=0)
>>> reg = OrthogonalMatchingPursuit().fit(X, y)
>>> reg.score(X, y)
0.9991...
array([-78.3854...])
Methods

diction.
__init__(n_nonzero_coefs=None, tol=None, fit_intercept=True, normalize=True, precom-

pute=’auto’)

fit(X, y)
Parameters
y [array-like, shape (n_samples,) or (n_samples, n_targets)] Target values. Will be cast to
X’s dtype if necessary
Returns
Parameters
Returns
predict(X)
Parameters
Returns
Parameters
Returns
of a nested object.
Returns

self
Examples using sklearn.linear_model.OrthogonalMatchingPursuit
6.21.13 sklearn.linear_model.PassiveAggressiveClassifier
class sklearn.linear_model.PassiveAggressiveClassifier(C=1.0, fit_intercept=True,

max_iter=None, tol=None,
early_stopping=False,
n_iter_no_change=5,
shuffle=True, verbose=0,
loss=’hinge’, n_jobs=None,
random_state=None,
warm_start=False,
class_weight=None, aver-
age=False, n_iter=None)
Passive Aggressive Classifier
Parameters
C [float] Maximum step size (regularization). Defaults to 1.0.
fit_intercept [bool, default=False] Whether the intercept should be estimated or not. If False,
the data is assumed to be already centered.
max_iter [int, optional] The maximum number of passes over the training data (aka epochs). It
only impacts the behavior in the fit method, and not the partial_fit. Defaults to 5. Defaults
to 1000 from 0.21, or if tol is not None.
tol [float or None, optional] The stopping criterion. If it is not None, the iterations will stop
when (loss > previous_loss - tol). Defaults to None. Defaults to 1e-3 from 0.21.
early_stopping [bool, default=False] Whether to use early stopping to terminate training when
validation. score is not improving. If set to True, it will automatically set aside a fraction of
training data as validation and terminate training when validation score is not improving by
at least tol for n_iter_no_change consecutive epochs.
validation_fraction [float, default=0.1] The proportion of training data to set aside as validation
set for early stopping. Must be between 0 and 1. Only used if early_stopping is True.
n_iter_no_change [int, default=5] Number of iterations with no improvement to wait before
early stopping.
shuffle [bool, default=True] Whether or not the training data should be shuffled after each
epoch.

verbose [integer, optional] The verbosity level

loss [string, optional] The loss function to be used: hinge: equivalent to PA-I in the reference
paper. squared_hinge: equivalent to PA-II in the reference paper.
n_jobs [int or None, optional (default=None)] The number of CPUs to use to do the OVA (One
Versus All, for multi-class problems) computation. None means 1 unless in a joblib.
details.
random_state [int, RandomState instance or None, optional, default=None] The seed of the
Repeatedly calling fit or partial_fit when warm_start is True can result in a different solution
than when calling fit a single time because of the way the data is shuffled.
class_weight [dict, {class_label: weight} or “balanced” or None, optional] Preset for the
class_weight fit parameter.
Weights associated with classes. If not given, all classes are supposed to have weight one.
bincount(y))
New in version 0.17: parameter class_weight to automatically weight samples.
average [bool or int, optional] When set to True, computes the averaged SGD weights and
stores the result in the coef_ attribute. If set to an int greater than 1, averaging will begin
once the total number of samples seen reaches average. So average=10 will begin averaging
after seeing 10 samples.
New in version 0.19: parameter average to use weights averaging in SGD
n_iter [int, optional] The number of passes over the training data (aka epochs). Defaults to
None. Deprecated, will be removed in 0.21.
Changed in version 0.19: Deprecated
Attributes
coef_ [array, shape = [1, n_features] if n_classes == 2 else [n_classes, n_features]] Weights
assigned to the features.
intercept_ [array, shape = [1] if n_classes == 2 else [n_classes]] Constants in decision function.
n_iter_ [int] The actual number of iterations to reach the stopping criterion. For multiclass fits,
it is the maximum over every binary fit.
See also:
SGDClassifier, Perceptron

References
Online Passive-Aggressive Algorithms <http://jmlr.csail.mit.edu/papers/volume7/crammer06a/crammer06a.

pdf> K. Crammer, O. Dekel, J. Keshat, S. Shalev-Shwartz, Y. Singer - JMLR (2006)
Examples
>>> from sklearn.linear_model import PassiveAggressiveClassifier

>>> X, y = make_classification(n_features=4, random_state=0)

>>> clf = PassiveAggressiveClassifier(max_iter=1000, random_state=0,
... tol=1e-3)
>>> clf.fit(X, y)
PassiveAggressiveClassifier(C=1.0, average=False, class_weight=None,
early_stopping=False, fit_intercept=True, loss='hinge',
max_iter=1000, n_iter=None, n_iter_no_change=5, n_jobs=None,
random_state=0, shuffle=True, tol=0.001,
[[-0.6543424 1.54603022 1.35361642 0.22199435]]
[0.63310933]
[1]
Methods

fit(X, y[, coef_init, intercept_init]) Fit linear model with Passive Aggressive algorithm.
partial_fit(X, y[, classes]) Fit linear model with Passive Aggressive algorithm.
labels.
set_params(*args, **kwargs)
__init__(C=1.0, fit_intercept=True, max_iter=None, tol=None, early_stopping=False, vali-

dation_fraction=0.1, n_iter_no_change=5, shuffle=True, verbose=0, loss=’hinge’,
n_jobs=None, random_state=None, warm_start=False, class_weight=None, aver-
Parameters

Returns
densify()
Returns
self [estimator]
fit(X, y, coef_init=None, intercept_init=None)
Fit linear model with Passive Aggressive algorithm.
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] Training data
y [numpy array of shape [n_samples]] Target values
coef_init [array, shape = [n_classes,n_features]] The initial coefficients to warm-start the
optimization.
intercept_init [array, shape = [n_classes]] The initial intercept to warm-start the optimiza-
tion.
Returns
Parameters
Returns
loss_function
DEPRECATED: Attribute loss_function was deprecated in version 0.19 and will be removed in 0.21. Use
loss_function_ instead
partial_fit(X, y, classes=None)
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] Subset of the training data
y [numpy array of shape [n_samples]] Subset of the target values
classes [array, shape = [n_classes]] Classes across all calls to partial_fit. Can be obtained by
via np.unique(y_all), where y_all is the target vector of the entire dataset. This argument
is required for the first call to partial_fit and can be omitted in the subsequent calls. Note
that y doesn’t need to contain all labels in classes.

Returns
predict(X)
Parameters
Returns
Parameters
Returns
sparsify()
Returns
self [estimator]
Notes
densify.
Examples using sklearn.linear_model.PassiveAggressiveClassifier


6.21.14 sklearn.linear_model.PassiveAggressiveRegressor
class sklearn.linear_model.PassiveAggressiveRegressor(C=1.0, fit_intercept=True,

max_iter=None, tol=None,
early_stopping=False,
n_iter_no_change=5, shuf-
fle=True, verbose=0,
loss=’epsilon_insensitive’, ep-
silon=0.1, random_state=None,
warm_start=False, aver-
Passive Aggressive Regressor
Parameters
C [float] Maximum step size (regularization). Defaults to 1.0.
assumed to be already centered. Defaults to True.
early stopping.
shuffle [bool, default=True] Whether or not the training data should be shuffled after each
epoch.
loss [string, optional] The loss function to be used: epsilon_insensitive: equivalent to PA-I in
the reference paper. squared_epsilon_insensitive: equivalent to PA-II in the reference paper.
epsilon [float] If the difference between the current prediction and the correct label is below
this threshold, the model is not updated.

random_state [int, RandomState instance or None, optional, default=None] The seed of the
than when calling fit a single time because of the way the data is shuffled.
stores the result in the coef_ attribute. If set to an int greater than 1, averaging will begin
once the total number of samples seen reaches average. So average=10 will begin averaging
after seeing 10 samples.
New in version 0.19: parameter average to use weights averaging in SGD
Attributes
n_iter_ [int] The actual number of iterations to reach the stopping criterion.
See also:
SGDRegressor
References
Online Passive-Aggressive Algorithms <http://jmlr.csail.mit.edu/papers/volume7/crammer06a/crammer06a.

pdf> K. Crammer, O. Dekel, J. Keshat, S. Shalev-Shwartz, Y. Singer - JMLR (2006)
Examples
>>> from sklearn.linear_model import PassiveAggressiveRegressor


>>> regr = PassiveAggressiveRegressor(max_iter=100, random_state=0,
... tol=1e-3)
>>> regr.fit(X, y)
PassiveAggressiveRegressor(C=1.0, average=False, early_stopping=False,
epsilon=0.1, fit_intercept=True, loss='epsilon_insensitive',
max_iter=100, n_iter=None, n_iter_no_change=5,
random_state=0, shuffle=True, tol=0.001,
[20.48736655 34.18818427 67.59122734 87.94731329]

[-0.02306214]
[-0.02306214]
Methods

fit(X, y[, coef_init, intercept_init]) Fit linear model with Passive Aggressive algorithm.
partial_fit(X, y) Fit linear model with Passive Aggressive algorithm.
diction.
__init__(C=1.0, fit_intercept=True, max_iter=None, tol=None, early_stopping=False,

validation_fraction=0.1, n_iter_no_change=5, shuffle=True, verbose=0,
loss=’epsilon_insensitive’, epsilon=0.1, random_state=None, warm_start=False, av-
erage=False, n_iter=None)
densify()
Returns
self [estimator]
fit(X, y, coef_init=None, intercept_init=None)
Parameters
coef_init [array, shape = [n_features]] The initial coefficients to warm-start the optimization.
intercept_init [array, shape = [1]] The initial intercept to warm-start the optimization.
Returns
Parameters
Returns


partial_fit(X, y)
Parameters
X [{array-like, sparse matrix}, shape = [n_samples, n_features]] Subset of training data
y [numpy array of shape [n_samples]] Subset of target values
Returns
predict(X)
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)]
Returns
array, shape (n_samples,) Predicted target values per element in X.
Parameters
Returns
sparsify()
Returns
self [estimator]
Notes

densify.
6.21.15 sklearn.linear_model.Perceptron
class sklearn.linear_model.Perceptron(penalty=None, alpha=0.0001, fit_intercept=True,

max_iter=None, tol=None, shuffle=True, ver-
bose=0, eta0=1.0, n_jobs=None, random_state=0,
early_stopping=False, validation_fraction=0.1,
n_iter_no_change=5, class_weight=None,
warm_start=False, n_iter=None)
Parameters
penalty [None, ‘l2’ or ‘l1’ or ‘elasticnet’] The penalty (aka regularization term) to be used.
Defaults to None.
alpha [float] Constant that multiplies the regularization term if regularization is used. Defaults
to 0.0001
shuffle [bool, optional, default True] Whether or not the training data should be shuffled after
each epoch.
eta0 [double] Constant by which the updates are multiplied. Defaults to 1.
details.

early stopping.
bincount(y))
Attributes
See also:
SGDClassifier
Notes
Perceptron is a classification algorithm which shares the same underlying implementation with
SGDClassifier. In fact, Perceptron() is equivalent to SGDClassifier(loss=”perceptron”, eta0=1,
learning_rate=”constant”, penalty=None).
References
https://en.wikipedia.org/wiki/Perceptron and references therein.
Examples

>>> from sklearn.linear_model import Perceptron
>>> clf = Perceptron(tol=1e-3, random_state=0)
>>> clf.fit(X, y)

Perceptron(alpha=0.0001, class_weight=None, early_stopping=False, eta0=1.0,

fit_intercept=True, max_iter=None, n_iter=None, n_iter_no_change=5,
n_jobs=None, penalty=None, random_state=0, shuffle=True, tol=0.001,
>>> clf.score(X, y)
0.946...
Methods

fit(X, y[, coef_init, intercept_init, . . . ]) Fit linear model with Stochastic Gradient Descent.
partial_fit(X, y[, classes, sample_weight]) Perform one epoch of stochastic gradient descent on
given samples.
labels.
__init__(penalty=None, alpha=0.0001, fit_intercept=True, max_iter=None, tol=None, shuf-

fle=True, verbose=0, eta0=1.0, n_jobs=None, random_state=0, early_stopping=False,
validation_fraction=0.1, n_iter_no_change=5, class_weight=None, warm_start=False,
n_iter=None)
Parameters
Returns
densify()
Returns
self [estimator]
fit(X, y, coef_init=None, intercept_init=None, sample_weight=None)
Fit linear model with Stochastic Gradient Descent.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Training data

y [numpy array, shape (n_samples,)] Target values

coef_init [array, shape (n_classes, n_features)] The initial coefficients to warm-start the op-
timization.
intercept_init [array, shape (n_classes,)] The initial intercept to warm-start the optimiza-
tion.
sample_weight [array-like, shape (n_samples,), optional] Weights applied to individual
samples. If not provided, uniform weights are assumed. These weights will be multiplied
with class_weight (passed through the constructor) if class_weight is specified
Returns
Parameters
Returns
loss_function
partial_fit(X, y, classes=None, sample_weight=None)
Perform one epoch of stochastic gradient descent on given samples.
Internally, this method uses max_iter = 1. Therefore, it is not guaranteed that a minimum of the cost
function is reached after calling it once. Matters such as objective convergence and early stopping should
be handled by the user.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Subset of the training data
y [numpy array, shape (n_samples,)] Subset of the target values
classes [array, shape (n_classes,)] Classes across all calls to partial_fit. Can be obtained by
samples. If not provided, uniform weights are assumed.
Returns
predict(X)
Parameters
Returns


Parameters
Returns
sparsify()
Returns
self [estimator]
Notes
densify.
Examples using sklearn.linear_model.Perceptron

6.21.16 sklearn.linear_model.RANSACRegressor
class sklearn.linear_model.RANSACRegressor(base_estimator=None, min_samples=None,

residual_threshold=None, is_data_valid=None,
is_model_valid=None, max_trials=100,
max_skips=inf, stop_n_inliers=inf,
stop_score=inf, stop_probability=0.99,
loss=’absolute_loss’, random_state=None)
RANSAC (RANdom SAmple Consensus) algorithm.

RANSAC is an iterative algorithm for the robust estimation of parameters from a subset of inliers from the
complete data set. More information can be found in the general documentation of linear models.
A detailed description of the algorithm can be found in the documentation of the linear_model sub-package.
Parameters
base_estimator [object, optional] Base estimator object which implements the following meth-
ods:
• fit(X, y): Fit model to given training data and target values.
• score(X, y): Returns the mean accuracy on the given test data, which is used for the stop
criterion defined by stop_score. Additionally, the score is used to decide which of two
equally large consensus sets is chosen as the better one.
• predict(X): Returns predicted values using the linear model, which is used to compute
residual error using loss function.
If base_estimator is None, then base_estimator=sklearn.linear_model.
LinearRegression() is used for target values of dtype float.
Note that the current implementation only supports regression estimators.
min_samples [int (>= 1) or float ([0, 1]), optional] Minimum number of samples chosen ran-
domly from original data. Treated as an absolute number of samples for min_samples
>= 1, treated as a relative number ceil(min_samples * X.shape[0]) for min_samples < 1.
This is typically chosen as the minimal number of samples necessary to estimate the given
base_estimator. By default a sklearn.linear_model.LinearRegression() es-
timator is assumed and min_samples is chosen as X.shape[1] + 1.
residual_threshold [float, optional] Maximum residual for a data sample to be classified as an
inlier. By default the threshold is chosen as the MAD (median absolute deviation) of the
target values y.
is_data_valid [callable, optional] This function is called with the randomly selected data before
the model is fitted to it: is_data_valid(X, y). If its return value is False the current randomly
chosen sub-sample is skipped.
is_model_valid [callable, optional] This function is called with the estimated model and the
randomly selected data: is_model_valid(model, X, y). If its return value is False the current
randomly chosen sub-sample is skipped. Rejecting samples with this function is computa-
tionally costlier than with is_data_valid. is_model_valid should therefore only be used if
the estimated model is needed for making the rejection decision.
max_trials [int, optional] Maximum number of iterations for random sample selection.
max_skips [int, optional] Maximum number of iterations that can be skipped due to finding
zero inliers or invalid data defined by is_data_valid or invalid models defined by
is_model_valid.
stop_n_inliers [int, optional] Stop iteration if at least this number of inliers are found.
stop_score [float, optional] Stop iteration if score is greater equal than this threshold.
stop_probability [float in range [0, 1], optional] RANSAC iteration stops if at least one outlier-
free set of the training data is sampled in RANSAC. This requires to generate at least N
samples (iterations):

N >= log(1 - probability) / log(1 - e**m)
where the probability (confidence) is typically set to high value such as 0.99 (the default)
and e is the current fraction of inliers w.r.t. the total number of samples.
loss [string, callable, optional, default “absolute_loss”] String inputs, “absolute_loss” and
“squared_loss” are supported which find the absolute loss and squared loss per sample re-
spectively.
If loss is a callable, then it should be a function that takes two arrays as inputs, the true
and predicted value and returns a 1-D array with the i-th value of the array corresponding to
the loss on X[i].
If the loss on a sample is greater than the residual_threshold, then this sample is
classified as an outlier.
random_state [int, RandomState instance or None, optional, default None] The generator used
to initialize the centers. If int, random_state is the seed used by the random number gener-
ator; If RandomState instance, random_state is the random number generator; If None, the
random number generator is the RandomState instance used by np.random.
Attributes
estimator_ [object] Best fitted model (copy of the base_estimator object).
n_trials_ [int] Number of random selection trials until one of the stop criteria is met. It is
always <= max_trials.
inlier_mask_ [bool array of shape [n_samples]] Boolean mask of inliers classified as True.
n_skips_no_inliers_ [int] Number of iterations skipped due to finding zero inliers.
n_skips_invalid_data_ [int] Number of iterations skipped due to invalid data defined by
is_data_valid.
n_skips_invalid_model_ [int] Number of iterations skipped due to an invalid model defined by
is_model_valid.
References
[1], [2], [3]
Examples
>>> from sklearn.linear_model import RANSACRegressor

>>> X, y = make_regression(
... n_samples=200, n_features=2, noise=4.0, random_state=0)
>>> reg = RANSACRegressor(random_state=0).fit(X, y)
>>> reg.score(X, y)
0.9885...
array([-31.9417...])

Methods
fit(X, y[, sample_weight]) Fit estimator using RANSAC algorithm.

predict(X) Predict using the estimated model.
score(X, y) Returns the score of the prediction.
__init__(base_estimator=None, min_samples=None, residual_threshold=None,

is_data_valid=None, is_model_valid=None, max_trials=100, max_skips=inf,
stop_n_inliers=inf, stop_score=inf, stop_probability=0.99, loss=’absolute_loss’, ran-
dom_state=None)
Fit estimator using RANSAC algorithm.
Parameters
X [array-like or sparse matrix, shape [n_samples, n_features]] Training data.
y [array-like, shape = [n_samples] or [n_samples, n_targets]] Target values.
sample_weight [array-like, shape = [n_samples]] Individual weights for each sample raises
error if sample_weight is passed and base_estimator fit method does not support it.
Raises
ValueError If no valid consensus set could be found. This occurs if is_data_valid and
is_model_valid return False for all max_trials randomly chosen sub-samples.
Parameters
Returns
predict(X)
Predict using the estimated model.
This is a wrapper for estimator_.predict(X).
Parameters
X [numpy array of shape [n_samples, n_features]]
Returns
y [array, shape = [n_samples] or [n_samples, n_targets]] Returns predicted values.
score(X, y)
Returns the score of the prediction.
This is a wrapper for estimator_.score(X, y).
Parameters
X [numpy array or sparse matrix of shape [n_samples, n_features]] Training data.

y [array, shape = [n_samples] or [n_samples, n_targets]] Target values.

Returns
z [float] Score of the prediction.
of a nested object.
Returns
self
Examples using sklearn.linear_model.RANSACRegressor

6.21.17 sklearn.linear_model.Ridge
class sklearn.linear_model.Ridge(alpha=1.0, fit_intercept=True, normalize=False,

copy_X=True, max_iter=None, tol=0.001, solver=’auto’,
random_state=None)
Linear least squares with l2 regularization.
Minimizes the objective function:
||y - Xw||^2_2 + alpha * ||w||^2_2
This model solves a regression model where the loss function is the linear least squares function and regulariza-
tion is given by the l2-norm. Also known as Ridge Regression or Tikhonov regularization. This estimator has
built-in support for multi-variate regression (i.e., when y is a 2d-array of shape [n_samples, n_targets]).
Parameters
alpha [{float, array-like}, shape (n_targets)] Regularization strength; must be a positive float.
Regularization improves the conditioning of the problem and reduces the variance of the es-
timates. Larger values specify stronger regularization. Alpha corresponds to C^-1 in other
linear models such as LogisticRegression or LinearSVC. If an array is passed, penalties are
assumed to be specific to the targets. Hence they must correspond in number.

max_iter [int, optional] Maximum number of iterations for conjugate gradient solver. For
‘sparse_cg’ and ‘lsqr’ solvers, the default value is determined by scipy.sparse.linalg. For
‘sag’ solver, the default value is 1000.
tol [float] Precision of the solution.
solver [{‘auto’, ‘svd’, ‘cholesky’, ‘lsqr’, ‘sparse_cg’, ‘sag’, ‘saga’}] Solver to use in the com-
putational routines:
• ‘auto’ chooses the solver automatically based on the type of data.
• ‘svd’ uses a Singular Value Decomposition of X to compute the Ridge coefficients. More
stable for singular matrices than ‘cholesky’.
• ‘cholesky’ uses the standard scipy.linalg.solve function to obtain a closed-form solution.
• ‘sparse_cg’ uses the conjugate gradient solver as found in scipy.sparse.linalg.cg. As an
iterative algorithm, this solver is more appropriate than ‘cholesky’ for large-scale data
(possibility to set tol and max_iter).
• ‘lsqr’ uses the dedicated regularized least-squares routine scipy.sparse.linalg.lsqr. It is the
fastest and uses an iterative procedure.
• ‘sag’ uses a Stochastic Average Gradient descent, and ‘saga’ uses its improved, unbiased
version named SAGA. Both methods also use an iterative procedure, and are often faster
than other solvers when both n_samples and n_features are large. Note that ‘sag’ and
‘saga’ fast convergence is only guaranteed on features with approximately the same scale.
You can preprocess the data with a scaler from sklearn.preprocessing.
All last five solvers support both dense and sparse data. However, only ‘sag’ and ‘saga’
supports sparse input when fit_intercept is True.
instance used by np.random. Used when solver == ‘sag’.
New in version 0.17: random_state to support Stochastic Average Gradient.
Attributes
coef_ [array, shape (n_features,) or (n_targets, n_features)] Weight vector(s).
n_iter_ [array or None, shape (n_targets,)] Actual number of iterations for each target. Avail-
able only for sag and lsqr solvers. Other solvers will return None.
See also:

RidgeCV Ridge regression with built-in cross validation
sklearn.kernel_ridge.KernelRidge Kernel ridge regression combines ridge regression with the
kernel trick

Examples
>>> from sklearn.linear_model import Ridge

>>> y = np.random.randn(n_samples)
>>> X = np.random.randn(n_samples, n_features)
>>> clf = Ridge(alpha=1.0)
>>> clf.fit(X, y)
Ridge(alpha=1.0, copy_X=True, fit_intercept=True, max_iter=None,
normalize=False, random_state=None, solver='auto', tol=0.001)
Methods
fit(X, y[, sample_weight]) Fit Ridge regression model

diction.
__init__(alpha=1.0, fit_intercept=True, normalize=False, copy_X=True, max_iter=None, tol=0.001,

solver=’auto’, random_state=None)
Fit Ridge regression model
Parameters
sample_weight [float or numpy array of shape [n_samples]] Individual weights for each
sample
Returns
Parameters
Returns
predict(X)
Parameters

Returns
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.linear_model.Ridge

• Plot Ridge coefficients as a function of the regularization
• Ordinary Least Squares and Ridge Regression Variance
• Polynomial interpolation
6.21.18 sklearn.linear_model.RidgeClassifier
class sklearn.linear_model.RidgeClassifier(alpha=1.0, fit_intercept=True, normalize=False,

copy_X=True, max_iter=None, tol=0.001,
class_weight=None, solver=’auto’, ran-
dom_state=None)
Classifier using Ridge regression.

Parameters
alpha [float] Regularization strength; must be a positive float. Regularization improves the con-
ditioning of the problem and reduces the variance of the estimates. Larger values specify
stronger regularization. Alpha corresponds to C^-1 in other linear models such as Logisti-
cRegression or LinearSVC.
max_iter [int, optional] Maximum number of iterations for conjugate gradient solver. The
default value is determined by scipy.sparse.linalg.
bincount(y))
• ‘cholesky’ uses the standard scipy.linalg.solve function to obtain a closed-form solution.
• ‘sag’ uses a Stochastic Average Gradient descent, and ‘saga’ uses its unbiased and more
flexible version named SAGA. Both methods use an iterative procedure, and are often
faster than other solvers when both n_samples and n_features are large. Note that ‘sag’
and ‘saga’ fast convergence is only guaranteed on features with approximately the same
scale. You can preprocess the data with a scaler from sklearn.preprocessing.

Attributes
coef_ [array, shape (n_features,) or (n_classes, n_features)] Weight vector(s).
n_iter_ [array or None, shape (n_targets,)] Actual number of iterations for each target. Avail-
able only for sag and lsqr solvers. Other solvers will return None.
See also:

RidgeClassifierCV Ridge classifier with built-in cross validation
Notes
For multi-class classification, n_class classifiers are trained in a one-versus-all approach. Concretely, this is
implemented by taking advantage of the multi-variate response support in Ridge.
Examples

>>> from sklearn.linear_model import RidgeClassifier
>>> clf = RidgeClassifier().fit(X, y)
>>> clf.score(X, y)
0.9595...
Methods

fit(X, y[, sample_weight]) Fit Ridge regression model.
labels.
__init__(alpha=1.0, fit_intercept=True, normalize=False, copy_X=True, max_iter=None, tol=0.001,

class_weight=None, solver=’auto’, random_state=None)
Parameters
Returns

Fit Ridge regression model.
Parameters
X [{array-like, sparse matrix}, shape = [n_samples,n_features]] Training data
y [array-like, shape = [n_samples]] Target values
sample_weight [float or numpy array of shape (n_samples,)] Sample weight.
New in version 0.17: sample_weight support to Classifier.
Returns
Parameters
Returns
predict(X)
Parameters
Returns
Parameters
Returns
of a nested object.

Returns
self
Examples using sklearn.linear_model.RidgeClassifier
6.21.19 sklearn.linear_model.SGDClassifier
class sklearn.linear_model.SGDClassifier(loss=’hinge’, penalty=’l2’, alpha=0.0001,

l1_ratio=0.15, fit_intercept=True, max_iter=None,
tol=None, shuffle=True, verbose=0, epsilon=0.1,
n_jobs=None, random_state=None, learn-
ing_rate=’optimal’, eta0=0.0, power_t=0.5,
n_iter_no_change=5, class_weight=None,
warm_start=False, average=False, n_iter=None)
Linear classifiers (SVM, logistic regression, a.o.) with SGD training.
This estimator implements regularized linear models with stochastic gradient descent (SGD) learning: the gra-
dient of the loss is estimated each sample at a time and the model is updated along the way with a decreasing
strength schedule (aka learning rate). SGD allows minibatch (online/out-of-core) learning, see the partial_fit
method. For best results using the default learning rate schedule, the data should have zero mean and unit
variance.
This implementation works with data represented as dense or sparse arrays of floating point values for the
features. The model it fits can be controlled with the loss parameter; by default, it fits a linear support vector
machine (SVM).
The regularizer is a penalty added to the loss function that shrinks model parameters towards the zero vector
using either the squared euclidean norm L2 or the absolute norm L1 or a combination of both (Elastic Net). If
the parameter update crosses the 0.0 value because of the regularizer, the update is truncated to 0.0 to allow for
learning sparse models and achieve online feature selection.
Parameters
loss [str, default: ‘hinge’] The loss function to be used. Defaults to ‘hinge’, which gives a linear
SVM.
The possible options are ‘hinge’, ‘log’, ‘modified_huber’, ‘squared_hinge’, ‘per-
ceptron’, or a regression loss: ‘squared_loss’, ‘huber’, ‘epsilon_insensitive’, or
‘squared_epsilon_insensitive’.
The ‘log’ loss gives logistic regression, a probabilistic classifier. ‘modified_huber’ is
another smooth loss that brings tolerance to outliers as well as probability estimates.
‘squared_hinge’ is like hinge but is quadratically penalized. ‘perceptron’ is the linear loss
used by the perceptron algorithm. The other losses are designed for regression but can be
useful in classification as well; see SGDRegressor for a description.
penalty [str, ‘none’, ‘l2’, ‘l1’, or ‘elasticnet’] The penalty (aka regularization term) to be used.
Defaults to ‘l2’ which is the standard regularizer for linear SVM models. ‘l1’ and ‘elasticnet’
might bring sparsity to the model (feature selection) not achievable with ‘l2’.
alpha [float] Constant that multiplies the regularization term. Defaults to 0.0001 Also used to
compute learning_rate when set to ‘optimal’.

l1_ratio [float] The Elastic Net mixing parameter, with 0 <= l1_ratio <= 1. l1_ratio=0 corre-
sponds to L2 penalty, l1_ratio=1 to L1. Defaults to 0.15.
shuffle [bool, optional] Whether or not the training data should be shuffled after each epoch.
Defaults to True.
epsilon [float] Epsilon in the epsilon-insensitive loss functions; only if loss is ‘huber’, ‘ep-
silon_insensitive’, or ‘squared_epsilon_insensitive’. For ‘huber’, determines the thresh-
old at which it becomes less important to get the prediction exactly right. For epsilon-
insensitive, any differences between the current prediction and the correct label are ignored
if they are less than this threshold.
details.
learning_rate [string, optional] The learning rate schedule:
‘constant’: eta = eta0
‘optimal’: [default] eta = 1.0 / (alpha * (t + t0)) where t0 is chosen by a heuristic proposed
by Leon Bottou.
‘invscaling’: eta = eta0 / pow(t, power_t)
‘adaptive’: eta = eta0, as long as the training keeps decreasing. Each time
n_iter_no_change consecutive epochs fail to decrease the training loss by tol or fail to
increase validation score by tol if early_stopping is True, the current learning rate is di-
vided by 5.
eta0 [double] The initial learning rate for the ‘constant’, ‘invscaling’ or ‘adaptive’ schedules.
The default value is 0.0 as eta0 is not used by the default schedule ‘optimal’.
power_t [double] The exponent for inverse scaling learning rate [default 0.5].
validation score is not improving. If set to True, it will automatically set aside a fraction of


early stopping.
bincount(y))
than when calling fit a single time because of the way the data is shuffled. If a dynamic
learning rate is used, the learning rate is adapted depending on the number of samples al-
ready seen. Calling fit resets this counter, while partial_fit will result in increasing
the existing counter.
stores the result in the coef_ attribute. If set to an int greater than 1, averaging will be-
gin once the total number of samples seen reaches average. So average=10 will begin
averaging after seeing 10 samples.
Attributes
coef_ [array, shape (1, n_features) if n_classes == 2 else (n_classes, n_features)] Weights as-
signed to the features.
intercept_ [array, shape (1,) if n_classes == 2 else (n_classes,)] Constants in decision function.
loss_function_ [concrete LossFunction]
See also:
sklearn.svm.LinearSVC, LogisticRegression, Perceptron
Examples

>>> X = np.array([[-1, -1], [-2, -1], [1, 1], [2, 1]])
>>> Y = np.array([1, 1, 2, 2])

>>> clf = linear_model.SGDClassifier(max_iter=1000, tol=1e-3)

>>> clf.fit(X, Y)
...

[1]
Methods

partial_fit(X, y[, classes, sample_weight]) Perform one epoch of stochastic gradient descent on
given samples.
labels.
__init__(loss=’hinge’, penalty=’l2’, alpha=0.0001, l1_ratio=0.15, fit_intercept=True,

max_iter=None, tol=None, shuffle=True, verbose=0, epsilon=0.1, n_jobs=None, ran-
dom_state=None, learning_rate=’optimal’, eta0=0.0, power_t=0.5, early_stopping=False,
validation_fraction=0.1, n_iter_no_change=5, class_weight=None, warm_start=False,
average=False, n_iter=None)
Parameters
Returns
densify()
Returns

self [estimator]
Parameters
coef_init [array, shape (n_classes, n_features)] The initial coefficients to warm-start the op-
timization.
intercept_init [array, shape (n_classes,)] The initial intercept to warm-start the optimiza-
tion.
samples. If not provided, uniform weights are assumed. These weights will be multiplied
with class_weight (passed through the constructor) if class_weight is specified
Returns
Parameters
Returns
loss_function
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Subset of the training data
y [numpy array, shape (n_samples,)] Subset of the target values
classes [array, shape (n_classes,)] Classes across all calls to partial_fit. Can be obtained by
Returns

predict(X)
Parameters
Returns
predict_log_proba
This method is only available for log loss and modified Huber loss.
When loss=”modified_huber”, probability estimates may be hard zeros and ones, so taking the logarithm
is not possible.
See predict_proba for details.
Parameters
Returns
T [array-like, shape (n_samples, n_classes)] Returns the log-probability of the sample for
predict_proba
This method is only available for log loss and modified Huber loss.
Multiclass probability estimates are derived from binary (one-vs.-rest) estimates by simple normalization,
as recommended by Zadrozny and Elkan.
Binary probability estimates for loss=”modified_huber” are given by (clip(decision_function(X), -1, 1) +
1) / 2. For other loss functions it is necessary to perform proper probability calibration by wrapping the
classifier with sklearn.calibration.CalibratedClassifierCV instead.
Parameters
Returns
array, shape (n_samples, n_classes) Returns the probability of the sample for each class in
the model, where classes are ordered as they are in self.classes_.
References
Zadrozny and Elkan, “Transforming classifier scores into multiclass probability estimates”, SIGKDD‘02,
http://www.research.ibm.com/people/z/zadrozny/kdd2002-Transf.pdf
The justification for the formula in the loss=”modified_huber” case is in the appendix B in: http://jmlr.
csail.mit.edu/papers/volume2/zhang02c/zhang02c.pdf

Parameters
Returns
sparsify()
Returns
self [estimator]
Notes
densify.
Examples using sklearn.linear_model.SGDClassifier

• SGD: Maximum margin separating hyperplane
• SGD: Weighted samples

6.21.20 sklearn.linear_model.SGDRegressor
class sklearn.linear_model.SGDRegressor(loss=’squared_loss’, penalty=’l2’, alpha=0.0001,

l1_ratio=0.15, fit_intercept=True, max_iter=None,
tol=None, shuffle=True, verbose=0, epsilon=0.1,
random_state=None, learning_rate=’invscaling’,
eta0=0.01, power_t=0.25, early_stopping=False,
validation_fraction=0.1, n_iter_no_change=5,
warm_start=False, average=False, n_iter=None)
Linear model fitted by minimizing a regularized empirical loss with SGD
SGD stands for Stochastic Gradient Descent: the gradient of the loss is estimated each sample at a time and the
model is updated along the way with a decreasing strength schedule (aka learning rate).
The regularizer is a penalty added to the loss function that shrinks model parameters towards the zero vector
using either the squared euclidean norm L2 or the absolute norm L1 or a combination of both (Elastic Net). If
the parameter update crosses the 0.0 value because of the regularizer, the update is truncated to 0.0 to allow for
learning sparse models and achieve online feature selection.
This implementation works with data represented as dense numpy arrays of floating point values for the features.
Parameters
loss [str, default: ‘squared_loss’] The loss function to be used. The possible values are
‘squared_loss’, ‘huber’, ‘epsilon_insensitive’, or ‘squared_epsilon_insensitive’
The ‘squared_loss’ refers to the ordinary least squares fit. ‘huber’ modifies ‘squared_loss’ to
focus less on getting outliers correct by switching from squared to linear loss past a distance
of epsilon. ‘epsilon_insensitive’ ignores errors less than epsilon and is linear past that; this
is the loss function used in SVR. ‘squared_epsilon_insensitive’ is the same but becomes
squared loss past a tolerance of epsilon.
penalty [str, ‘none’, ‘l2’, ‘l1’, or ‘elasticnet’] The penalty (aka regularization term) to be used.
Defaults to ‘l2’ which is the standard regularizer for linear SVM models. ‘l1’ and ‘elasticnet’
might bring sparsity to the model (feature selection) not achievable with ‘l2’.
alpha [float] Constant that multiplies the regularization term. Defaults to 0.0001 Also used to
compute learning_rate when set to ‘optimal’.
l1_ratio [float] The Elastic Net mixing parameter, with 0 <= l1_ratio <= 1. l1_ratio=0 corre-
sponds to L2 penalty, l1_ratio=1 to L1. Defaults to 0.15.
shuffle [bool, optional] Whether or not the training data should be shuffled after each epoch.
Defaults to True.
verbose [integer, optional] The verbosity level.

epsilon [float] Epsilon in the epsilon-insensitive loss functions; only if loss is ‘huber’, ‘ep-
silon_insensitive’, or ‘squared_epsilon_insensitive’. For ‘huber’, determines the thresh-
old at which it becomes less important to get the prediction exactly right. For epsilon-
insensitive, any differences between the current prediction and the correct label are ignored
if they are less than this threshold.
learning_rate [string, optional] The learning rate schedule:
‘constant’: eta = eta0
‘optimal’: eta = 1.0 / (alpha * (t + t0)) where t0 is chosen by a heuristic proposed by Leon
Bottou.
‘invscaling’: [default] eta = eta0 / pow(t, power_t)
‘adaptive’: eta = eta0, as long as the training keeps decreasing. Each time
n_iter_no_change consecutive epochs fail to decrease the training loss by tol or fail to
increase validation score by tol if early_stopping is True, the current learning rate is di-
vided by 5.
eta0 [double] The initial learning rate for the ‘constant’, ‘invscaling’ or ‘adaptive’ schedules.
The default value is 0.0 as eta0 is not used by the default schedule ‘optimal’.
power_t [double] The exponent for inverse scaling learning rate [default 0.5].
validation score is not improving. If set to True, it will automatically set aside a fraction of
early stopping.
than when calling fit a single time because of the way the data is shuffled. If a dynamic
learning rate is used, the learning rate is adapted depending on the number of samples al-
ready seen. Calling fit resets this counter, while partial_fit will result in increasing
the existing counter.
stores the result in the coef_ attribute. If set to an int greater than 1, averaging will be-
gin once the total number of samples seen reaches average. So average=10 will begin
averaging after seeing 10 samples.

Attributes
coef_ [array, shape (n_features,)] Weights assigned to the features.
intercept_ [array, shape (1,)] The intercept term.
average_coef_ [array, shape (n_features,)] Averaged weights assigned to the features.
average_intercept_ [array, shape (1,)] The averaged intercept term.
n_iter_ [int] The actual number of iterations to reach the stopping criterion.
See also:
Ridge, ElasticNet, Lasso, sklearn.svm.SVR
Examples

>>> clf = linear_model.SGDRegressor(max_iter=1000, tol=1e-3)
>>> clf.fit(X, y)
...
SGDRegressor(alpha=0.0001, average=False, early_stopping=False,
epsilon=0.1, eta0=0.01, fit_intercept=True, l1_ratio=0.15,
learning_rate='invscaling', loss='squared_loss', max_iter=1000,
n_iter=None, n_iter_no_change=5, penalty='l2', power_t=0.25,
random_state=None, shuffle=True, tol=0.001, validation_fraction=0.1,
verbose=0, warm_start=False)
Methods

partial_fit(X, y[, sample_weight]) Perform one epoch of stochastic gradient descent on
given samples.
diction.
__init__(loss=’squared_loss’, penalty=’l2’, alpha=0.0001, l1_ratio=0.15, fit_intercept=True,

max_iter=None, tol=None, shuffle=True, verbose=0, epsilon=0.1, random_state=None,
learning_rate=’invscaling’, eta0=0.01, power_t=0.25, early_stopping=False, valida-
tion_fraction=0.1, n_iter_no_change=5, warm_start=False, average=False, n_iter=None)

densify()
Returns
self [estimator]
Parameters
coef_init [array, shape (n_features,)] The initial coefficients to warm-start the optimization.
intercept_init [array, shape (1,)] The initial intercept to warm-start the optimization.
samples (1. for unweighted).
Returns
Parameters
Returns
partial_fit(X, y, sample_weight=None)
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Subset of training data
y [numpy array of shape (n_samples,)] Subset of target values
Returns
predict(X)
Parameters


Returns
array, shape (n_samples,) Predicted target values per element in X.
Parameters
Returns
sparsify()
Returns
self [estimator]
Notes
densify.
Examples using sklearn.linear_model.SGDRegressor

6.21.21 sklearn.linear_model.TheilSenRegressor
class sklearn.linear_model.TheilSenRegressor(fit_intercept=True, copy_X=True,

max_subpopulation=10000.0,
n_subsamples=None, max_iter=300,
tol=0.001, random_state=None,
n_jobs=None, verbose=False)
Theil-Sen Estimator: robust multivariate regression model.
The algorithm calculates least square solutions on subsets with size n_subsamples of the samples in X. Any value
of n_subsamples between the number of features and samples leads to an estimator with a compromise between
robustness and efficiency. Since the number of least square solutions is “n_samples choose n_subsamples”, it
can be extremely large and can therefore be limited with max_subpopulation. If this limit is reached, the subsets
are chosen randomly. In a final step, the spatial median (or L1 median) is calculated of all least square solutions.
Parameters
fit_intercept [boolean, optional, default True] Whether to calculate the intercept for this model.
If set to false, no intercept will be used in calculations.
max_subpopulation [int, optional, default 1e4] Instead of computing with a set of cardinality
‘n choose k’, where n is the number of samples and k is the number of subsamples (at
least number of features), consider only a stochastic subpopulation of a given maximal size
if ‘n choose k’ is larger than max_subpopulation. For other than small problem sizes this
parameter will determine memory usage and runtime if n_subsamples is not changed.
n_subsamples [int, optional, default None] Number of samples to calculate the parameters.
This is at least the number of features (plus 1 if fit_intercept=True) and the number of sam-
ples as a maximum. A lower number leads to a higher breakdown point and a low efficiency
while a high number leads to a low breakdown point and a high efficiency. If None, take the
minimum number of subsamples leading to maximal robustness. If n_subsamples is set to
n_samples, Theil-Sen is identical to least squares.
max_iter [int, optional, default 300] Maximum number of iterations for the calculation of spa-
tial median.
tol [float, optional, default 1.e-3] Tolerance when calculating spatial median.
random_state [int, RandomState instance or None, optional, default None] A random number
generator instance to define the state of the random permutations generator. If int, ran-
Attributes
coef_ [array, shape = (n_features)] Coefficients of the regression model (median of distribution).
intercept_ [float] Estimated intercept of regression model.
breakdown_ [float] Approximated breakdown point.
n_iter_ [int] Number of iterations needed for the spatial median.

n_subpopulation_ [int] Number of combinations taken into account from ‘n choose k’, where
n is the number of samples and k is the number of subsamples.
References
• Theil-Sen Estimators in a Multiple Linear Regression Model, 2009 Xin Dang, Hanxiang Peng, Xueqin
Wang and Heping Zhang http://home.olemiss.edu/~xdang/papers/MTSE.pdf
Examples
>>> from sklearn.linear_model import TheilSenRegressor

>>> X, y = make_regression(
... n_samples=200, n_features=2, noise=4.0, random_state=0)
>>> reg = TheilSenRegressor(random_state=0).fit(X, y)
>>> reg.score(X, y)
0.9884...
array([-31.5871...])
Methods
fit(X, y) Fit linear model.

diction.
__init__(fit_intercept=True, copy_X=True, max_subpopulation=10000.0, n_subsamples=None,

max_iter=300, tol=0.001, random_state=None, n_jobs=None, verbose=False)
fit(X, y)
Fit linear model.
Parameters
X [numpy array of shape [n_samples, n_features]] Training data
Returns
Parameters
Returns


predict(X)
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.linear_model.TheilSenRegressor
linear_model.enet_path(X, y[, l1_ratio, . . . ]) Compute elastic net path with coordinate descent
linear_model.lars_path(X, y[, Xy, Gram, . . . ]) Compute Least Angle Regression or Lasso path using
LARS algorithm [1]
linear_model.lasso_path(X, y[, eps, . . . ]) Compute Lasso path with coordinate descent
linear_model.logistic_regression_path(X, Compute a Logistic Regression model for a list of regular-
y) ization parameters.
linear_model.orthogonal_mp(X, y[, . . . ]) Orthogonal Matching Pursuit (OMP)
linear_model.orthogonal_mp_gram(Gram, Xy[, Gram Orthogonal Matching Pursuit (OMP)
. . . ])


linear_model.ridge_regression(X, y, alpha[, Solve the ridge equation by the method of normal equa-
. . . ]) tions.
6.21.22 sklearn.linear_model.enet_path
sklearn.linear_model.enet_path(X, y, l1_ratio=0.5, eps=0.001, n_alphas=100, alphas=None,

precompute=’auto’, Xy=None, copy_X=True, coef_init=None,
verbose=False, return_n_iter=False, positive=False,
1 / (2 * n_samples) * ||y - Xw||^2_2

+ 0.5 * alpha * (1 - l1_ratio) * ||w||^2_2
(1 / (2 * n_samples)) * ||Y - XW||^Fro_2

Where:

Parameters
contiguous data to avoid unnecessary memory duplication. If y is mono-output then X can
be sparse.
l1_ratio [float, optional] float between 0 and 1 passed to elastic net (scaling between l1 and l2
penalties). l1_ratio=1 corresponds to the Lasso
1e-3
alphas [ndarray, optional] List of alphas where to compute the models. If None alphas are set
automatically
as argument.
Xy [array-like, optional] Xy = np.dot(X.T, y) that can be precomputed. It is useful only when
the Gram matrix is precomputed.

when y.ndim == 1).
check_input [bool, default True] Skip input validation checks, including the Gram matrix when
provided assuming there are handled by the caller when check_input=False.
Returns
along the path.
alpha.
See also:
Notes
Examples using sklearn.linear_model.enet_path
6.21.23 sklearn.linear_model.lars_path
sklearn.linear_model.lars_path(X, y, Xy=None, Gram=None, max_iter=500, alpha_min=0,

method=’lar’, copy_X=True, eps=2.220446049250313e-
16, copy_Gram=True, verbose=0, return_path=True, re-
turn_n_iter=False, positive=False)
Compute Least Angle Regression or Lasso path using LARS algorithm [1]
The optimization objective for the case method=’lasso’ is:
in the case of method=’lars’, the objective function is only known in the form of an implicit equation (see
discussion in [1])
Parameters
X [array, shape: (n_samples, n_features)] Input data.

y [array, shape: (n_samples)] Input targets.

Xy [array-like, shape (n_samples,) or (n_samples, n_targets), optional] Xy = np.dot(X.T, y) that
can be precomputed. It is useful only when the Gram matrix is precomputed.
Gram [None, ‘auto’, array, shape: (n_features, n_features), optional] Precomputed Gram ma-
trix (X’ * X), if 'auto', the Gram matrix is precomputed from the given X, if there are
more samples than features.
max_iter [integer, optional (default=500)] Maximum number of iterations to perform, set to
infinity for no limit.
alpha_min [float, optional (default=0)] Minimum correlation along the path. It corresponds to
the regularization parameter alpha parameter in the Lasso.
method [{‘lar’, ‘lasso’}, optional (default=’lar’)] Specifies the returned model. Select 'lar'
for Least Angle Regression, 'lasso' for the Lasso.
copy_X [bool, optional (default=True)] If False, X is overwritten.
eps [float, optional (default=‘‘np.finfo(np.float).eps‘‘)] The machine-precision regularization in
the computation of the Cholesky diagonal factors. Increase this for very ill-conditioned
systems.
copy_Gram [bool, optional (default=True)] If False, Gram is overwritten.
verbose [int (default=0)] Controls output verbosity.
return_path [bool, optional (default=True)] If return_path==True returns the entire path,
else returns only the last point of the path.
return_n_iter [bool, optional (default=False)] Whether to return the number of iterations.
positive [boolean (default=False)] Restrict coefficients to be >= 0. This option is only allowed
with method ‘lasso’. Note that the model coefficients will not converge to the ordinary-
least-squares solution for small values of alpha. Only coefficients up to the smallest alpha
value (alphas_[alphas_ > 0.].min() when fit_path=True) reached by the step-
wise Lars-Lasso algorithm are typically in congruence with the solution of the coordinate
descent lasso_path function.
Returns
alphas [array, shape: [n_alphas + 1]] Maximum of covariances (in absolute value) at each
iteration. n_alphas is either max_iter, n_features or the number of nodes in the
path with alpha >= alpha_min, whichever is smaller.
active [array, shape [n_alphas]] Indices of active variables at the end of the path.
coefs [array, shape (n_features, n_alphas + 1)] Coefficients along the path
See also:
lasso_path, LassoLars, Lars, LassoLarsCV , LarsCV , sklearn.decomposition.
sparse_encode
References
[1], [2], [3]

Examples using sklearn.linear_model.lars_path
6.21.24 sklearn.linear_model.lasso_path
sklearn.linear_model.lasso_path(X, y, eps=0.001, n_alphas=100, alphas=None, precom-

pute=’auto’, Xy=None, copy_X=True, coef_init=None, ver-
bose=False, return_n_iter=False, positive=False, **params)
Where:

Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Training data. Pass directly as
Fortran-contiguous data to avoid unnecessary memory duplication. If y is mono-output then
X can be sparse.
= 1e-3
alphas [ndarray, optional] List of alphas where to compute the models. If None alphas are set
automatically
as argument.
Xy [array-like, optional] Xy = np.dot(X.T, y) that can be precomputed. It is useful only when
the Gram matrix is precomputed.
when y.ndim == 1).


Returns
along the path.
alpha.
n_iters [array-like, shape (n_alphas,)] The number of iterations taken by the coordinate descent
optimizer to reach the specified tolerance for each alpha.
See also:
sparse_encode
Notes

Note that in certain cases, the Lars solver may be significantly faster to implement this functionality. In particu-
lar, linear interpolation can be used to retrieve model coefficients between the values output by lars_path
Examples
>>> X = np.array([[1, 2, 3.1], [2.3, 5.4, 4.3]]).T

>>> y = np.array([1, 2, 3.1])
[[0. 0. 0.46874778]
[0.2159048 0.4425765 0.23689075]]

>>> # same path
[[0. 0. 0.46915237]
[0.2159048 0.4425765 0.23668876]]
Examples using sklearn.linear_model.lasso_path

6.21.25 sklearn.linear_model.logistic_regression_path
sklearn.linear_model.logistic_regression_path(X, y, pos_class=None, Cs=10,

fit_intercept=True, max_iter=100,
tol=0.0001, verbose=0, solver=’lbfgs’,
coef=None, class_weight=None,
dual=False, penalty=’l2’, inter-
cept_scaling=1.0, multi_class=’warn’,
random_state=None, check_input=True,
max_squared_sum=None, sam-
ple_weight=None)
Compute a Logistic Regression model for a list of regularization parameters.
This is an implementation that uses the result of the previous model to speed up computations along the set of
solutions, making it faster than sequentially calling LogisticRegression for the different parameters. Note that
there will be no speedup with liblinear solver, since it does not handle warm-starting.
Parameters
X [array-like or sparse matrix, shape (n_samples, n_features)] Input data.
y [array-like, shape (n_samples,) or (n_samples, n_targets)] Input data, target values.
pos_class [int, None] The class with respect to which we perform a one-vs-all fit. If None, then
it is assumed that the given problem is binary.
Cs [int | array-like, shape (n_cs,)] List of values for the regularization parameter or integer
specifying the number of regularization parameters that should be used. In this case, the
parameters will be chosen in a logarithmic scale between 1e-4 and 1e4.
fit_intercept [bool] Whether to fit an intercept for the model. In this case the shape of the
returned array is (n_cs, n_features + 1).
max_iter [int] Maximum number of iterations for the solver.
tol [float] Stopping criterion. For the newton-cg and lbfgs solvers, the iteration will stop when
max{|g_i | i = 1, ..., n} <= tol where g_i is the i-th component of the
gradient.
verbose [int] For the liblinear and lbfgs solvers set verbose to any positive number for verbosity.
solver [{‘lbfgs’, ‘newton-cg’, ‘liblinear’, ‘sag’, ‘saga’}] Numerical solver to use.
coef [array-like, shape (n_features,), default None] Initialization value for coefficients of logis-
tic regression. Useless for liblinear solver.
bincount(y)).
dual [bool] Dual or primal formulation. Dual formulation is only implemented for l2 penalty
with liblinear solver. Prefer dual=False when n_samples > n_features.
penalty [str, ‘l1’ or ‘l2’] Used to specify the norm used in the penalization. The ‘newton-cg’,
‘sag’ and ‘lbfgs’ solvers support only l2 penalties.

instance used by np.random. Used when solver == ‘sag’ or ‘liblinear’.
check_input [bool, default True] If False, the input arrays X and y will not be checked.
max_squared_sum [float, default None] Maximum squared sum of X over samples. Used only
in SAG solver. If None, it will be computed, going through all the samples. The value should
be precomputed to speed up cross validation.
sample_weight [array-like, shape(n_samples,) optional] Array of weights that are assigned to
individual samples. If not provided, then each sample is given unit weight.
Returns
coefs [ndarray, shape (n_cs, n_features) or (n_cs, n_features + 1)] List of coefficients
for the Logistic Regression model. If fit_intercept is set to True then the sec-
ond dimension will be n_features + 1, where the last item represents the intercept.
For multiclass='multinomial', the shape is (n_classes, n_cs, n_features) or
(n_classes, n_cs, n_features + 1).
Cs [ndarray] Grid of Cs used for cross-validation.
n_iter [array, shape (n_cs,)] Actual number of iteration for each Cs.
Notes
You might get slightly different results with the solver liblinear than with the others since this uses LIBLINEAR
which penalizes the intercept.
Changed in version 0.19: The “copy” parameter was removed.

6.21.26 sklearn.linear_model.orthogonal_mp
sklearn.linear_model.orthogonal_mp(X, y, n_nonzero_coefs=None, tol=None, precom-

pute=False, copy_X=True, return_path=False, re-
turn_n_iter=False)
Orthogonal Matching Pursuit (OMP)
Solves n_targets Orthogonal Matching Pursuit problems. An instance of the problem has the form:
When parametrized by the number of non-zero coefficients using n_nonzero_coefs: argmin ||y - Xgamma||^2
subject to ||gamma||_0 <= n_{nonzero coefs}
When parametrized by error using the parameter tol: argmin ||gamma||_0 subject to ||y - Xgamma||^2 <= tol
Parameters
X [array, shape (n_samples, n_features)] Input data. Columns are assumed to have unit norm.
y [array, shape (n_samples,) or (n_samples, n_targets)] Input targets
n_nonzero_coefs [int] Desired number of non-zero entries in the solution. If None (by default)
this value is set to 10% of n_features.
tol [float] Maximum norm of the residual. If not None, overrides n_nonzero_coefs.
precompute [{True, False, ‘auto’},] Whether to perform precomputations. Improves perfor-
mance when n_targets or n_samples is very large.
copy_X [bool, optional] Whether the design matrix X must be copied by the algorithm. A false
value is only helpful if X is already Fortran-ordered, otherwise a copy is made anyway.
return_path [bool, optional. Default: False] Whether to return every value of the nonzero
coefficients along the forward path. Useful for cross-validation.
return_n_iter [bool, optional default False] Whether or not to return the number of iterations.
Returns
coef [array, shape (n_features,) or (n_features, n_targets)] Coefficients of the OMP solution.
If return_path=True, this contains the whole coefficient path. In this case its shape is
(n_features, n_features) or (n_features, n_targets, n_features) and iterating over the last axis
yields coefficients in increasing order of active features.
n_iters [array-like or int] Number of active features across every target. Returned only if re-
See also:
OrthogonalMatchingPursuit, orthogonal_mp_gram, lars_path, decomposition.
sparse_encode
Notes
Orthogonal matching pursuit was introduced in S. Mallat, Z. Zhang, Matching pursuits with time-frequency

6.21.27 sklearn.linear_model.orthogonal_mp_gram
sklearn.linear_model.orthogonal_mp_gram(Gram, Xy, n_nonzero_coefs=None, tol=None,

norms_squared=None, copy_Gram=True,
copy_Xy=True, return_path=False, re-
turn_n_iter=False)
Gram Orthogonal Matching Pursuit (OMP)
Solves n_targets Orthogonal Matching Pursuit problems using only the Gram matrix X.T * X and the product
X.T * y.
Parameters
Gram [array, shape (n_features, n_features)] Gram matrix of the input data: X.T * X
Xy [array, shape (n_features,) or (n_features, n_targets)] Input targets multiplied by X: X.T * y
n_nonzero_coefs [int] Desired number of non-zero entries in the solution. If None (by default)
this value is set to 10% of n_features.
tol [float] Maximum norm of the residual. If not None, overrides n_nonzero_coefs.
norms_squared [array-like, shape (n_targets,)] Squared L2 norms of the lines of y. Required
if tol is not None.
copy_Gram [bool, optional] Whether the gram matrix must be copied by the algorithm. A false
value is only helpful if it is already Fortran-ordered, otherwise a copy is made anyway.
copy_Xy [bool, optional] Whether the covariance vector Xy must be copied by the algorithm.
If False, it may be overwritten.
return_path [bool, optional. Default: False] Whether to return every value of the nonzero
coefficients along the forward path. Useful for cross-validation.
return_n_iter [bool, optional default False] Whether or not to return the number of iterations.
Returns
coef [array, shape (n_features,) or (n_features, n_targets)] Coefficients of the OMP solution.
If return_path=True, this contains the whole coefficient path. In this case its shape is
(n_features, n_features) or (n_features, n_targets, n_features) and iterating over the last axis
yields coefficients in increasing order of active features.
n_iters [array-like or int] Number of active features across every target. Returned only if re-
See also:
OrthogonalMatchingPursuit, orthogonal_mp, lars_path, decomposition.
sparse_encode
Notes
Orthogonal matching pursuit was introduced in G. Mallat, Z. Zhang, Matching pursuits with time-frequency

6.21.28 sklearn.linear_model.ridge_regression
sklearn.linear_model.ridge_regression(X, y, alpha, sample_weight=None, solver=’auto’,

max_iter=None, tol=0.001, verbose=0, ran-
dom_state=None, return_n_iter=False, re-
turn_intercept=False)
Solve the ridge equation by the method of normal equations.
Parameters
X [{array-like, sparse matrix, LinearOperator},] shape = [n_samples, n_features] Training data
alpha [{float, array-like},] shape = [n_targets] if array-like Regularization strength; must be
a positive float. Regularization improves the conditioning of the problem and reduces the
variance of the estimates. Larger values specify stronger regularization. Alpha corresponds
to C^-1 in other linear models such as LogisticRegression or LinearSVC. If an array is
passed, penalties are assumed to be specific to the targets. Hence they must correspond in
number.
sample_weight [float or numpy array of shape [n_samples]] Individual weights for each sam-
ple. If sample_weight is not None and solver=’auto’, the solver will be set to ‘cholesky’.
• ‘cholesky’ uses the standard scipy.linalg.solve function to obtain a closed-form solution
via a Cholesky decomposition of dot(X.T, X)
• ‘sag’ uses a Stochastic Average Gradient descent, and ‘saga’ uses its improved, unbiased
version named SAGA. Both methods also use an iterative procedure, and are often faster
than other solvers when both n_samples and n_features are large. Note that ‘sag’ and
‘saga’ fast convergence is only guaranteed on features with approximately the same scale.
You can preprocess the data with a scaler from sklearn.preprocessing.
All last five solvers support both dense and sparse data. However, only ‘sag’ and ‘saga’
supports sparse input when‘fit_intercept‘ is True.
max_iter [int, optional] Maximum number of iterations for conjugate gradient solver. For the
‘sparse_cg’ and ‘lsqr’ solvers, the default value is determined by scipy.sparse.linalg. For
‘sag’ and saga solver, the default value is 1000.

verbose [int] Verbosity level. Setting verbose > 0 will display additional information depending
on the solver used.
return_n_iter [boolean, default False] If True, the method also returns n_iter, the actual num-
ber of iteration performed by the solver.
return_intercept [boolean, default False] If True and if X is sparse, the method also re-
turns the intercept, and the solver is automatically changed to ‘sag’. This is only
a temporary fix for fitting the intercept with sparse data. For dense data, use
sklearn.linear_model._preprocess_data before your regression.
Returns
coef [array, shape = [n_features] or [n_targets, n_features]] Weight vector(s).
n_iter [int, optional] The actual number of iteration performed by the solver. Only returned if
return_n_iter is True.
intercept [float or array, shape = [n_targets]] The intercept of the model. Only returned if
return_intercept is True and if X is a scipy sparse array.
Notes
This function won’t compute the intercept.
6.22 sklearn.manifold: Manifold Learning
The sklearn.manifold module implements data embedding techniques.

User guide: See the Manifold learning section for further details.
manifold.Isomap([n_neighbors, n_components, . . . ]) Isomap Embedding

manifold.LocallyLinearEmbedding([. . . ]) Locally Linear Embedding
manifold.MDS([n_components, metric, n_init, . . . ]) Multidimensional scaling
manifold.SpectralEmbedding([n_components, Spectral embedding for non-linear dimensionality reduc-
. . . ]) tion.
manifold.TSNE([n_components, perplexity, . . . ]) t-distributed Stochastic Neighbor Embedding.
6.22.1 sklearn.manifold.Isomap
class sklearn.manifold.Isomap(n_neighbors=5, n_components=2, eigen_solver=’auto’, tol=0,

max_iter=None, path_method=’auto’, neighbors_algorithm=’auto’,
n_jobs=None)
Isomap Embedding
6.22. sklearn.manifold: Manifold Learning 1865

Non-linear dimensionality reduction through Isometric Mapping

Parameters
n_neighbors [integer] number of neighbors to consider for each point.
n_components [integer] number of coordinates for the manifold
eigen_solver [[‘auto’|’arpack’|’dense’]] ‘auto’ : Attempt to choose the most efficient solver for
the given problem.
‘arpack’ : Use Arnoldi decomposition to find the eigenvalues and eigenvectors.
‘dense’ : Use a direct solver (i.e. LAPACK) for the eigenvalue decomposition.
tol [float] Convergence tolerance passed to arpack or lobpcg. not used if eigen_solver ==
‘dense’.
max_iter [integer] Maximum number of iterations for the arpack solver. not used if
eigen_solver == ‘dense’.
path_method [string [‘auto’|’FW’|’D’]] Method to use in finding shortest path.
‘auto’ : attempt to choose the best algorithm automatically.
‘FW’ : Floyd-Warshall algorithm.
‘D’ : Dijkstra’s algorithm.
neighbors_algorithm [string [‘auto’|’brute’|’kd_tree’|’ball_tree’]] Algorithm to use for nearest
neighbors search, passed to neighbors.NearestNeighbors instance.
Attributes
embedding_ [array-like, shape (n_samples, n_components)] Stores the embedding vectors.
kernel_pca_ [object] KernelPCA object used to implement the embedding.
training_data_ [array-like, shape (n_samples, n_features)] Stores the training data.
nbrs_ [sklearn.neighbors.NearestNeighbors instance] Stores nearest neighbors instance, in-
cluding BallTree or KDtree if applicable.
dist_matrix_ [array-like, shape (n_samples, n_samples)] Stores the geodesic distance matrix
of training data.
References
[1]
Examples

>>> from sklearn.manifold import Isomap
>>> X.shape

(1797, 64)
>>> embedding = Isomap(n_components=2)
>>> X_transformed = embedding.fit_transform(X[:100])
(100, 2)
Methods
fit(X[, y]) Compute the embedding vectors for data X

reconstruction_error() Compute the reconstruction error for the embedding.
transform(X) Transform X.
__init__(n_neighbors=5, n_components=2, eigen_solver=’auto’, tol=0, max_iter=None,

path_method=’auto’, neighbors_algorithm=’auto’, n_jobs=None)
fit(X, y=None)
Compute the embedding vectors for data X
Parameters
X [{array-like, sparse matrix, BallTree, KDTree, NearestNeighbors}] Sample data, shape
= (n_samples, n_features), in the form of a numpy array, precomputed tree, or Nearest-
Neighbors object.
y [Ignored]
Returns
Parameters
X [{array-like, sparse matrix, BallTree, KDTree}] Training vector, where n_samples in the
y [Ignored]
Returns
Parameters
Returns

reconstruction_error()
Compute the reconstruction error for the embedding.
Returns
reconstruction_error [float]
Notes
The cost function of an isomap embedding is

E = frobenius_norm[K(D) - K(D_fit)] / n_samples
Where D is the matrix of distances for the input data X, D_fit is the matrix of distances for the output
embedding X_fit, and K is the isomap kernel:
K(D) = -0.5 * (I - 1/n_samples) * D^2 * (I - 1/n_samples)
of a nested object.
Returns
self
transform(X)
Transform X.
This is implemented by linking the points X into the graph of geodesic distances of the training data. First
the n_neighbors nearest neighbors of X are found in the training data, and from these the shortest geodesic
distances from each point in X to each point in the training data are computed in order to construct the
kernel. The embedding of X is the projection of this kernel onto the embedding vectors of the training set.
Parameters
Returns
Examples using sklearn.manifold.Isomap

• Manifold Learning methods on a severed sphere

6.22.2 sklearn.manifold.LocallyLinearEmbedding
class sklearn.manifold.LocallyLinearEmbedding(n_neighbors=5, n_components=2,

reg=0.001, eigen_solver=’auto’, tol=1e-
06, max_iter=100, method=’standard’,
hessian_tol=0.0001, modified_tol=1e-
12, neighbors_algorithm=’auto’, ran-
dom_state=None, n_jobs=None)
Locally Linear Embedding
Parameters
n_components [integer] number of coordinates for the manifold
reg [float] regularization constant, multiplies the trace of the local covariance matrix of the
distances.
eigen_solver [string, {‘auto’, ‘arpack’, ‘dense’}] auto : algorithm will attempt to choose the
best method for input data
arpack [use arnoldi iteration in shift-invert mode.] For this method, M may be a dense
matrix, sparse matrix, or general linear operator. Warning: ARPACK can be unstable for
some problems. It is best to try several random seeds in order to check results.
dense [use standard dense matrix operations for the eigenvalue] decomposition. For this
method, M must be an array or matrix type. This method should be avoided for large
problems.
tol [float, optional] Tolerance for ‘arpack’ method Not used if eigen_solver==’dense’.
max_iter [integer] maximum number of iterations for the arpack solver. Not used if
eigen_solver==’dense’.
method [string (‘standard’, ‘hessian’, ‘modified’ or ‘ltsa’)]
standard [use the standard locally linear embedding algorithm. see] reference [1]
hessian [use the Hessian eigenmap method. This method requires] n_neighbors >
n_components * (1 + (n_components + 1) / 2 see reference [2]
modified [use the modified locally linear embedding algorithm.] see reference [3]
ltsa [use local tangent space alignment algorithm] see reference [4]
hessian_tol [float, optional] Tolerance for Hessian eigenmapping method. Only used if
method == 'hessian'
modified_tol [float, optional] Tolerance for modified LLE method. Only used if method ==
'modified'
neighbors_algorithm [string [‘auto’|’brute’|’kd_tree’|’ball_tree’]] algorithm to use for nearest
neighbors search, passed to neighbors.NearestNeighbors instance
the RandomState instance used by np.random. Used when eigen_solver == ‘arpack’.

Attributes
embedding_ [array-like, shape [n_samples, n_components]] Stores the embedding vectors
reconstruction_error_ [float] Reconstruction error associated with embedding_
nbrs_ [NearestNeighbors object] Stores nearest neighbors instance, including BallTree or
KDtree if applicable.
References
[1], [2], [3], [4]
Examples

>>> from sklearn.manifold import LocallyLinearEmbedding
>>> X.shape
(1797, 64)
>>> embedding = LocallyLinearEmbedding(n_components=2)
(100, 2)
Methods
fit(X[, y]) Compute the embedding vectors for data X

fit_transform(X[, y]) Compute the embedding vectors for data X and trans-
form X.
transform(X) Transform new points into embedding space.
__init__(n_neighbors=5, n_components=2, reg=0.001, eigen_solver=’auto’, tol=1e-06,

max_iter=100, method=’standard’, hessian_tol=0.0001, modified_tol=1e-12, neigh-
bors_algorithm=’auto’, random_state=None, n_jobs=None)
fit(X, y=None)
Compute the embedding vectors for data X
Parameters
X [array-like of shape [n_samples, n_features]] training set.
y [Ignored]
Returns

Compute the embedding vectors for data X and transform X.
Parameters
X [array-like of shape [n_samples, n_features]] training set.
y [Ignored]
Returns
Parameters
Returns
of a nested object.
Returns
self
transform(X)
Transform new points into embedding space.
Parameters
Returns
X_new [array, shape = [n_samples, n_components]]
Notes
Because of scaling performed by this method, it is discouraged to use it together with methods that are not
scale-invariant (like SVMs)
Examples using sklearn.manifold.LocallyLinearEmbedding


6.22.3 sklearn.manifold.MDS
class sklearn.manifold.MDS(n_components=2, metric=True, n_init=4, max_iter=300, ver-

bose=0, eps=0.001, n_jobs=None, random_state=None, dissimilar-
ity=’euclidean’)
Multidimensional scaling
Parameters
n_components [int, optional, default: 2] Number of dimensions in which to immerse the dis-
similarities.
metric [boolean, optional, default: True] If True, perform metric MDS; otherwise, perform
nonmetric MDS.
n_init [int, optional, default: 4] Number of times the SMACOF algorithm will be run with
different initializations. The final results will be the best output of the runs, determined by
the run with the smallest final stress.
max_iter [int, optional, default: 300] Maximum number of iterations of the SMACOF algo-
rithm for a single run.
verbose [int, optional, default: 0] Level of verbosity.
eps [float, optional, default: 1e-3] Relative tolerance with respect to stress at which to declare
convergence.
If multiple initializations are used (n_init), each run of the algorithm is computed in
parallel.
random_state [int, RandomState instance or None, optional, default: None] The generator
dissimilarity [‘euclidean’ | ‘precomputed’, optional, default: ‘euclidean’] Dissimilarity mea-
sure to use:
• ‘euclidean’: Pairwise Euclidean distances between points in the dataset.
• ‘precomputed’: Pre-computed dissimilarities are passed directly to fit and
fit_transform.
Attributes
embedding_ [array-like, shape (n_samples, n_components)] Stores the position of the dataset
in the embedding space.
stress_ [float] The final value of the stress (sum of squared distance of the disparities and the
distances for all constrained points).
References
“Modern Multidimensional Scaling - Theory and Applications” Borg, I.; Groenen P. Springer Series in Statistics
(1997)

“Nonmetric multidimensional scaling: a numerical method” Kruskal, J. Psychometrika, 29 (1964)

“Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis” Kruskal, J. Psychometrika,
29, (1964)
Examples

>>> from sklearn.manifold import MDS
>>> X.shape
(1797, 64)
>>> embedding = MDS(n_components=2)
(100, 2)
Methods
fit(X[, y, init]) Computes the position of the points in the embedding

space
fit_transform(X[, y, init]) Fit the data from X, and returns the embedded coordi-
nates
__init__(n_components=2, metric=True, n_init=4, max_iter=300, verbose=0, eps=0.001,

n_jobs=None, random_state=None, dissimilarity=’euclidean’)
fit(X, y=None, init=None)
Computes the position of the points in the embedding space
Parameters
X [array, shape (n_samples, n_features) or (n_samples, n_samples)] Input data. If
dissimilarity=='precomputed', the input should be the dissimilarity matrix.
y [Ignored]
init [ndarray, shape (n_samples,), optional, default: None] Starting configuration of the em-
bedding to initialize the SMACOF algorithm. By default, the algorithm is initialized with
a randomly chosen array.
fit_transform(X, y=None, init=None)
Fit the data from X, and returns the embedded coordinates
Parameters
X [array, shape (n_samples, n_features) or (n_samples, n_samples)] Input data. If
dissimilarity=='precomputed', the input should be the dissimilarity matrix.
y [Ignored]
init [ndarray, shape (n_samples,), optional, default: None] Starting configuration of the em-
bedding to initialize the SMACOF algorithm. By default, the algorithm is initialized with
a randomly chosen array.

Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.manifold.MDS
• Multi-dimensional scaling
6.22.4 sklearn.manifold.SpectralEmbedding
class sklearn.manifold.SpectralEmbedding(n_components=2, affinity=’nearest_neighbors’,

gamma=None, random_state=None,
eigen_solver=None, n_neighbors=None,
n_jobs=None)
Spectral embedding for non-linear dimensionality reduction.
Forms an affinity matrix given by the specified function and applies spectral decomposition to the corresponding
graph laplacian. The resulting transformation is given by the value of the eigenvectors for each data point.
Note : Laplacian Eigenmaps is the actual algorithm implemented here.
Parameters
n_components [integer, default: 2] The dimension of the projected subspace.
affinity [string or callable, default]
How to construct the affinity matrix.
• ‘nearest_neighbors’ : construct affinity matrix by knn graph
• ‘rbf’ : construct affinity matrix by rbf kernel
• ‘precomputed’ : interpret X as precomputed affinity matrix
• callable : use passed in function as affinity the function takes in data matrix (n_samples,
n_features) and return affinity matrix (n_samples, n_samples).

gamma [float, optional, default] Kernel coefficient for rbf kernel.

random_state [int, RandomState instance or None, optional, default: None] A pseudo random
number generator used for the initialization of the lobpcg eigenvectors. If int, random_state
instance used by np.random. Used when solver == ‘amg’.
but may also lead to instabilities.
n_neighbors [int, default] Number of nearest neighbors for nearest_neighbors graph building.
Attributes
embedding_ [array, shape = (n_samples, n_components)] Spectral embedding of the training
matrix.
affinity_matrix_ [array, shape = (n_samples, n_samples)] Affinity_matrix constructed from
samples or precomputed.
References

summary?doi=10.1.1.165.9323
• On Spectral Clustering: Analysis and an algorithm, 2001 Andrew Y. Ng, Michael I. Jordan, Yair Weiss
http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.19.8100
Examples

>>> from sklearn.manifold import SpectralEmbedding
>>> X.shape
(1797, 64)
>>> embedding = SpectralEmbedding(n_components=2)
(100, 2)
Methods



__init__(n_components=2, affinity=’nearest_neighbors’, gamma=None, random_state=None,

eigen_solver=None, n_neighbors=None, n_jobs=None)
fit(X, y=None)
Parameters
X [array-like, shape (n_samples, n_features)] Training vector, where n_samples is the num-
If affinity is “precomputed” X : array-like, shape (n_samples, n_samples), Interpret X as
precomputed adjacency graph computed from samples.
Returns
Parameters
X [array-like, shape (n_samples, n_features)] Training vector, where n_samples is the num-
If affinity is “precomputed” X : array-like, shape (n_samples, n_samples), Interpret X as
precomputed adjacency graph computed from samples.
Returns
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.manifold.SpectralEmbedding



6.22.5 sklearn.manifold.TSNE
class sklearn.manifold.TSNE(n_components=2, perplexity=30.0, early_exaggeration=12.0,

learning_rate=200.0, n_iter=1000, n_iter_without_progress=300,
min_grad_norm=1e-07, metric=’euclidean’, init=’random’, verbose=0,
random_state=None, method=’barnes_hut’, angle=0.5)
t-distributed Stochastic Neighbor Embedding.
t-SNE [1] is a tool to visualize high-dimensional data. It converts similarities between data points to joint
probabilities and tries to minimize the Kullback-Leibler divergence between the joint probabilities of the low-
dimensional embedding and the high-dimensional data. t-SNE has a cost function that is not convex, i.e. with
different initializations we can get different results.
It is highly recommended to use another dimensionality reduction method (e.g. PCA for dense data or Truncat-
edSVD for sparse data) to reduce the number of dimensions to a reasonable amount (e.g. 50) if the number of
features is very high. This will suppress some noise and speed up the computation of pairwise distances between
samples. For more tips see Laurens van der Maaten’s FAQ [2].
Parameters
n_components [int, optional (default: 2)] Dimension of the embedded space.
perplexity [float, optional (default: 30)] The perplexity is related to the number of nearest
neighbors that is used in other manifold learning algorithms. Larger datasets usually require
a larger perplexity. Consider selecting a value between 5 and 50. The choice is not extremely
critical since t-SNE is quite insensitive to this parameter.
early_exaggeration [float, optional (default: 12.0)] Controls how tight natural clusters in the
original space are in the embedded space and how much space will be between them. For
larger values, the space between natural clusters will be larger in the embedded space.
Again, the choice of this parameter is not very critical. If the cost function increases during
initial optimization, the early exaggeration factor or the learning rate might be too high.
learning_rate [float, optional (default: 200.0)] The learning rate for t-SNE is usually in the
range [10.0, 1000.0]. If the learning rate is too high, the data may look like a ‘ball’ with any
point approximately equidistant from its nearest neighbours. If the learning rate is too low,
most points may look compressed in a dense cloud with few outliers. If the cost function
gets stuck in a bad local minimum increasing the learning rate may help.
n_iter [int, optional (default: 1000)] Maximum number of iterations for the optimization.
Should be at least 250.
n_iter_without_progress [int, optional (default: 300)] Maximum number of iterations without
progress before we abort the optimization, used after 250 initial iterations with early exag-
geration. Note that progress is only checked every 50 iterations so this value is rounded to
the next multiple of 50.
New in version 0.17: parameter n_iter_without_progress to control stopping criteria.
min_grad_norm [float, optional (default: 1e-7)] If the gradient norm is below this threshold,
the optimization will be stopped.

metric [string or callable, optional] The metric to use when calculating distance between
instances in a feature array. If metric is a string, it must be one of the options al-
lowed by scipy.spatial.distance.pdist for its metric parameter, or a metric listed in pair-
wise.PAIRWISE_DISTANCE_FUNCTIONS. If metric is “precomputed”, X is assumed to
be a distance matrix. Alternatively, if metric is a callable function, it is called on each pair
of instances (rows) and the resulting value recorded. The callable should take two arrays
from X as input and return a value indicating the distance between them. The default is
“euclidean” which is interpreted as squared euclidean distance.
init [string or numpy array, optional (default: “random”)] Initialization of embedding. Possible
options are ‘random’, ‘pca’, and a numpy array of shape (n_samples, n_components). PCA
initialization cannot be used with precomputed distances and is usually more globally stable
than random initialization.
verbose [int, optional (default: 0)] Verbosity level.
random_state [int, RandomState instance or None, optional (default: None)] If int, ran-
the RandomState instance used by np.random. Note that different initializations might re-
sult in different local minima of the cost function.
method [string (default: ‘barnes_hut’)] By default the gradient calculation algorithm uses
Barnes-Hut approximation running in O(NlogN) time. method=’exact’ will run on the
slower, but exact, algorithm in O(N^2) time. The exact algorithm should be used when
nearest-neighbor errors need to be better than 3%. However, the exact method cannot scale
to millions of examples.
New in version 0.17: Approximate optimization method via the Barnes-Hut.
angle [float (default: 0.5)] Only used if method=’barnes_hut’ This is the trade-off between
speed and accuracy for Barnes-Hut T-SNE. ‘angle’ is the angular size (referred to as theta
in [3]) of a distant node as measured from a point. If this size is below ‘angle’ then it is used
as a summary node of all points contained within it. This method is not very sensitive to
changes in this parameter in the range of 0.2 - 0.8. Angle less than 0.2 has quickly increasing
computation time and angle greater 0.8 has quickly increasing error.
Attributes
embedding_ [array-like, shape (n_samples, n_components)] Stores the embedding vectors.
kl_divergence_ [float] Kullback-Leibler divergence after optimization.
References
[1] van der Maaten, L.J.P.; Hinton, G.E. Visualizing High-Dimensional Data Using t-SNE. Journal of Ma-
chine Learning Research 9:2579-2605, 2008.
[2] van der Maaten, L.J.P. t-Distributed Stochastic Neighbor Embedding https://lvdmaaten.github.io/tsne/
[3] L.J.P. van der Maaten. Accelerating t-SNE using Tree-Based Algorithms. Journal of Machine Learn-
ing Research 15(Oct):3221-3245, 2014. https://lvdmaaten.github.io/publications/papers/JMLR_2014.pdf

Examples

>>> from sklearn.manifold import TSNE
>>> X = np.array([[0, 0, 0], [0, 1, 1], [1, 0, 1], [1, 1, 1]])
>>> X_embedded = TSNE(n_components=2).fit_transform(X)
>>> X_embedded.shape
(4, 2)
Methods
fit(X[, y]) Fit X into an embedded space.

fit_transform(X[, y]) Fit X into an embedded space and return that trans-
formed output.
__init__(n_components=2, perplexity=30.0, early_exaggeration=12.0, learning_rate=200.0,

n_iter=1000, n_iter_without_progress=300, min_grad_norm=1e-07, metric=’euclidean’,
init=’random’, verbose=0, random_state=None, method=’barnes_hut’, angle=0.5)
fit(X, y=None)
Fit X into an embedded space.
Parameters
X [array, shape (n_samples, n_features) or (n_samples, n_samples)] If the metric is ‘pre-
computed’ X must be a square distance matrix. Otherwise it contains a sample per row. If
the method is ‘exact’, X may be a sparse matrix of type ‘csr’, ‘csc’ or ‘coo’.
y [Ignored]
Fit X into an embedded space and return that transformed output.
Parameters
X [array, shape (n_samples, n_features) or (n_samples, n_samples)] If the metric is ‘pre-
computed’ X must be a square distance matrix. Otherwise it contains a sample per row.
y [Ignored]
Returns
X_new [array, shape (n_samples, n_components)] Embedding of the training data in low-
dimensional space.
Parameters
Returns

n_iter_final
DEPRECATED: Attribute n_iter_final was deprecated in version 0.19 and will be removed in 0.21. Use
n_iter_ instead
of a nested object.
Returns
self
Examples using sklearn.manifold.TSNE

manifold.locally_linear_embedding(X, . . . [, Perform a Locally Linear Embedding analysis on the data.

. . . ])
manifold.smacof(dissimilarities[, metric, . . . ]) Computes multidimensional scaling using the SMACOF
algorithm.
manifold.spectral_embedding(adjacency[, . . . ]) Project the sample on the first eigenvectors of the graph
Laplacian.
6.22.6 sklearn.manifold.locally_linear_embedding
sklearn.manifold.locally_linear_embedding(X, n_neighbors, n_components, reg=0.001,

eigen_solver=’auto’, tol=1e-06, max_iter=100,
method=’standard’, hessian_tol=0.0001,
modified_tol=1e-12, random_state=None,
n_jobs=None)
Perform a Locally Linear Embedding analysis on the data.
Parameters
X [{array-like, NearestNeighbors}] Sample data, shape = (n_samples, n_features), in the form
of a numpy array or a NearestNeighbors object.
n_components [integer] number of coordinates for the manifold.
reg [float] regularization constant, multiplies the trace of the local covariance matrix of the
distances.
eigen_solver [string, {‘auto’, ‘arpack’, ‘dense’}] auto : algorithm will attempt to choose the
best method for input data

arpack [use arnoldi iteration in shift-invert mode.] For this method, M may be a dense
matrix, sparse matrix, or general linear operator. Warning: ARPACK can be unstable for
some problems. It is best to try several random seeds in order to check results.
dense [use standard dense matrix operations for the eigenvalue] decomposition. For this
method, M must be an array or matrix type. This method should be avoided for large
problems.
tol [float, optional] Tolerance for ‘arpack’ method Not used if eigen_solver==’dense’.
max_iter [integer] maximum number of iterations for the arpack solver.
method [{‘standard’, ‘hessian’, ‘modified’, ‘ltsa’}]
standard [use the standard locally linear embedding algorithm.] see reference [1]
hessian [use the Hessian eigenmap method. This method requires] n_neighbors >
n_components * (1 + (n_components + 1) / 2. see reference [2]
modified [use the modified locally linear embedding algorithm.] see reference [3]
ltsa [use local tangent space alignment algorithm] see reference [4]
hessian_tol [float, optional] Tolerance for Hessian eigenmapping method. Only used if method
== ‘hessian’
modified_tol [float, optional] Tolerance for modified LLE method. Only used if method ==
‘modified’
the RandomState instance used by np.random. Used when solver == ‘arpack’.
Returns
Y [array-like, shape [n_samples, n_components]] Embedding vectors.
squared_error [float] Reconstruction error for the embedding vectors. Equivalent to norm(Y
- W Y, 'fro')**2, where W are the reconstruction weights.
References
[1], [2], [3], [4]
Examples using sklearn.manifold.locally_linear_embedding
• Swiss Roll reduction with LLE
6.22.7 sklearn.manifold.smacof
sklearn.manifold.smacof(dissimilarities, metric=True, n_components=2, init=None, n_init=8,

n_jobs=None, max_iter=300, verbose=0, eps=0.001, random_state=None,
return_n_iter=False)
Computes multidimensional scaling using the SMACOF algorithm.

The SMACOF (Scaling by MAjorizing a COmplicated Function) algorithm is a multidimensional scaling algo-
rithm which minimizes an objective function (the stress) using a majorization technique. Stress majorization,
also known as the Guttman Transform, guarantees a monotone convergence of stress, and is more powerful than
traditional techniques such as gradient descent.
The SMACOF algorithm for metric MDS can summarized by the following steps:
1. Set an initial start configuration, randomly or not.
2. Compute the stress
3. Compute the Guttman Transform
4. Iterate 2 and 3 until convergence.
The nonmetric algorithm adds a monotonic regression step before computing the stress.
Parameters
dissimilarities [ndarray, shape (n_samples, n_samples)] Pairwise dissimilarities between the
points. Must be symmetric.
metric [boolean, optional, default: True] Compute metric or nonmetric SMACOF algorithm.
n_components [int, optional, default: 2] Number of dimensions in which to immerse the dis-
similarities. If an init array is provided, this option is overridden and the shape of init
is used to determine the dimensionality of the embedding space.
init [ndarray, shape (n_samples, n_components), optional, default: None] Starting configura-
tion of the embedding to initialize the algorithm. By default, the algorithm is initialized
with a randomly chosen array.
n_init [int, optional, default: 8] Number of times the SMACOF algorithm will be run with
different initializations. The final results will be the best output of the runs, determined by
the run with the smallest final stress. If init is provided, this option is overridden and a
single run is performed.
If multiple initializations are used (n_init), each run of the algorithm is computed in
parallel.
max_iter [int, optional, default: 300] Maximum number of iterations of the SMACOF algo-
rithm for a single run.
verbose [int, optional, default: 0] Level of verbosity.
eps [float, optional, default: 1e-3] Relative tolerance with respect to stress at which to declare
convergence.
return_n_iter [bool, optional, default: False] Whether or not to return the number of iterations.
Returns
X [ndarray, shape (n_samples, n_components)] Coordinates of the points in a
n_components-space.

stress [float] The final value of the stress (sum of squared distance of the disparities and the
distances for all constrained points).
n_iter [int] The number of iterations corresponding to the best stress. Returned only if
return_n_iter is set to True.
Notes
“Modern Multidimensional Scaling - Theory and Applications” Borg, I.; Groenen P. Springer Series in Statistics
(1997)
“Nonmetric multidimensional scaling: a numerical method” Kruskal, J. Psychometrika, 29 (1964)
“Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis” Kruskal, J. Psychometrika,
29, (1964)
6.22.8 sklearn.manifold.spectral_embedding
sklearn.manifold.spectral_embedding(adjacency, n_components=8, eigen_solver=None, ran-

dom_state=None, eigen_tol=0.0, norm_laplacian=True,
drop_first=True)
Project the sample on the first eigenvectors of the graph Laplacian.
The adjacency matrix is used to compute a normalized graph Laplacian whose spectrum (especially the eigen-
vectors associated to the smallest eigenvalues) has an interpretation in terms of minimal number of cuts neces-
sary to split the graph into comparably sized components.
This embedding can also ‘work’ even if the adjacency variable is not strictly the adjacency matrix of a graph
but more generally an affinity or similarity matrix between samples (for instance the heat kernel of a euclidean
distance matrix or a k-NN matrix).
However care must taken to always make the affinity matrix symmetric so that the eigenvector decomposition
works as expected.
Note : Laplacian Eigenmaps is the actual algorithm implemented here.
Parameters
adjacency [array-like or sparse matrix, shape: (n_samples, n_samples)] The adjacency matrix
of the graph to embed.
n_components [integer, optional, default 8] The dimension of the projection subspace.
eigen_solver [{None, ‘arpack’, ‘lobpcg’, or ‘amg’}, default None] The eigenvalue decompo-
sition strategy to use. AMG requires pyamg to be installed. It can be faster on very large,
sparse problems, but may also lead to instabilities.
random_state [int, RandomState instance or None, optional, default: None] A pseudo random
number generator used for the initialization of the lobpcg eigenvectors decomposition. If int,
random_state is the seed used by the random number generator; If RandomState instance,
random_state is the random number generator; If None, the random number generator is the
RandomState instance used by np.random. Used when solver == ‘amg’.
eigen_tol [float, optional, default=0.0] Stopping criterion for eigendecomposition of the Lapla-
norm_laplacian [bool, optional, default=True] If True, then compute normalized Laplacian.

drop_first [bool, optional, default=True] Whether to drop the first eigenvector. For spectral em-
bedding, this should be True as the first eigenvector should be constant vector for connected
graph, but for spectral clustering, this should be kept as False to retain the first eigenvector.
Returns
embedding [array, shape=(n_samples, n_components)] The reduced samples.
Notes
Spectral Embedding (Laplacian Eigenmaps) is most useful when the graph has one connected component. If
there graph has many components, the first few eigenvectors will simply uncover the connected components of
the graph.
References
• https://en.wikipedia.org/wiki/LOBPCG
• Toward the Optimal Preconditioned Eigensolver: Locally Optimal Block Preconditioned Conjugate Gra-
dient Method Andrew V. Knyazev https://doi.org/10.1137%2FS1064827500366124
6.23 sklearn.metrics: Metrics
See the Model evaluation: quantifying the quality of predictions section and the Pairwise metrics, Affinities and
Kernels section of the user guide for further details. The sklearn.metrics module includes score functions,
performance metrics and pairwise metrics and distance computations.
6.23.1 Model Selection Interface
See the The scoring parameter: defining model evaluation rules section of the user guide for further details.
metrics.check_scoring(estimator[, scoring, . . . ]) Determine scorer from user options.

metrics.get_scorer(scoring) Get a scorer from string
metrics.make_scorer(score_func[, . . . ]) Make a scorer from a performance metric or loss function.
sklearn.metrics.check_scoring
sklearn.metrics.check_scoring(estimator, scoring=None, allow_none=False)

Determine scorer from user options.
A TypeError will be thrown if the estimator cannot be scored.
Parameters
estimator [estimator object implementing ‘fit’] The object to use to fit the data.
X, y).
allow_none [boolean, optional, default: False] If no scoring is specified and the estimator has
no score function, we can either return None or raise an exception.

Returns
scoring [callable] A scorer callable object / function with signature scorer(estimator,
X, y).
sklearn.metrics.get_scorer
sklearn.metrics.get_scorer(scoring)
Get a scorer from string
Parameters
scoring [str | callable] scoring method as string. If callable it is returned as is.
Returns
scorer [callable] The scorer.
sklearn.metrics.make_scorer
sklearn.metrics.make_scorer(score_func, greater_is_better=True, needs_proba=False,

needs_threshold=False, **kwargs)
Make a scorer from a performance metric or loss function.
This factory function wraps scoring functions for use in GridSearchCV and cross_val_score. It takes
a score function, such as accuracy_score, mean_squared_error, adjusted_rand_index or
average_precision and returns a callable that scores an estimator’s output.
Parameters
score_func [callable,] Score function (or loss function) with signature score_func(y,
y_pred, **kwargs).
greater_is_better [boolean, default=True] Whether score_func is a score function (default),
meaning high is good, or a loss function, meaning low is good. In the latter case, the scorer
object will sign-flip the outcome of the score_func.
needs_proba [boolean, default=False] Whether score_func requires predict_proba to get prob-
ability estimates out of a classifier.
needs_threshold [boolean, default=False] Whether score_func takes a continuous decision cer-
tainty. This only works for binary classification using estimators that have either a deci-
sion_function or predict_proba method.
For example average_precision or the area under the roc curve can not be computed
using discrete predictions alone.
**kwargs [additional arguments] Additional parameters to be passed to score_func.
Returns
scorer [callable] Callable object that returns a scalar score; greater is better.
Examples
6.23. sklearn.metrics: Metrics 1885

>>> from sklearn.metrics import fbeta_score, make_scorer

>>> ftwo_scorer = make_scorer(fbeta_score, beta=2)
>>> ftwo_scorer
make_scorer(fbeta_score, beta=2)
>>> grid = GridSearchCV(LinearSVC(), param_grid={'C': [1, 10]},
... scoring=ftwo_scorer)
Examples using sklearn.metrics.make_scorer
6.23.2 Classification metrics
See the Classification metrics section of the user guide for further details.
metrics.accuracy_score(y_true, y_pred[, . . . ]) Accuracy classification score.

metrics.auc(x, y[, reorder]) Compute Area Under the Curve (AUC) using the trape-
zoidal rule
metrics.average_precision_score(y_true, Compute average precision (AP) from prediction scores
y_score)
metrics.balanced_accuracy_score(y_true, Compute the balanced accuracy
y_pred)
metrics.brier_score_loss(y_true, y_prob[, . . . ]) Compute the Brier score.
metrics.classification_report(y_true, Build a text report showing the main classification metrics
y_pred)
metrics.cohen_kappa_score(y1, y2[, labels, . . . ]) Cohen’s kappa: a statistic that measures inter-annotator
agreement.
metrics.confusion_matrix(y_true, y_pred[, . . . ]) Compute confusion matrix to evaluate the accuracy of a
classification
metrics.f1_score(y_true, y_pred[, labels, . . . ]) Compute the F1 score, also known as balanced F-score or
F-measure
metrics.fbeta_score(y_true, y_pred, beta[, . . . ]) Compute the F-beta score
metrics.hamming_loss(y_true, y_pred[, . . . ]) Compute the average Hamming loss.
metrics.hinge_loss(y_true, pred_decision[, . . . ]) Average hinge loss (non-regularized)
metrics.jaccard_similarity_score(y_true, Jaccard similarity coefficient score
y_pred)
metrics.log_loss(y_true, y_pred[, eps, . . . ]) Log loss, aka logistic loss or cross-entropy loss.
metrics.matthews_corrcoef(y_true, y_pred[, Compute the Matthews correlation coefficient (MCC)
. . . ])
metrics.precision_recall_curve(y_true, . . . ) Compute precision-recall pairs for different probability
thresholds
metrics.precision_recall_fscore_support(. . .Compute
) precision, recall, F-measure and support for each
class
metrics.precision_score(y_true, y_pred[, . . . ]) Compute the precision
metrics.recall_score(y_true, y_pred[, . . . ]) Compute the recall
metrics.roc_auc_score(y_true, y_score[, . . . ]) Compute Area Under the Receiver Operating Characteris-
tic Curve (ROC AUC) from prediction scores.
metrics.roc_curve(y_true, y_score[, . . . ]) Compute Receiver operating characteristic (ROC)


metrics.zero_one_loss(y_true, y_pred[, . . . ]) Zero-one classification loss.
sklearn.metrics.accuracy_score
sklearn.metrics.accuracy_score(y_true, y_pred, normalize=True, sample_weight=None)

Accuracy classification score.
In multilabel classification, this function computes subset accuracy: the set of labels predicted for a sample must
exactly match the corresponding set of labels in y_true.
Parameters
y_true [1d array-like, or label indicator array / sparse matrix] Ground truth (correct) labels.
y_pred [1d array-like, or label indicator array / sparse matrix] Predicted labels, as returned by
a classifier.
normalize [bool, optional (default=True)] If False, return the number of correctly classified
samples. Otherwise, return the fraction of correctly classified samples.
Returns
score [float] If normalize == True, return the fraction of correctly classified samples
(float), else returns the number of correctly classified samples (int).
The best performance is 1 with normalize == True and the number of samples with
normalize == False.
See also:
jaccard_similarity_score, hamming_loss, zero_one_loss
Notes
In binary and multiclass classification, this function is equal to the jaccard_similarity_score function.
Examples

>>> y_pred = [0, 2, 1, 3]
>>> y_true = [0, 1, 2, 3]
>>> accuracy_score(y_true, y_pred)
0.5
>>> accuracy_score(y_true, y_pred, normalize=False)
2
>>> accuracy_score(np.array([[0, 1], [1, 1]]), np.ones((2, 2)))

0.5

Examples using sklearn.metrics.accuracy_score

sklearn.metrics.auc
sklearn.metrics.auc(x, y, reorder=’deprecated’)
Compute Area Under the Curve (AUC) using the trapezoidal rule
This is a general function, given points on a curve. For computing the area under the ROC-
curve, see roc_auc_score. For an alternative way to summarize a precision-recall curve, see
average_precision_score.
Parameters
x [array, shape = [n]] x coordinates. These must be either monotonic increasing or monotonic
decreasing.
y [array, shape = [n]] y coordinates.
reorder [boolean, optional (default=’deprecated’)] Whether to sort x before computing. If
False, assume that x must be either monotonic increasing or monotonic decreasing. If True,
y is used to break ties when sorting x. Make sure that y has a monotonic relation to x when
setting reorder to True.
Deprecated since version 0.20: Parameter reorder has been deprecated in version 0.20
and will be removed in 0.22. It’s introduced for roc_auc_score (not for general use) and is
no longer used there. What’s more, the result from auc will be significantly influenced if
x is sorted unexpectedly due to slight floating point error (See issue #9786). Future (and
default) behavior is equivalent to reorder=False.
Returns
auc [float]
See also:
roc_auc_score Compute the area under the ROC curve

average_precision_score Compute average precision from prediction scores
precision_recall_curve Compute precision-recall pairs for different probability thresholds
Examples

>>> y = np.array([1, 1, 2, 2])
>>> pred = np.array([0.1, 0.4, 0.35, 0.8])
>>> fpr, tpr, thresholds = metrics.roc_curve(y, pred, pos_label=2)

>>> metrics.auc(fpr, tpr)

0.75
Examples using sklearn.metrics.auc

sklearn.metrics.average_precision_score
sklearn.metrics.average_precision_score(y_true, y_score, average=’macro’, pos_label=1,

sample_weight=None)
Compute average precision (AP) from prediction scores
AP summarizes a precision-recall curve as the weighted mean of precisions achieved at each threshold, with the
increase in recall from the previous threshold used as the weight:
∑︁
AP = (𝑅𝑛 − 𝑅𝑛−1 )𝑃𝑛
𝑛
where 𝑃𝑛 and 𝑅𝑛 are the precision and recall at the nth threshold [1]. This implementation is not interpolated
and is different from computing the area under the precision-recall curve with the trapezoidal rule, which uses
linear interpolation and can be too optimistic.
Note: this implementation is restricted to the binary classification task or multilabel classification task.
Parameters
y_true [array, shape = [n_samples] or [n_samples, n_classes]] True binary labels or binary label
indicators.
y_score [array, shape = [n_samples] or [n_samples, n_classes]] Target scores, can either be
probability estimates of the positive class, confidence values, or non-thresholded measure
of decisions (as returned by “decision_function” on some classifiers).
average [string, [None, ‘micro’, ‘macro’ (default), ‘samples’, ‘weighted’]] If None, the scores
for each class are returned. Otherwise, this determines the type of averaging performed on
the data:
'micro': Calculate metrics globally by considering each element of the label indicator
matrix as a label.
'macro': Calculate metrics for each label, and find their unweighted mean. This does not
take label imbalance into account.
'weighted': Calculate metrics for each label, and find their average, weighted by sup-
port (the number of true instances for each label).
'samples': Calculate metrics for each instance, and find their average.
Will be ignored when y_true is binary.
pos_label [int or str (default=1)] The label of the positive class. Only applied to binary
y_true. For multilabel-indicator y_true, pos_label is fixed to 1.


Returns
average_precision [float]
See also:

precision_recall_curve Compute precision-recall pairs for different probability thresholds
Notes
Changed in version 0.19: Instead of linearly interpolating between operating points, precisions are weighted by
the change in recall since the last operating point.
References
[1]
Examples

>>> from sklearn.metrics import average_precision_score
>>> y_true = np.array([0, 0, 1, 1])
>>> y_scores = np.array([0.1, 0.4, 0.35, 0.8])
>>> average_precision_score(y_true, y_scores)
0.83...
Examples using sklearn.metrics.average_precision_score
sklearn.metrics.balanced_accuracy_score
sklearn.metrics.balanced_accuracy_score(y_true, y_pred, sample_weight=None, ad-

justed=False)
Compute the balanced accuracy
The balanced accuracy in binary and multiclass classification problems to deal with imbalanced datasets. It is
defined as the average of recall obtained on each class.
The best value is 1 and the worst value is 0 when adjusted=False.
Parameters
y_true [1d array-like] Ground truth (correct) target values.
y_pred [1d array-like] Estimated targets as returned by a classifier.

adjusted [bool, default=False] When true, the result is adjusted for chance, so that random
performance would score 0, and perfect performance scores 1.
Returns
balanced_accuracy [float]
See also:
recall_score, roc_auc_score
Notes
Some literature promotes alternative definitions of balanced accuracy. Our definition is equivalent to
accuracy_score with class-balanced sample weights, and shares desirable properties with the binary case.
See the User Guide.
References
[1], [2]
Examples
>>> from sklearn.metrics import balanced_accuracy_score

>>> y_true = [0, 1, 0, 0, 1, 0]
>>> y_pred = [0, 1, 0, 0, 0, 1]
>>> balanced_accuracy_score(y_true, y_pred)
0.625
sklearn.metrics.brier_score_loss
sklearn.metrics.brier_score_loss(y_true, y_prob, sample_weight=None, pos_label=None)

Compute the Brier score. The smaller the Brier score, the better, hence the naming with “loss”. Across all
items in a set N predictions, the Brier score measures the mean squared difference between (1) the predicted
probability assigned to the possible outcomes for item i, and (2) the actual outcome. Therefore, the lower
the Brier score is for a set of predictions, the better the predictions are calibrated. Note that the Brier score
always takes on a value between zero and one, since this is the largest possible difference between a predicted
probability (which must be between zero and one) and the actual outcome (which can take on values of only 0
and 1). The Brier loss is composed of refinement loss and calibration loss. The Brier score is appropriate for
binary and categorical outcomes that can be structured as true or false, but is inappropriate for ordinal variables
which can take on three or more values (this is because the Brier score assumes that all possible outcomes are
equivalently “distant” from one another). Which label is considered to be the positive label is controlled via the
parameter pos_label, which defaults to 1. Read more in the User Guide.
Parameters
y_true [array, shape (n_samples,)] True targets.
y_prob [array, shape (n_samples,)] Probabilities of the positive class.
pos_label [int or str, default=None] Label of the positive class. If None, the maximum label is
used as positive class

Returns
score [float] Brier score
References
[1]
Examples

>>> from sklearn.metrics import brier_score_loss
>>> y_true = np.array([0, 1, 1, 0])
>>> y_true_categorical = np.array(["spam", "ham", "ham", "spam"])
>>> y_prob = np.array([0.1, 0.9, 0.8, 0.3])
>>> brier_score_loss(y_true, y_prob)
0.037...
>>> brier_score_loss(y_true, 1-y_prob, pos_label=0)
0.037...
>>> brier_score_loss(y_true_categorical, y_prob, pos_
˓→label="ham")
0.037...
>>> brier_score_loss(y_true, np.array(y_prob) > 0.5)
0.0
Examples using sklearn.metrics.brier_score_loss

sklearn.metrics.classification_report
sklearn.metrics.classification_report(y_true, y_pred, labels=None, target_names=None,

sample_weight=None, digits=2, output_dict=False)
Build a text report showing the main classification metrics
Parameters
y_true [1d array-like, or label indicator array / sparse matrix] Ground truth (correct) target
values.
y_pred [1d array-like, or label indicator array / sparse matrix] Estimated targets as returned by
a classifier.
labels [array, shape = [n_labels]] Optional list of label indices to include in the report.
target_names [list of strings] Optional display names matching the labels (same order).
digits [int] Number of digits for formatting output floating point values. When output_dict
is True, this will be ignored and the returned values will not be rounded.

output_dict [bool (default = False)] If True, return output as dict

Returns
report [string / dict] Text summary of the precision, recall, F1 score for each class. Dictionary
returned if output_dict is True. Dictionary has the following structure:
{'label 1': {'precision':0.5,

'recall':1.0,
'f1-score':0.67,
'support':1},
'label 2': { ... },
...
}
The reported averages include micro average (averaging the total true positives, false neg-
atives and false positives), macro average (averaging the unweighted mean per label),
weighted average (averaging the support-weighted mean per label) and sample average (only
for multilabel classification). See also precision_recall_fscore_support for
more details on averages.
Note that in binary classification, recall of the positive class is also known as “sensitivity”;
recall of the negative class is “specificity”.
Examples
>>> from sklearn.metrics import classification_report

>>> y_true = [0, 1, 2, 2, 2]
>>> y_pred = [0, 0, 2, 2, 1]
>>> target_names = ['class 0', 'class 1', 'class 2']
>>> print(classification_report(y_true, y_pred, target_names=target_names))
class 0 0.50 1.00 0.67 1

class 1 0.00 0.00 0.00 1
class 2 1.00 0.67 0.80 3
micro avg 0.60 0.60 0.60 5

macro avg 0.50 0.56 0.49 5
weighted avg 0.70 0.60 0.61 5
Examples using sklearn.metrics.classification_report


sklearn.metrics.cohen_kappa_score
sklearn.metrics.cohen_kappa_score(y1, y2, labels=None, weights=None, sam-

ple_weight=None)
Cohen’s kappa: a statistic that measures inter-annotator agreement.
This function computes Cohen’s kappa [1], a score that expresses the level of agreement between two annotators
on a classification problem. It is defined as
𝜅 = (𝑝𝑜 − 𝑝𝑒 )/(1 − 𝑝𝑒 )
where 𝑝𝑜 is the empirical probability of agreement on the label assigned to any sample (the observed agreement
ratio), and 𝑝𝑒 is the expected agreement when both annotators assign labels randomly. 𝑝𝑒 is estimated using a
per-annotator empirical prior over the class labels [2].
Parameters
y1 [array, shape = [n_samples]] Labels assigned by the first annotator.
y2 [array, shape = [n_samples]] Labels assigned by the second annotator. The kappa statistic is
symmetric, so swapping y1 and y2 doesn’t change the value.
labels [array, shape = [n_classes], optional] List of labels to index the matrix. This may be used
to select a subset of labels. If None, all labels that appear at least once in y1 or y2 are used.
weights [str, optional] List of weighting type to calculate the score. None means no weighted;
“linear” means linear weighted; “quadratic” means quadratic weighted.
Returns
kappa [float] The kappa statistic, which is a number between -1 and 1. The maximum value
means complete agreement; zero or lower means chance agreement.
References
[1], [2], [3]
sklearn.metrics.confusion_matrix
sklearn.metrics.confusion_matrix(y_true, y_pred, labels=None, sample_weight=None)

Compute confusion matrix to evaluate the accuracy of a classification
By definition a confusion matrix 𝐶 is such that 𝐶𝑖,𝑗 is equal to the number of observations known to be in group
𝑖 but predicted to be in group 𝑗.
Thus in binary classification, the count of true negatives is 𝐶0,0 , false negatives is 𝐶1,0 , true positives is 𝐶1,1
and false positives is 𝐶0,1 .
Parameters
y_true [array, shape = [n_samples]] Ground truth (correct) target values.

y_pred [array, shape = [n_samples]] Estimated targets as returned by a classifier.

labels [array, shape = [n_classes], optional] List of labels to index the matrix. This may be used
to reorder or select a subset of labels. If none is given, those that appear at least once in
y_true or y_pred are used in sorted order.
Returns
C [array, shape = [n_classes, n_classes]] Confusion matrix
References
[1]
Examples

>>> y_true = [2, 0, 2, 2, 0, 1]
>>> y_pred = [0, 0, 2, 2, 0, 2]
>>> confusion_matrix(y_true, y_pred)
array([[2, 0, 0],
[0, 0, 1],
[1, 0, 2]])
>>> y_true = ["cat", "ant", "cat", "cat", "ant", "bird"]

>>> y_pred = ["ant", "ant", "cat", "cat", "ant", "cat"]
>>> confusion_matrix(y_true, y_pred, labels=["ant", "bird", "cat"])
array([[2, 0, 0],
[0, 0, 1],
[1, 0, 2]])
In the binary case, we can extract true positives, etc as follows:
>>> tn, fp, fn, tp = confusion_matrix([0, 1, 0, 1], [1, 1, 1, 0]).ravel()

>>> (tn, fp, fn, tp)
(0, 2, 1, 1)
Examples using sklearn.metrics.confusion_matrix


sklearn.metrics.f1_score
sklearn.metrics.f1_score(y_true, y_pred, labels=None, pos_label=1, average=’binary’, sam-

ple_weight=None)
Compute the F1 score, also known as balanced F-score or F-measure
The F1 score can be interpreted as a weighted average of the precision and recall, where an F1 score reaches its
best value at 1 and worst score at 0. The relative contribution of precision and recall to the F1 score are equal.
The formula for the F1 score is:
F1 = 2 * (precision * recall) / (precision + recall)
In the multi-class and multi-label case, this is the average of the F1 score of each class with weighting depending
on the average parameter.
Parameters
values.
a classifier.
labels [list, optional] The set of labels to include when average != 'binary', and their
order if average is None. Labels present in the data can be excluded, for example to
calculate a multiclass average ignoring a majority negative class, while labels not present in
the data will result in 0 components in a macro average. For multilabel targets, labels are
column indices. By default, all labels in y_true and y_pred are used in sorted order.
Changed in version 0.17: parameter labels improved for multiclass problem.
pos_label [str or int, 1 by default] The class to report if average='binary' and the
data is binary. If the data are multiclass or multilabel, this will be ignored; setting
labels=[pos_label] and average != 'binary' will report scores for that la-
bel only.
average [string, [None, ‘binary’ (default), ‘micro’, ‘macro’, ‘samples’, ‘weighted’]] This pa-
rameter is required for multiclass/multilabel targets. If None, the scores for each class are
returned. Otherwise, this determines the type of averaging performed on the data:
'binary': Only report results for the class specified by pos_label. This is applicable
only if targets (y_{true,pred}) are binary.
'micro': Calculate metrics globally by counting the total true positives, false negatives
and false positives.
'weighted': Calculate metrics for each label, and find their average weighted by support
(the number of true instances for each label). This alters ‘macro’ to account for label
imbalance; it can result in an F-score that is not between precision and recall.
'samples': Calculate metrics for each instance, and find their average (only meaningful
for multilabel classification where this differs from accuracy_score).
Returns

f1_score [float or array of float, shape = [n_unique_labels]] F1 score of the positive class in
binary classification or weighted average of the F1 scores of each class for the multiclass
task.
References
[1]
Examples
>>> from sklearn.metrics import f1_score

>>> y_true = [0, 1, 2, 0, 1, 2]
>>> y_pred = [0, 2, 1, 0, 0, 1]
>>> f1_score(y_true, y_pred, average='macro')
0.26...
>>> f1_score(y_true, y_pred, average='micro')
0.33...
>>> f1_score(y_true, y_pred, average='weighted')
0.26...
>>> f1_score(y_true, y_pred, average=None)
array([0.8, 0. , 0. ])
Examples using sklearn.metrics.f1_score
sklearn.metrics.fbeta_score
sklearn.metrics.fbeta_score(y_true, y_pred, beta, labels=None, pos_label=1, average=’binary’,

sample_weight=None)
Compute the F-beta score
The F-beta score is the weighted harmonic mean of precision and recall, reaching its optimal value at 1 and its
worst value at 0.
The beta parameter determines the weight of precision in the combined score. beta < 1 lends more weight to
precision, while beta > 1 favors recall (beta -> 0 considers only precision, beta -> inf only recall).
Parameters
values.
a classifier.
beta [float] Weight of precision in harmonic mean.


bel only.
Returns
fbeta_score [float (if average is not None) or array of float, shape = [n_unique_labels]] F-beta
score of the positive class in binary classification or weighted average of the F-beta score of
each class for the multiclass task.
References
[1], [2]
Examples
>>> from sklearn.metrics import fbeta_score

>>> y_true = [0, 1, 2, 0, 1, 2]
>>> y_pred = [0, 2, 1, 0, 0, 1]
>>> fbeta_score(y_true, y_pred, average='macro', beta=0.5)
...
0.23...
>>> fbeta_score(y_true, y_pred, average='micro', beta=0.5)
...
0.33...
>>> fbeta_score(y_true, y_pred, average='weighted', beta=0.5)
...
0.23...
>>> fbeta_score(y_true, y_pred, average=None, beta=0.5)
...
array([0.71..., 0. , 0. ])

sklearn.metrics.hamming_loss
sklearn.metrics.hamming_loss(y_true, y_pred, labels=None, sample_weight=None)

Compute the average Hamming loss.
The Hamming loss is the fraction of labels that are incorrectly predicted.
Parameters
a classifier.
labels [array, shape = [n_labels], optional (default=None)] Integer array of labels. If not pro-
vided, labels will be inferred from y_true and y_pred.
Returns
loss [float or int,] Return the average Hamming loss between element of y_true and y_pred.
See also:
accuracy_score, jaccard_similarity_score, zero_one_loss
Notes
In multiclass classification, the Hamming loss corresponds to the Hamming distance between y_true and
y_pred which is equivalent to the subset zero_one_loss function.
In multilabel classification, the Hamming loss is different from the subset zero-one loss. The zero-one loss
considers the entire set of labels for a given sample incorrect if it does entirely match the true set of labels.
Hamming loss is more forgiving in that it penalizes the individual labels.
The Hamming loss is upperbounded by the subset zero-one loss. When normalized over samples, the Hamming
loss is always between 0 and 1.
References
[1], [2]
Examples
>>> from sklearn.metrics import hamming_loss

>>> y_pred = [1, 2, 3, 4]
>>> y_true = [2, 2, 3, 4]
>>> hamming_loss(y_true, y_pred)
0.25

>>> hamming_loss(np.array([[0, 1], [1, 1]]), np.zeros((2, 2)))

0.75
Examples using sklearn.metrics.hamming_loss
sklearn.metrics.hinge_loss
sklearn.metrics.hinge_loss(y_true, pred_decision, labels=None, sample_weight=None)

Average hinge loss (non-regularized)
In binary class case, assuming labels in y_true are encoded with +1 and -1, when a prediction mistake is
made, margin = y_true * pred_decision is always negative (since the signs disagree), implying
1 - margin is always greater than 1. The cumulated hinge loss is therefore an upper bound of the number of
mistakes made by the classifier.
In multiclass case, the function expects that either all the labels are included in y_true or an optional labels
argument is provided which contains all the labels. The multilabel margin is calculated according to Crammer-
Singer’s method. As in the binary case, the cumulated hinge loss is an upper bound of the number of mistakes
made by the classifier.
Parameters
y_true [array, shape = [n_samples]] True target, consisting of integers of two values. The
positive label must be greater than the negative label.
pred_decision [array, shape = [n_samples] or [n_samples, n_classes]] Predicted decisions, as
output by decision_function (floats).
labels [array, optional, default None] Contains all the labels for the problem. Used in multiclass
hinge loss.
Returns
loss [float]
References
[1], [2], [3]
Examples

>>> from sklearn.metrics import hinge_loss
>>> X = [[0], [1]]
>>> y = [-1, 1]
>>> est = svm.LinearSVC(random_state=0)
>>> est.fit(X, y)


multi_class='ovr', penalty='l2', random_state=0, tol=0.0001,
verbose=0)
>>> pred_decision = est.decision_function([[-2], [3], [0.5]])
>>> pred_decision
array([-2.18..., 2.36..., 0.09...])
>>> hinge_loss([-1, 1, 1], pred_decision)
0.30...
In the multiclass case:
>>> X = np.array([[0], [1], [2], [3]])

>>> Y = np.array([0, 1, 2, 3])
>>> labels = np.array([0, 1, 2, 3])
>>> est = svm.LinearSVC()
>>> est.fit(X, Y)
verbose=0)
>>> pred_decision = est.decision_function([[-1], [2], [3]])
>>> y_true = [0, 2, 3]
>>> hinge_loss(y_true, pred_decision, labels)
0.56...
sklearn.metrics.jaccard_similarity_score
sklearn.metrics.jaccard_similarity_score(y_true, y_pred, normalize=True, sam-

ple_weight=None)
Jaccard similarity coefficient score
The Jaccard index [1], or Jaccard similarity coefficient, defined as the size of the intersection divided by the size
of the union of two label sets, is used to compare set of predicted labels for a sample to the corresponding set of
labels in y_true.
Parameters
a classifier.
normalize [bool, optional (default=True)] If False, return the sum of the Jaccard similarity
coefficient over the sample set. Otherwise, return the average of Jaccard similarity coeffi-
cient.
Returns
score [float] If normalize == True, return the average Jaccard similarity coefficient, else
it returns the sum of the Jaccard similarity coefficient over the sample set.
The best performance is 1 with normalize == True and the number of samples with
normalize == False.
See also:

accuracy_score, hamming_loss, zero_one_loss
Notes
In binary and multiclass classification, this function is equivalent to the accuracy_score. It differs in the
multilabel classification problem.
References
[1]
Examples

>>> from sklearn.metrics import jaccard_similarity_score
>>> y_pred = [0, 2, 1, 3]
>>> y_true = [0, 1, 2, 3]
>>> jaccard_similarity_score(y_true, y_pred)
0.5
>>> jaccard_similarity_score(y_true, y_pred, normalize=False)
2
>>> jaccard_similarity_score(np.array([[0, 1], [1, 1]]), np.ones((2, 2)))

0.75
Examples using sklearn.metrics.jaccard_similarity_score
sklearn.metrics.log_loss
sklearn.metrics.log_loss(y_true, y_pred, eps=1e-15, normalize=True, sample_weight=None, la-

bels=None)
Log loss, aka logistic loss or cross-entropy loss.
This is the loss function used in (multinomial) logistic regression and extensions of it such as neural networks,
defined as the negative log-likelihood of the true labels given a probabilistic classifier’s predictions. The log loss
is only defined for two or more labels. For a single sample with true label yt in {0,1} and estimated probability
yp that yt = 1, the log loss is
-log P(yt|yp) = -(yt log(yp) + (1 - yt) log(1 - yp))
Parameters
y_true [array-like or label indicator matrix] Ground truth (correct) labels for n_samples sam-
ples.

y_pred [array-like of float, shape = (n_samples, n_classes) or (n_samples,)] Predicted prob-

abilities, as returned by a classifier’s predict_proba method. If y_pred.shape =
(n_samples,) the probabilities provided are assumed to be that of the positive class. The
labels in y_pred are assumed to be ordered alphabetically, as done by preprocessing.
LabelBinarizer.
eps [float] Log loss is undefined for p=0 or p=1, so probabilities are clipped to max(eps, min(1
- eps, p)).
normalize [bool, optional (default=True)] If true, return the mean loss per sample. Otherwise,
return the sum of the per-sample losses.
labels [array-like, optional (default=None)] If not provided, labels will be inferred from y_true.
If labels is None and y_pred has shape (n_samples,) the labels are assumed to be
binary and are inferred from y_true. .. versionadded:: 0.18
Returns
loss [float]
Notes
The logarithm used is the natural logarithm (base-e).
References
C.M. Bishop (2006). Pattern Recognition and Machine Learning. Springer, p. 209.
Examples
>>> log_loss(["spam", "ham", "ham", "spam"],

... [[.1, .9], [.9, .1], [.8, .2], [.35, .65]])
0.21616...
Examples using sklearn.metrics.log_loss

sklearn.metrics.matthews_corrcoef
sklearn.metrics.matthews_corrcoef(y_true, y_pred, sample_weight=None)

Compute the Matthews correlation coefficient (MCC)
The Matthews correlation coefficient is used in machine learning as a measure of the quality of binary and
multiclass classifications. It takes into account true and false positives and negatives and is generally regarded
as a balanced measure which can be used even if the classes are of very different sizes. The MCC is in essence a
correlation coefficient value between -1 and +1. A coefficient of +1 represents a perfect prediction, 0 an average
random prediction and -1 an inverse prediction. The statistic is also known as the phi coefficient. [source:
Wikipedia]

Binary and multiclass labels are supported. Only in the binary case does this relate to information about true
and false positives and negatives. See references below.
Parameters
y_true [array, shape = [n_samples]] Ground truth (correct) target values.
y_pred [array, shape = [n_samples]] Estimated targets as returned by a classifier.
sample_weight [array-like of shape = [n_samples], default None] Sample weights.
Returns
mcc [float] The Matthews correlation coefficient (+1 represents a perfect prediction, 0 an aver-
age random prediction and -1 and inverse prediction).
References
[1], [2], [3], [4]
Examples
>>> from sklearn.metrics import matthews_corrcoef

>>> y_true = [+1, +1, +1, -1]
>>> y_pred = [+1, -1, +1, +1]
>>> matthews_corrcoef(y_true, y_pred)
-0.33...
sklearn.metrics.precision_recall_curve
sklearn.metrics.precision_recall_curve(y_true, probas_pred, pos_label=None, sam-

ple_weight=None)
Compute precision-recall pairs for different probability thresholds
Note: this implementation is restricted to the binary classification task.
The precision is the ratio tp / (tp + fp) where tp is the number of true positives and fp the number of
false positives. The precision is intuitively the ability of the classifier not to label as positive a sample that is
negative.
The recall is the ratio tp / (tp + fn) where tp is the number of true positives and fn the number of false
negatives. The recall is intuitively the ability of the classifier to find all the positive samples.
The last precision and recall values are 1. and 0. respectively and do not have a corresponding threshold. This
ensures that the graph starts on the y axis.
Parameters
y_true [array, shape = [n_samples]] True targets of binary classification in range {-1, 1} or {0,
1}.
probas_pred [array, shape = [n_samples]] Estimated probabilities or decision function.
pos_label [int or str, default=None] The label of the positive class

Returns
precision [array, shape = [n_thresholds + 1]] Precision values such that element i is the preci-
sion of predictions with score >= thresholds[i] and the last element is 1.
recall [array, shape = [n_thresholds + 1]] Decreasing recall values such that element i is the
recall of predictions with score >= thresholds[i] and the last element is 0.
thresholds [array, shape = [n_thresholds <= len(np.unique(probas_pred))]] Increasing thresh-
olds on the decision function used to compute precision and recall.
See also:
average_precision_score Compute average precision from prediction scores

roc_curve Compute Receiver operating characteristic (ROC) curve
Examples

>>> from sklearn.metrics import precision_recall_curve
>>> y_true = np.array([0, 0, 1, 1])
>>> y_scores = np.array([0.1, 0.4, 0.35, 0.8])
>>> precision, recall, thresholds = precision_recall_curve(
... y_true, y_scores)
>>> precision
array([0.66666667, 0.5 , 1. , 1. ])
>>> recall
array([1. , 0.5, 0.5, 0. ])
>>> thresholds
array([0.35, 0.4 , 0.8 ])
Examples using sklearn.metrics.precision_recall_curve
sklearn.metrics.precision_recall_fscore_support
sklearn.metrics.precision_recall_fscore_support(y_true, y_pred, beta=1.0, la-

bels=None, pos_label=1, aver-
age=None, warn_for=(‘precision’,
’recall’, ’f-score’), sam-
ple_weight=None)
Compute precision, recall, F-measure and support for each class
negative.
The F-beta score can be interpreted as a weighted harmonic mean of the precision and recall, where an F-beta
score reaches its best value at 1 and worst score at 0.

The F-beta score weights recall more than precision by a factor of beta. beta == 1.0 means recall and
precision are equally important.
The support is the number of occurrences of each class in y_true.
If pos_label is None and in binary classification, this function returns the average precision, recall and
F-measure if average is one of 'micro', 'macro', 'weighted' or 'samples'.
Parameters
values.
a classifier.
beta [float, 1.0 by default] The strength of recall versus precision in the F-score.
bel only.
average [string, [None (default), ‘binary’, ‘micro’, ‘macro’, ‘samples’, ‘weighted’]] If None,
the scores for each class are returned. Otherwise, this determines the type of averaging
performed on the data:
warn_for [tuple or set, for internal use] This determines which warnings will be made in the
case that this function is being used to return only one of its metrics.
Returns
precision [float (if average is not None) or array of float, shape = [n_unique_labels]]
recall [float (if average is not None) or array of float, , shape = [n_unique_labels]]
fbeta_score [float (if average is not None) or array of float, shape = [n_unique_labels]]

support [int (if average is not None) or array of int, shape = [n_unique_labels]] The number of
occurrences of each label in y_true.
References
[1], [2], [3]
Examples
>>> from sklearn.metrics import precision_recall_fscore_support

>>> y_true = np.array(['cat', 'dog', 'pig', 'cat', 'dog', 'pig'])
>>> y_pred = np.array(['cat', 'pig', 'dog', 'cat', 'cat', 'dog'])
>>> precision_recall_fscore_support(y_true, y_pred, average='macro')
...
(0.22..., 0.33..., 0.26..., None)
>>> precision_recall_fscore_support(y_true, y_pred, average='micro')
...
(0.33..., 0.33..., 0.33..., None)
>>> precision_recall_fscore_support(y_true, y_pred, average='weighted')
...
(0.22..., 0.33..., 0.26..., None)
It is possible to compute per-label precisions, recalls, F1-scores and supports instead of averaging:
>>> precision_recall_fscore_support(y_true, y_pred, average=None,

... labels=['pig', 'dog', 'cat'])
...
(array([0. , 0. , 0.66...]),
array([0., 0., 1.]), array([0. , 0. , 0.8]),
array([2, 2, 2]))
sklearn.metrics.precision_score
sklearn.metrics.precision_score(y_true, y_pred, labels=None, pos_label=1, average=’binary’,

sample_weight=None)
Compute the precision
negative.
The best value is 1 and the worst value is 0.
Parameters
values.
a classifier.

bel only.
Returns
precision [float (if average is not None) or array of float, shape = [n_unique_labels]] Precision
of the positive class in binary classification or weighted average of the precision of each
class for the multiclass task.
Examples
>>> from sklearn.metrics import precision_score

>>> y_true = [0, 1, 2, 0, 1, 2]
>>> y_pred = [0, 2, 1, 0, 0, 1]
>>> precision_score(y_true, y_pred, average='macro')
0.22...
>>> precision_score(y_true, y_pred, average='micro')
0.33...
>>> precision_score(y_true, y_pred, average='weighted')
...
0.22...
>>> precision_score(y_true, y_pred, average=None)
array([0.66..., 0. , 0. ])
Examples using sklearn.metrics.precision_score

sklearn.metrics.recall_score
sklearn.metrics.recall_score(y_true, y_pred, labels=None, pos_label=1, average=’binary’, sam-

ple_weight=None)
Compute the recall
The best value is 1 and the worst value is 0.
Parameters
values.
a classifier.
bel only.
Returns
recall [float (if average is not None) or array of float, shape = [n_unique_labels]] Recall of the
positive class in binary classification or weighted average of the recall of each class for the
multiclass task.

Examples
>>> from sklearn.metrics import recall_score

>>> y_true = [0, 1, 2, 0, 1, 2]
>>> y_pred = [0, 2, 1, 0, 0, 1]
>>> recall_score(y_true, y_pred, average='macro')
0.33...
>>> recall_score(y_true, y_pred, average='micro')
0.33...
>>> recall_score(y_true, y_pred, average='weighted')
0.33...
>>> recall_score(y_true, y_pred, average=None)
array([1., 0., 0.])
Examples using sklearn.metrics.recall_score
sklearn.metrics.roc_auc_score
sklearn.metrics.roc_auc_score(y_true, y_score, average=’macro’, sample_weight=None,

max_fpr=None)
Compute Area Under the Receiver Operating Characteristic Curve (ROC AUC) from prediction scores.
Note: this implementation is restricted to the binary classification task or multilabel classification task in label
indicator format.
Parameters
y_true [array, shape = [n_samples] or [n_samples, n_classes]] True binary labels or binary label
indicators.
y_score [array, shape = [n_samples] or [n_samples, n_classes]] Target scores, can either be
probability estimates of the positive class, confidence values, or non-thresholded measure
of decisions (as returned by “decision_function” on some classifiers). For binary y_true,
y_score is supposed to be the score of the class with greater label.
average [string, [None, ‘micro’, ‘macro’ (default), ‘samples’, ‘weighted’]] If None, the scores
for each class are returned. Otherwise, this determines the type of averaging performed on
the data:
'micro': Calculate metrics globally by considering each element of the label indicator
matrix as a label.
'weighted': Calculate metrics for each label, and find their average, weighted by sup-
port (the number of true instances for each label).
'samples': Calculate metrics for each instance, and find their average.
Will be ignored when y_true is binary.

max_fpr [float > 0 and <= 1, optional] If not None, the standardized partial AUC [3] over the
range [0, max_fpr] is returned.
Returns
auc [float]
See also:
average_precision_score Area under the precision-recall curve

roc_curve Compute Receiver operating characteristic (ROC) curve
References
[1], [2], [3]
Examples

>>> from sklearn.metrics import roc_auc_score
>>> y_true = np.array([0, 0, 1, 1])
>>> y_scores = np.array([0.1, 0.4, 0.35, 0.8])
>>> roc_auc_score(y_true, y_scores)
0.75
sklearn.metrics.roc_curve
sklearn.metrics.roc_curve(y_true, y_score, pos_label=None, sample_weight=None,

drop_intermediate=True)
Compute Receiver operating characteristic (ROC)
Note: this implementation is restricted to the binary classification task.
Parameters
y_true [array, shape = [n_samples]] True binary labels. If labels are not either {-1, 1} or {0, 1},
then pos_label should be explicitly given.
y_score [array, shape = [n_samples]] Target scores, can either be probability estimates of the
positive class, confidence values, or non-thresholded measure of decisions (as returned by
“decision_function” on some classifiers).
pos_label [int or str, default=None] Label considered as positive and others are considered
negative.
drop_intermediate [boolean, optional (default=True)] Whether to drop some suboptimal
thresholds which would not appear on a plotted ROC curve. This is useful in order to
create lighter ROC curves.
New in version 0.17: parameter drop_intermediate.
Returns

fpr [array, shape = [>2]] Increasing false positive rates such that element i is the false positive
rate of predictions with score >= thresholds[i].
tpr [array, shape = [>2]] Increasing true positive rates such that element i is the true positive
rate of predictions with score >= thresholds[i].
thresholds [array, shape = [n_thresholds]] Decreasing thresholds on the decision function used
to compute fpr and tpr. thresholds[0] represents no instances being predicted and is arbi-
trarily set to max(y_score) + 1.
See also:
Notes
Since the thresholds are sorted from low to high values, they are reversed upon returning them to ensure they
correspond to both fpr and tpr, which are sorted in reversed order during their calculation.
References
[1], [2]
Examples

>>> y = np.array([1, 1, 2, 2])
>>> scores = np.array([0.1, 0.4, 0.35, 0.8])
>>> fpr, tpr, thresholds = metrics.roc_curve(y, scores, pos_label=2)
>>> fpr
array([0. , 0. , 0.5, 0.5, 1. ])
>>> tpr
array([0. , 0.5, 0.5, 1. , 1. ])
>>> thresholds
array([1.8 , 0.8 , 0.4 , 0.35, 0.1 ])
Examples using sklearn.metrics.roc_curve

sklearn.metrics.zero_one_loss
sklearn.metrics.zero_one_loss(y_true, y_pred, normalize=True, sample_weight=None)

Zero-one classification loss.

If normalize is True, return the fraction of misclassifications (float), else it returns the number of misclassifica-
tions (int). The best performance is 0.
Parameters
a classifier.
normalize [bool, optional (default=True)] If False, return the number of misclassifications.
Otherwise, return the fraction of misclassifications.
Returns
loss [float or int,] If normalize == True, return the fraction of misclassifications (float),
else it returns the number of misclassifications (int).
See also:
accuracy_score, hamming_loss, jaccard_similarity_score
Notes
In multilabel classification, the zero_one_loss function corresponds to the subset zero-one loss: for each sample,
the entire set of labels must be correctly predicted, otherwise the loss for that sample is equal to one.
Examples
>>> from sklearn.metrics import zero_one_loss

>>> y_pred = [1, 2, 3, 4]
>>> y_true = [2, 2, 3, 4]
>>> zero_one_loss(y_true, y_pred)
0.25
>>> zero_one_loss(y_true, y_pred, normalize=False)
1
>>> zero_one_loss(np.array([[0, 1], [1, 1]]), np.ones((2, 2)))

0.5
Examples using sklearn.metrics.zero_one_loss
6.23.3 Regression metrics
See the Regression metrics section of the user guide for further details.

metrics.explained_variance_score(y_true, Explained variance regression score function

y_pred)
metrics.mean_absolute_error(y_true, y_pred) Mean absolute error regression loss
metrics.mean_squared_error(y_true, y_pred[, Mean squared error regression loss
. . . ])
metrics.mean_squared_log_error(y_true, Mean squared logarithmic error regression loss
y_pred)
metrics.median_absolute_error(y_true, Median absolute error regression loss
y_pred)
metrics.r2_score(y_true, y_pred[, . . . ]) R^2 (coefficient of determination) regression score func-
tion.
sklearn.metrics.explained_variance_score
sklearn.metrics.explained_variance_score(y_true, y_pred, sample_weight=None, multiout-

put=’uniform_average’)
Explained variance regression score function
Best possible score is 1.0, lower values are worse.
Parameters
y_true [array-like of shape = (n_samples) or (n_samples, n_outputs)] Ground truth (correct)
target values.
y_pred [array-like of shape = (n_samples) or (n_samples, n_outputs)] Estimated target values.
sample_weight [array-like of shape = (n_samples), optional] Sample weights.
multioutput [string in [‘raw_values’, ‘uniform_average’, ‘variance_weighted’] or array-like of
shape (n_outputs)] Defines aggregating of multiple output scores. Array-like value defines
weights used to average scores.
‘raw_values’ : Returns a full set of scores in case of multioutput input.
‘uniform_average’ : Scores of all outputs are averaged with uniform weight.
‘variance_weighted’ : Scores of all outputs are averaged, weighted by the variances of
each individual output.
Returns
score [float or ndarray of floats] The explained variance or ndarray if ‘multioutput’ is
‘raw_values’.
Notes
This is not a symmetric function.
Examples
>>> from sklearn.metrics import explained_variance_score

>>> y_true = [3, -0.5, 2, 7]
>>> y_pred = [2.5, 0.0, 2, 8]

>>> explained_variance_score(y_true, y_pred)

0.957...
>>> y_true = [[0.5, 1], [-1, 1], [7, -6]]
>>> y_pred = [[0, 2], [-1, 2], [8, -5]]
>>> explained_variance_score(y_true, y_pred, multioutput='uniform_average')
...
0.983...
sklearn.metrics.mean_absolute_error
sklearn.metrics.mean_absolute_error(y_true, y_pred, sample_weight=None, multiout-

Mean absolute error regression loss
Parameters
target values.
multioutput [string in [‘raw_values’, ‘uniform_average’]] or array-like of shape (n_outputs)
Defines aggregating of multiple output values. Array-like value defines weights used to
average errors.
‘raw_values’ : Returns a full set of errors in case of multioutput input.
‘uniform_average’ : Errors of all outputs are averaged with uniform weight.
Returns
loss [float or ndarray of floats] If multioutput is ‘raw_values’, then mean absolute error is re-
turned for each output separately. If multioutput is ‘uniform_average’ or an ndarray of
weights, then the weighted average of all output errors is returned.
MAE output is non-negative floating point. The best value is 0.0.
Examples
>>> from sklearn.metrics import mean_absolute_error

>>> y_true = [3, -0.5, 2, 7]
>>> y_pred = [2.5, 0.0, 2, 8]
0.5
>>> y_true = [[0.5, 1], [-1, 1], [7, -6]]
>>> y_pred = [[0, 2], [-1, 2], [8, -5]]
0.75
>>> mean_absolute_error(y_true, y_pred, multioutput='raw_values')
array([0.5, 1. ])
>>> mean_absolute_error(y_true, y_pred, multioutput=[0.3, 0.7])
...
0.85...

sklearn.metrics.mean_squared_error
sklearn.metrics.mean_squared_error(y_true, y_pred, sample_weight=None, multiout-

Mean squared error regression loss
Parameters
target values.
multioutput [string in [‘raw_values’, ‘uniform_average’]] or array-like of shape (n_outputs)
average errors.
‘raw_values’ : Returns a full set of errors in case of multioutput input.
Returns
loss [float or ndarray of floats] A non-negative floating point value (the best value is 0.0), or an
array of floating point values, one for each individual target.
Examples

>>> y_true = [3, -0.5, 2, 7]
>>> y_pred = [2.5, 0.0, 2, 8]
0.375
>>> y_true = [[0.5, 1],[-1, 1],[7, -6]]
>>> y_pred = [[0, 2],[-1, 2],[8, -5]]
0.708...
>>> mean_squared_error(y_true, y_pred, multioutput='raw_values')
...
array([0.41666667, 1. ])
>>> mean_squared_error(y_true, y_pred, multioutput=[0.3, 0.7])
...
0.825...
Examples using sklearn.metrics.mean_squared_error


sklearn.metrics.mean_squared_log_error
sklearn.metrics.mean_squared_log_error(y_true, y_pred, sample_weight=None, multiout-

Mean squared logarithmic error regression loss
Parameters
target values.
multioutput [string in [‘raw_values’, ‘uniform_average’] or array-like of shape = (n_outputs)]
average errors.
‘raw_values’ : Returns a full set of errors when the input is of multioutput format.
Returns
loss [float or ndarray of floats] A non-negative floating point value (the best value is 0.0), or an
array of floating point values, one for each individual target.
Examples
>>> from sklearn.metrics import mean_squared_log_error

>>> y_true = [3, 5, 2.5, 7]
>>> y_pred = [2.5, 5, 4, 8]
0.039...
>>> y_true = [[0.5, 1], [1, 2], [7, 6]]
>>> y_pred = [[0.5, 2], [1, 2.5], [8, 8]]
0.044...
>>> mean_squared_log_error(y_true, y_pred, multioutput='raw_values')
...
array([0.00462428, 0.08377444])
>>> mean_squared_log_error(y_true, y_pred, multioutput=[0.3, 0.7])
...
0.060...
sklearn.metrics.median_absolute_error
sklearn.metrics.median_absolute_error(y_true, y_pred)
Median absolute error regression loss
Parameters
y_true [array-like of shape = (n_samples)] Ground truth (correct) target values.
y_pred [array-like of shape = (n_samples)] Estimated target values.

Returns
loss [float] A positive floating point value (the best value is 0.0).
Examples
>>> from sklearn.metrics import median_absolute_error

>>> y_true = [3, -0.5, 2, 7]
>>> y_pred = [2.5, 0.0, 2, 8]
>>> median_absolute_error(y_true, y_pred)
0.5
Examples using sklearn.metrics.median_absolute_error
sklearn.metrics.r2_score
sklearn.metrics.r2_score(y_true, y_pred, sample_weight=None, multioutput=’uniform_average’)

R^2 (coefficient of determination) regression score function.
Best possible score is 1.0 and it can be negative (because the model can be arbitrarily worse). A constant model
that always predicts the expected value of y, disregarding the input features, would get a R^2 score of 0.0.
Parameters
target values.
multioutput [string in [‘raw_values’, ‘uniform_average’, ‘variance_weighted’] or None or
array-like of shape (n_outputs)] Defines aggregating of multiple output scores. Array-like
value defines weights used to average scores. Default is “uniform_average”.
‘raw_values’ : Returns a full set of scores in case of multioutput input.
‘uniform_average’ : Scores of all outputs are averaged with uniform weight.
‘variance_weighted’ : Scores of all outputs are averaged, weighted by the variances of
each individual output.
Changed in version 0.19: Default value of multioutput is ‘uniform_average’.
Returns
z [float or ndarray of floats] The R^2 score or ndarray of scores if ‘multioutput’ is ‘raw_values’.
Notes
This is not a symmetric function.

Unlike most other scores, R^2 score may be negative (it need not actually be the square of a quantity R).

References
[1]
Examples
>>> from sklearn.metrics import r2_score

>>> y_true = [3, -0.5, 2, 7]
>>> y_pred = [2.5, 0.0, 2, 8]
0.948...
>>> y_true = [[0.5, 1], [-1, 1], [7, -6]]
>>> y_pred = [[0, 2], [-1, 2], [8, -5]]
>>> r2_score(y_true, y_pred,
... multioutput='variance_weighted')
0.938...
>>> y_true = [1, 2, 3]
>>> y_pred = [1, 2, 3]
1.0
>>> y_true = [1, 2, 3]
>>> y_pred = [2, 2, 2]
0.0
>>> y_true = [1, 2, 3]
>>> y_pred = [3, 2, 1]
-3.0
Examples using sklearn.metrics.r2_score

6.23.4 Multilabel ranking metrics
See the Multilabel ranking metrics section of the user guide for further details.
metrics.coverage_error(y_true, y_score[, . . . ]) Coverage error measure

metrics.label_ranking_average_precision_score(. Compute
. . )ranking-based average precision
metrics.label_ranking_loss(y_true, y_score) Compute Ranking loss measure
sklearn.metrics.coverage_error
sklearn.metrics.coverage_error(y_true, y_score, sample_weight=None)

Coverage error measure
Compute how far we need to go through the ranked scores to cover all true labels. The best value is equal to the
average number of labels in y_true per sample.

Ties in y_scores are broken by giving maximal rank that would have been assigned to all tied values.
Note: Our implementation’s score is 1 greater than the one given in Tsoumakas et al., 2010. This extends it to
handle the degenerate case in which an instance has 0 true labels.
Parameters
y_true [array, shape = [n_samples, n_labels]] True binary labels in binary indicator format.
y_score [array, shape = [n_samples, n_labels]] Target scores, can either be probability esti-
mates of the positive class, confidence values, or non-thresholded measure of decisions (as
returned by “decision_function” on some classifiers).
Returns
coverage_error [float]
References
[1]
sklearn.metrics.label_ranking_average_precision_score
sklearn.metrics.label_ranking_average_precision_score(y_true, y_score, sam-

ple_weight=None)
Compute ranking-based average precision
Label ranking average precision (LRAP) is the average over each ground truth label assigned to each sample, of
the ratio of true vs. total labels with lower score.
This metric is used in multilabel ranking problem, where the goal is to give better rank to the labels associated
to each sample.
The obtained score is always strictly greater than 0 and the best value is 1.
Parameters
y_true [array or sparse matrix, shape = [n_samples, n_labels]] True binary labels in binary
indicator format.
Returns
score [float]
Examples


>>> from sklearn.metrics import label_ranking_average_precision_score
>>> y_true = np.array([[1, 0, 0], [0, 0, 1]])
>>> y_score = np.array([[0.75, 0.5, 1], [1, 0.2, 0.1]])
>>> label_ranking_average_precision_score(y_true, y_score)
0.416...
sklearn.metrics.label_ranking_loss
sklearn.metrics.label_ranking_loss(y_true, y_score, sample_weight=None)

Compute Ranking loss measure
Compute the average number of label pairs that are incorrectly ordered given y_score weighted by the size of
the label set and the number of labels not in the label set.
This is similar to the error set size, but weighted by the number of relevant and irrelevant labels. The best
performance is achieved with a ranking loss of zero.
New in version 0.17: A function label_ranking_loss
Parameters
y_true [array or sparse matrix, shape = [n_samples, n_labels]] True binary labels in binary
indicator format.
Returns
loss [float]
References
[1]
6.23.5 Clustering metrics
See the Clustering performance evaluation section of the user guide for further details. The sklearn.metrics.
cluster submodule contains evaluation metrics for cluster analysis results. There are two forms of evaluation:
• supervised, which uses a ground truth class values for each sample.
• unsupervised, which does not and measures the ‘quality’ of the model itself.
metrics.adjusted_mutual_info_score(. . . [, Adjusted Mutual Information between two clusterings.

. . . ])
metrics.adjusted_rand_score(labels_true, . . . ) Rand index adjusted for chance.
metrics.calinski_harabaz_score(X, labels) Compute the Calinski and Harabaz score.
metrics.davies_bouldin_score(X, labels) Computes the Davies-Bouldin score.


metrics.completeness_score(labels_true, . . . ) Completeness metric of a cluster labeling given a ground
truth.
metrics.cluster.contingency_matrix(. . . [, Build a contingency matrix describing the relationship be-
. . . ]) tween labels.
metrics.fowlkes_mallows_score(labels_true, Measure the similarity of two clusterings of a set of points.
...)
metrics.homogeneity_completeness_v_measure(. Compute
..) the homogeneity and completeness and V-
Measure scores at once.
metrics.homogeneity_score(labels_true, . . . ) Homogeneity metric of a cluster labeling given a ground
truth.
metrics.mutual_info_score(labels_true, . . . ) Mutual Information between two clusterings.
metrics.normalized_mutual_info_score(. . . [, Normalized Mutual Information between two clusterings.
. . . ])
metrics.silhouette_score(X, labels[, . . . ]) Compute the mean Silhouette Coefficient of all samples.
metrics.silhouette_samples(X, labels[, metric]) Compute the Silhouette Coefficient for each sample.
metrics.v_measure_score(labels_true, la- V-measure cluster labeling given a ground truth.
bels_pred)
sklearn.metrics.adjusted_mutual_info_score
sklearn.metrics.adjusted_mutual_info_score(labels_true, labels_pred, aver-

age_method=’warn’)
Adjusted Mutual Information between two clusterings.
Adjusted Mutual Information (AMI) is an adjustment of the Mutual Information (MI) score to account for
chance. It accounts for the fact that the MI is generally higher for two clusterings with a larger number of
clusters, regardless of whether there is actually more information shared. For two clusterings 𝑈 and 𝑉 , the AMI
is given as:
AMI(U, V) = [MI(U, V) - E(MI(U, V))] / [avg(H(U), H(V)) - E(MI(U, V))]
This metric is independent of the absolute values of the labels: a permutation of the class or cluster label values
won’t change the score value in any way.
This metric is furthermore symmetric: switching label_true with label_pred will return the same score
value. This can be useful to measure the agreement of two independent label assignments strategies on the same
dataset when the real ground truth is not known.
Be mindful that this function is an order of magnitude slower than other metrics, such as the Adjusted Rand
Index.
Parameters
labels_true [int array, shape = [n_samples]] A clustering of the data into disjoint subsets.
labels_pred [array, shape = [n_samples]] A clustering of the data into disjoint subsets.
average_method [string, optional (default: ‘warn’)] How to compute the normalizer in the
denominator. Possible options are ‘min’, ‘geometric’, ‘arithmetic’, and ‘max’. If ‘warn’,
‘max’ will be used. The default will change to ‘arithmetic’ in version 0.22.
Returns

ami: float (upperlimited by 1.0) The AMI returns a value of 1 when the two partitions are
identical (ie perfectly matched). Random partitions (independent labellings) have an ex-
pected AMI around 0 on average hence can be negative.
See also:
adjusted_rand_score Adjusted Rand Index

mutual_info_score Mutual Information (not adjusted for chance)
References
[1], [2]
Examples
Perfect labelings are both homogeneous and complete, hence have score 1.0:
>>> from sklearn.metrics.cluster import adjusted_mutual_info_score

>>> adjusted_mutual_info_score([0, 0, 1, 1], [0, 0, 1, 1])
...
1.0
...
1.0
If classes members are completely split across different clusters, the assignment is totally in-complete, hence
the AMI is null:

...
0.0
Examples using sklearn.metrics.adjusted_mutual_info_score

• Adjustment for chance in clustering performance evaluation
sklearn.metrics.adjusted_rand_score
sklearn.metrics.adjusted_rand_score(labels_true, labels_pred)
Rand index adjusted for chance.
The Rand Index computes a similarity measure between two clusterings by considering all pairs of samples and
counting pairs that are assigned in the same or different clusters in the predicted and true clusterings.
The raw RI score is then “adjusted for chance” into the ARI score using the following scheme:
ARI = (RI - Expected_RI) / (max(RI) - Expected_RI)

The adjusted Rand index is thus ensured to have a value close to 0.0 for random labeling independently of the
number of clusters and samples and exactly 1.0 when the clusterings are identical (up to a permutation).
ARI is a symmetric measure:
adjusted_rand_score(a, b) == adjusted_rand_score(b, a)

Parameters
labels_true [int array, shape = [n_samples]] Ground truth class labels to be used as a reference
labels_pred [array, shape = [n_samples]] Cluster labels to evaluate
Returns
ari [float] Similarity score between -1.0 and 1.0. Random labelings have an ARI close to 0.0.
1.0 stands for perfect match.
See also:
adjusted_mutual_info_score Adjusted Mutual Information
References
[Hubert1985], [wk]
Examples
Perfectly matching labelings have a score of 1 even
>>> from sklearn.metrics.cluster import adjusted_rand_score

>>> adjusted_rand_score([0, 0, 1, 1], [0, 0, 1, 1])
1.0
1.0
Labelings that assign all classes members to the same clusters are complete be not always pure, hence penalized:

0.57...
ARI is symmetric, so labelings that have pure clusters with members coming from the same classes but unnec-
essary splits are penalized:

0.57...
If classes members are completely split across different clusters, the assignment is totally incomplete, hence the
ARI is very low:

0.0

Examples using sklearn.metrics.adjusted_rand_score

sklearn.metrics.calinski_harabaz_score
sklearn.metrics.calinski_harabaz_score(X, labels)
Compute the Calinski and Harabaz score.
It is also known as the Variance Ratio Criterion.
The score is defined as ratio between the within-cluster dispersion and the between-cluster dispersion.
Parameters
X [array-like, shape (n_samples, n_features)] List of n_features-dimensional data
points. Each row corresponds to a single data point.
labels [array-like, shape (n_samples,)] Predicted labels for each sample.
Returns
score [float] The resulting Calinski-Harabaz score.
References
[1]
sklearn.metrics.davies_bouldin_score
sklearn.metrics.davies_bouldin_score(X, labels)
Computes the Davies-Bouldin score.
The score is defined as the ratio of within-cluster distances to between-cluster distances.
Parameters
X [array-like, shape (n_samples, n_features)] List of n_features-dimensional data
points. Each row corresponds to a single data point.
labels [array-like, shape (n_samples,)] Predicted labels for each sample.
Returns
score: float The resulting Davies-Bouldin score.

References
[1]
sklearn.metrics.completeness_score
sklearn.metrics.completeness_score(labels_true, labels_pred)
Completeness metric of a cluster labeling given a ground truth.
A clustering result satisfies completeness if all the data points that are members of a given class are elements of
the same cluster.
This metric is not symmetric: switching label_true with label_pred will return the
homogeneity_score which will be different in general.
Parameters
labels_true [int array, shape = [n_samples]] ground truth class labels to be used as a reference
labels_pred [array, shape = [n_samples]] cluster labels to evaluate
Returns
completeness [float] score between 0.0 and 1.0. 1.0 stands for perfectly complete labeling
See also:
homogeneity_score, v_measure_score
References
[1]
Examples
Perfect labelings are complete:
>>> from sklearn.metrics.cluster import completeness_score

>>> completeness_score([0, 0, 1, 1], [1, 1, 0, 0])
1.0
Non-perfect labelings that assign all classes members to the same clusters are still complete:
>>> print(completeness_score([0, 0, 1, 1], [0, 0, 0, 0]))

1.0
0.999...
If classes members are split across different clusters, the assignment cannot be complete:


0.0
0.0
Examples using sklearn.metrics.completeness_score

sklearn.metrics.cluster.contingency_matrix
sklearn.metrics.cluster.contingency_matrix(labels_true, labels_pred, eps=None,

sparse=False)
Build a contingency matrix describing the relationship between labels.
Parameters
labels_true [int array, shape = [n_samples]] Ground truth class labels to be used as a reference
labels_pred [array, shape = [n_samples]] Cluster labels to evaluate
eps [None or float, optional.] If a float, that value is added to all values in the contingency
matrix. This helps to stop NaN propagation. If None, nothing is adjusted.
sparse [boolean, optional.] If True, return a sparse CSR continency matrix. If eps is not
None, and sparse is True, will throw ValueError.
Returns
contingency [{array-like, sparse}, shape=[n_classes_true, n_classes_pred]] Matrix 𝐶 such that
𝐶𝑖,𝑗 is the number of samples in true class 𝑖 and in predicted class 𝑗. If eps is None,
the dtype of this array will be integer. If eps is given, the dtype will be float. Will be a
scipy.sparse.csr_matrix if sparse=True.
sklearn.metrics.fowlkes_mallows_score
sklearn.metrics.fowlkes_mallows_score(labels_true, labels_pred, sparse=False)

Measure the similarity of two clusterings of a set of points.
The Fowlkes-Mallows index (FMI) is defined as the geometric mean between of the precision and recall:
FMI = TP / sqrt((TP + FP) * (TP + FN))
Where TP is the number of True Positive (i.e. the number of pair of points that belongs in the same clusters
in both labels_true and labels_pred), FP is the number of False Positive (i.e. the number of pair of
points that belongs in the same clusters in labels_true and not in labels_pred) and FN is the number
of False Negative (i.e the number of pair of points that belongs in the same clusters in labels_pred and not
in labels_True).
The score ranges from 0 to 1. A high value indicates a good similarity between two clusters.


Parameters
labels_true [int array, shape = (n_samples,)] A clustering of the data into disjoint subsets.
labels_pred [array, shape = (n_samples, )] A clustering of the data into disjoint subsets.
sparse [bool] Compute contingency matrix internally with sparse matrix.
Returns
score [float] The resulting Fowlkes-Mallows score.
References
[1], [2]
Examples
>>> from sklearn.metrics.cluster import fowlkes_mallows_score

>>> fowlkes_mallows_score([0, 0, 1, 1], [0, 0, 1, 1])
1.0
1.0
If classes members are completely split across different clusters, the assignment is totally random, hence the
FMI is null:

0.0
sklearn.metrics.homogeneity_completeness_v_measure
sklearn.metrics.homogeneity_completeness_v_measure(labels_true, labels_pred)
Compute the homogeneity and completeness and V-Measure scores at once.
Those metrics are based on normalized conditional entropy measures of the clustering labeling to evaluate given
the knowledge of a Ground Truth class labels of the same samples.
A clustering result satisfies homogeneity if all of its clusters contain only data points which are members of a
single class.
A clustering result satisfies completeness if all the data points that are members of a given class are elements of
the same cluster.
Both scores have positive values between 0.0 and 1.0, larger values being desirable.
Those 3 metrics are independent of the absolute values of the labels: a permutation of the class or cluster label
values won’t change the score values in any way.
V-Measure is furthermore symmetric: swapping labels_true and label_pred will give the
same score. This does not hold for homogeneity and completeness. V-Measure is identical to
normalized_mutual_info_score with the arithmetic averaging method.

Parameters
Returns
homogeneity [float] score between 0.0 and 1.0. 1.0 stands for perfectly homogeneous labeling
completeness [float] score between 0.0 and 1.0. 1.0 stands for perfectly complete labeling
v_measure [float] harmonic mean of the first two
See also:
homogeneity_score, completeness_score, v_measure_score
sklearn.metrics.homogeneity_score
sklearn.metrics.homogeneity_score(labels_true, labels_pred)
Homogeneity metric of a cluster labeling given a ground truth.
A clustering result satisfies homogeneity if all of its clusters contain only data points which are members of a
single class.
This metric is not symmetric: switching label_true with label_pred will return the
completeness_score which will be different in general.
Parameters
Returns
homogeneity [float] score between 0.0 and 1.0. 1.0 stands for perfectly homogeneous labeling
See also:
completeness_score, v_measure_score
References
[1]
Examples
Perfect labelings are homogeneous:
>>> from sklearn.metrics.cluster import homogeneity_score

>>> homogeneity_score([0, 0, 1, 1], [1, 1, 0, 0])
1.0
Non-perfect labelings that further split classes into more clusters can be perfectly homogeneous:

>>> print("%.6f" % homogeneity_score([0, 0, 1, 1], [0, 0, 1, 2]))

...
1.000000
...
1.000000
Clusters that include samples from different classes do not make for an homogeneous labeling:

...
0.0...
...
0.0...
Examples using sklearn.metrics.homogeneity_score

sklearn.metrics.mutual_info_score
sklearn.metrics.mutual_info_score(labels_true, labels_pred, contingency=None)

Mutual Information between two clusterings.
The Mutual Information is a measure of the similarity between two labels of the same data. Where |𝑈𝑖 | is the
number of the samples in cluster 𝑈𝑖 and |𝑉𝑗 | is the number of the samples in cluster 𝑉𝑗 , the Mutual Information
between clusterings 𝑈 and 𝑉 is given as:
|𝑈 | |𝑉 |
∑︁ ∑︁ |𝑈𝑖 ∩ 𝑉𝑗 | 𝑁 |𝑈𝑖 ∩ 𝑉𝑗 |
𝑀 𝐼(𝑈, 𝑉 ) = log
𝑖=1 𝑗=1
𝑁 |𝑈𝑖 ||𝑉𝑗 |
Parameters
contingency [{None, array, sparse matrix}, shape = [n_classes_true, n_classes_pred]] A con-
tingency matrix given by the contingency_matrix function. If value is None, it will
be computed, otherwise the given value is used, with labels_true and labels_pred
ignored.

Returns
mi [float] Mutual information, a non-negative value
See also:
adjusted_mutual_info_score Adjusted against chance Mutual Information

normalized_mutual_info_score Normalized Mutual Information
Examples using sklearn.metrics.mutual_info_score
sklearn.metrics.normalized_mutual_info_score
sklearn.metrics.normalized_mutual_info_score(labels_true, labels_pred, aver-

age_method=’warn’)
Normalized Mutual Information between two clusterings.
Normalized Mutual Information (NMI) is a normalization of the Mutual Information (MI) score to scale the
results between 0 (no mutual information) and 1 (perfect correlation). In this function, mutual information
is normalized by some generalized mean of H(labels_true) and H(labels_pred)), defined by the
average_method.
This measure is not adjusted for chance. Therefore adjusted_mutual_info_score might be preferred.
Parameters
average_method [string, optional (default: ‘warn’)] How to compute the normalizer in the
denominator. Possible options are ‘min’, ‘geometric’, ‘arithmetic’, and ‘max’. If ‘warn’,
‘geometric’ will be used. The default will change to ‘arithmetic’ in version 0.22.
Returns
nmi [float] score between 0.0 and 1.0. 1.0 stands for perfectly complete labeling
See also:
v_measure_score V-Measure (NMI with arithmetic mean option.)

adjusted_rand_score Adjusted Rand Index
adjusted_mutual_info_score Adjusted Mutual Information (adjusted against chance)

Examples
>>> from sklearn.metrics.cluster import normalized_mutual_info_score

>>> normalized_mutual_info_score([0, 0, 1, 1], [0, 0, 1, 1])
...
1.0
>>> normalized_mutual_info_score([0, 0, 1, 1], [1, 1, 0, 0])
...
1.0
If classes members are completely split across different clusters, the assignment is totally in-complete, hence
the NMI is null:
>>> normalized_mutual_info_score([0, 0, 0, 0], [0, 1, 2, 3])i

...
0.0
sklearn.metrics.silhouette_score
sklearn.metrics.silhouette_score(X, labels, metric=’euclidean’, sample_size=None, ran-

dom_state=None, **kwds)
Compute the mean Silhouette Coefficient of all samples.
The Silhouette Coefficient is calculated using the mean intra-cluster distance (a) and the mean nearest-cluster
distance (b) for each sample. The Silhouette Coefficient for a sample is (b - a) / max(a, b). To clarify,
b is the distance between a sample and the nearest cluster that the sample is not a part of. Note that Silhouette
Coefficient is only defined if number of labels is 2 <= n_labels <= n_samples - 1.
This function returns the mean Silhouette Coefficient over all samples. To obtain the values for each sample,
use silhouette_samples.
The best value is 1 and the worst value is -1. Values near 0 indicate overlapping clusters. Negative values
generally indicate that a sample has been assigned to the wrong cluster, as a different cluster is more similar.
Parameters
X [array [n_samples_a, n_samples_a] if metric == “precomputed”, or, [n_samples_a,
n_features] otherwise] Array of pairwise distances between samples, or a feature array.
labels [array, shape = [n_samples]] Predicted labels for each sample.
metric [string, or callable] The metric to use when calculating distance between instances
in a feature array. If metric is a string, it must be one of the options allowed by
metrics.pairwise.pairwise_distances. If X is the distance array itself, use
metric="precomputed".
sample_size [int or None] The size of the sample to use when computing the Silhouette Co-
efficient on a random subset of the data. If sample_size is None, no sampling is
used.
random_state [int, RandomState instance or None, optional (default=None)] The generator
used to randomly select a subset of samples. If int, random_state is the seed used by the
random number generator; If RandomState instance, random_state is the random number
generator; If None, the random number generator is the RandomState instance used by
np.random. Used when sample_size is not None.

**kwds [optional keyword parameters] Any further parameters are passed directly to the dis-
tance function. If using a scipy.spatial.distance metric, the parameters are still metric de-
pendent. See the scipy docs for usage examples.
Returns
silhouette [float] Mean Silhouette Coefficient for all samples.
References
[1], [2]
Examples using sklearn.metrics.silhouette_score

sklearn.metrics.silhouette_samples
sklearn.metrics.silhouette_samples(X, labels, metric=’euclidean’, **kwds)

Compute the Silhouette Coefficient for each sample.
The Silhouette Coefficient is a measure of how well samples are clustered with samples that are similar to
themselves. Clustering models with a high Silhouette Coefficient are said to be dense, where samples in the
same cluster are similar to each other, and well separated, where samples in different clusters are not very
similar to each other.
The Silhouette Coefficient is calculated using the mean intra-cluster distance (a) and the mean nearest-cluster
distance (b) for each sample. The Silhouette Coefficient for a sample is (b - a) / max(a, b). Note that
Silhouette Coefficient is only defined if number of labels is 2 <= n_labels <= n_samples - 1.
This function returns the Silhouette Coefficient for each sample.
The best value is 1 and the worst value is -1. Values near 0 indicate overlapping clusters.
Parameters
labels [array, shape = [n_samples]] label values for each sample
a feature array. If metric is a string, it must be one of the options allowed by sklearn.
metrics.pairwise.pairwise_distances. If X is the distance array itself, use
“precomputed” as the metric.
‘**kwds‘ [optional keyword parameters] Any further parameters are passed directly to the dis-
tance function. If using a scipy.spatial.distance metric, the parameters are still
metric dependent. See the scipy docs for usage examples.

Returns
silhouette [array, shape = [n_samples]] Silhouette Coefficient for each samples.
References
[1], [2]
Examples using sklearn.metrics.silhouette_samples
sklearn.metrics.v_measure_score
sklearn.metrics.v_measure_score(labels_true, labels_pred)
V-measure cluster labeling given a ground truth.
This score is identical to normalized_mutual_info_score with the 'arithmetic' option for aver-
aging.
The V-measure is the harmonic mean between homogeneity and completeness:
v = 2 * (homogeneity * completeness) / (homogeneity + completeness)
Parameters
Returns
v_measure [float] score between 0.0 and 1.0. 1.0 stands for perfectly complete labeling
See also:
homogeneity_score, completeness_score, normalized_mutual_info_score
References
[1]
Examples

>>> from sklearn.metrics.cluster import v_measure_score

>>> v_measure_score([0, 0, 1, 1], [0, 0, 1, 1])
1.0
>>> v_measure_score([0, 0, 1, 1], [1, 1, 0, 0])
1.0
Labelings that assign all classes members to the same clusters are complete be not homogeneous, hence penal-
ized:
>>> print("%.6f" % v_measure_score([0, 0, 1, 2], [0, 0, 1, 1]))
...
0.8...
...
0.66...
Labelings that have pure clusters with members coming from the same classes are homogeneous but un-
necessary splits harms completeness and thus penalize V-measure as well:
...
0.8...
...
0.66...
If classes members are completely split across different clusters, the assignment is totally incomplete, hence the
V-Measure is null:
...
0.0...
Clusters that include samples from totally different classes totally destroy the homogeneity of the labeling,
hence:
...
0.0...
Examples using sklearn.metrics.v_measure_score

6.23.6 Biclustering metrics
See the Biclustering evaluation section of the user guide for further details.

metrics.consensus_score(a, b[, similarity]) The similarity of two sets of biclusters.
sklearn.metrics.consensus_score
sklearn.metrics.consensus_score(a, b, similarity=’jaccard’)
The similarity of two sets of biclusters.
Similarity between individual biclusters is computed. Then the best matching between sets is found using the
Hungarian algorithm. The final score is the sum of similarities divided by the size of the larger set.
Parameters
a [(rows, columns)] Tuple of row and column indicators for a set of biclusters.
b [(rows, columns)] Another set of biclusters like a.
similarity [string or function, optional, default: “jaccard”] May be the string “jaccard” to use
the Jaccard coefficient, or any function that takes four arguments, each of which is a 1d
indicator vector: (a_rows, a_columns, b_rows, b_columns).
References
• Hochreiter, Bodenhofer, et. al., 2010. FABIA: factor analysis for bicluster acquisition.
Examples using sklearn.metrics.consensus_score
• A demo of the Spectral Co-Clustering algorithm

• A demo of the Spectral Biclustering algorithm
6.23.7 Pairwise metrics
See the Pairwise metrics, Affinities and Kernels section of the user guide for further details.
metrics.pairwise.additive_chi2_kernel(X[, Computes the additive chi-squared kernel between obser-

Y]) vations in X and Y
metrics.pairwise.chi2_kernel(X[, Y, gamma]) Computes the exponential chi-squared kernel X and Y.
metrics.pairwise.cosine_similarity(X[, Y, Compute cosine similarity between samples in X and Y.
. . . ])
metrics.pairwise.cosine_distances(X[, Y]) Compute cosine distance between samples in X and Y.
metrics.pairwise.distance_metrics() Valid metrics for pairwise_distances.
metrics.pairwise.euclidean_distances(X[, Considering the rows of X (and Y=X) as vectors, compute
Y, . . . ]) the distance matrix between each pair of vectors.
metrics.pairwise.kernel_metrics() Valid metrics for pairwise_kernels
metrics.pairwise.laplacian_kernel(X[, Y, Compute the laplacian kernel between X and Y.
gamma])
metrics.pairwise.linear_kernel(X[, Y, . . . ]) Compute the linear kernel between X and Y.
metrics.pairwise.manhattan_distances(X[, Compute the L1 distances between the vectors in X and Y.
Y, . . . ])


metrics.pairwise.pairwise_kernels(X[, Y, Compute the kernel between arrays X and optional array Y.
. . . ])
metrics.pairwise.polynomial_kernel(X[, Y, Compute the polynomial kernel between X and Y:
. . . ])
metrics.pairwise.rbf_kernel(X[, Y, gamma]) Compute the rbf (gaussian) kernel between X and Y:
metrics.pairwise.sigmoid_kernel(X[, Y, . . . ]) Compute the sigmoid kernel between X and Y:
metrics.pairwise.paired_euclidean_distances(X, Computes the paired euclidean distances between X and Y
Y)
metrics.pairwise.paired_manhattan_distances(X, Compute the L1 distances between the vectors in X and Y.
Y)
metrics.pairwise.paired_cosine_distances(X, Computes the paired cosine distances between X and Y
Y)
metrics.pairwise.paired_distances(X, Y[, Computes the paired distances between X and Y.
metric])
metrics.pairwise_distances(X[, Y, metric, . . . ]) Compute the distance matrix from a vector array X and op-
tional Y.
metrics.pairwise_distances_argmin(X, Y[, Compute minimum distances between one point and a set
. . . ]) of points.
metrics.pairwise_distances_argmin_min(X, Compute minimum distances between one point and a set
Y) of points.
metrics.pairwise_distances_chunked(X[, Y, Generate a distance matrix chunk by chunk with optional
. . . ]) reduction
sklearn.metrics.pairwise.additive_chi2_kernel
sklearn.metrics.pairwise.additive_chi2_kernel(X, Y=None)
Computes the additive chi-squared kernel between observations in X and Y
The chi-squared kernel is computed between each pair of rows in X and Y. X and Y have to be non-negative.
This kernel is most commonly applied to histograms.
The chi-squared kernel is given by:
k(x, y) = -Sum [(x - y)^2 / (x + y)]
It can be interpreted as a weighted difference per entry.

Parameters
X [array-like of shape (n_samples_X, n_features)]
Y [array of shape (n_samples_Y, n_features)]
Returns
kernel_matrix [array of shape (n_samples_X, n_samples_Y)]
See also:
chi2_kernel The exponentiated version of the kernel, which is usually preferable.

sklearn.kernel_approximation.AdditiveChi2Sampler A Fourier approximation to this ker-
nel.

Notes
As the negative of a distance, this kernel is only conditionally positive definite.
References
sklearn.metrics.pairwise.chi2_kernel
sklearn.metrics.pairwise.chi2_kernel(X, Y=None, gamma=1.0)

Computes the exponential chi-squared kernel X and Y.
The chi-squared kernel is computed between each pair of rows in X and Y. X and Y have to be non-negative.
This kernel is most commonly applied to histograms.
The chi-squared kernel is given by:
k(x, y) = exp(-gamma Sum [(x - y)^2 / (x + y)])
It can be interpreted as a weighted difference per entry.

Parameters
X [array-like of shape (n_samples_X, n_features)]
gamma [float, default=1.] Scaling parameter of the chi2 kernel.
Returns
See also:
additive_chi2_kernel The additive version of this kernel

sklearn.kernel_approximation.AdditiveChi2Sampler A Fourier approximation to the addi-
tive version of this kernel.
References
sklearn.metrics.pairwise.cosine_similarity
sklearn.metrics.pairwise.cosine_similarity(X, Y=None, dense_output=True)

Compute cosine similarity between samples in X and Y.
Cosine similarity, or the cosine kernel, computes similarity as the normalized dot product of X and Y:

K(X, Y) = <X, Y> / (||X||*||Y||)

On L2-normalized data, this function is equivalent to linear_kernel.
Parameters
X [ndarray or sparse array, shape: (n_samples_X, n_features)] Input data.
Y [ndarray or sparse array, shape: (n_samples_Y, n_features)] Input data. If None, the output
will be the pairwise similarities between all samples in X.
dense_output [boolean (optional), default True] Whether to return dense output even when the
input is sparse. If False, the output is sparse if both input arrays are sparse.
New in version 0.17: parameter dense_output for dense output.
Returns
kernel matrix [array] An array with shape (n_samples_X, n_samples_Y).
sklearn.metrics.pairwise.cosine_distances
sklearn.metrics.pairwise.cosine_distances(X, Y=None)
Compute cosine distance between samples in X and Y.
Cosine distance is defined as 1.0 minus the cosine similarity.
Parameters
X [array_like, sparse matrix] with shape (n_samples_X, n_features).
Y [array_like, sparse matrix (optional)] with shape (n_samples_Y, n_features).
Returns
distance matrix [array] An array with shape (n_samples_X, n_samples_Y).
See also:
sklearn.metrics.pairwise.cosine_similarity, scipy.spatial.distance.cosine
sklearn.metrics.pairwise.distance_metrics
sklearn.metrics.pairwise.distance_metrics()
Valid metrics for pairwise_distances.
This function simply returns the valid pairwise distance metrics. It exists to allow for a description of the
mapping for each of the valid strings.
The valid distance metrics, and the function they map to, are:
metric Function
‘cityblock’ metrics.pairwise.manhattan_distances
‘cosine’ metrics.pairwise.cosine_distances
‘euclidean’ metrics.pairwise.euclidean_distances
‘l1’ metrics.pairwise.manhattan_distances
‘l2’ metrics.pairwise.euclidean_distances
‘manhattan’ metrics.pairwise.manhattan_distances

sklearn.metrics.pairwise.euclidean_distances
sklearn.metrics.pairwise.euclidean_distances(X, Y=None, Y_norm_squared=None,

squared=False, X_norm_squared=None)
Considering the rows of X (and Y=X) as vectors, compute the distance matrix between each pair of vectors.
For efficiency reasons, the euclidean distance between a pair of row vector x and y is computed as:
dist(x, y) = sqrt(dot(x, x) - 2 * dot(x, y) + dot(y, y))
This formulation has two advantages over other ways of computing distances. First, it is computationally ef-
ficient when dealing with sparse data. Second, if one argument varies but the other remains unchanged, then
dot(x, x) and/or dot(y, y) can be pre-computed.
However, this is not the most precise way of doing this computation, and the distance matrix returned by this
function may not be exactly symmetric as required by, e.g., scipy.spatial.distance functions.
Parameters
X [{array-like, sparse matrix}, shape (n_samples_1, n_features)]
Y [{array-like, sparse matrix}, shape (n_samples_2, n_features)]
Y_norm_squared [array-like, shape (n_samples_2, ), optional] Pre-computed dot-products of
vectors in Y (e.g., (Y**2).sum(axis=1))
squared [boolean, optional] Return squared Euclidean distances.
X_norm_squared [array-like, shape = [n_samples_1], optional] Pre-computed dot-products of
vectors in X (e.g., (X**2).sum(axis=1))
Returns
distances [{array, sparse matrix}, shape (n_samples_1, n_samples_2)]
See also:
paired_distances distances betweens pairs of elements of X and Y.
Examples
>>> from sklearn.metrics.pairwise import euclidean_distances

>>> X = [[0, 1], [1, 1]]
>>> # distance between rows of X
>>> euclidean_distances(X, X)
array([[0., 1.],
[1., 0.]])
>>> # get distance to origin
>>> euclidean_distances(X, [[0, 0]])
array([[1. ],
[1.41421356]])

sklearn.metrics.pairwise.kernel_metrics
sklearn.metrics.pairwise.kernel_metrics()
Valid metrics for pairwise_kernels
This function simply returns the valid pairwise distance metrics. It exists, however, to allow for a verbose
description of the mapping for each of the valid strings.
The valid distance metrics, and the function they map to, are:
metric Function
‘additive_chi2’ sklearn.pairwise.additive_chi2_kernel
‘chi2’ sklearn.pairwise.chi2_kernel
‘linear’ sklearn.pairwise.linear_kernel
‘poly’ sklearn.pairwise.polynomial_kernel
‘polynomial’ sklearn.pairwise.polynomial_kernel
‘rbf’ sklearn.pairwise.rbf_kernel
‘laplacian’ sklearn.pairwise.laplacian_kernel
‘sigmoid’ sklearn.pairwise.sigmoid_kernel
‘cosine’ sklearn.pairwise.cosine_similarity
sklearn.metrics.pairwise.laplacian_kernel
sklearn.metrics.pairwise.laplacian_kernel(X, Y=None, gamma=None)

Compute the laplacian kernel between X and Y.
The laplacian kernel is defined as:
K(x, y) = exp(-gamma ||x-y||_1)
for each pair of rows x in X and y in Y. Read more in the User Guide.
Parameters
X [array of shape (n_samples_X, n_features)]
gamma [float, default None] If None, defaults to 1.0 / n_features
Returns
sklearn.metrics.pairwise.linear_kernel
sklearn.metrics.pairwise.linear_kernel(X, Y=None, dense_output=True)

Compute the linear kernel between X and Y.
Parameters
X [array of shape (n_samples_1, n_features)]

Y [array of shape (n_samples_2, n_features)]

dense_output [boolean (optional), default True] Whether to return dense output even when the
input is sparse. If False, the output is sparse if both input arrays are sparse.
Returns
Gram matrix [array of shape (n_samples_1, n_samples_2)]
sklearn.metrics.pairwise.manhattan_distances
sklearn.metrics.pairwise.manhattan_distances(X, Y=None, sum_over_features=True,

size_threshold=None)
Compute the L1 distances between the vectors in X and Y.
With sum_over_features equal to False it returns the componentwise distances.
Parameters
X [array_like] An array with shape (n_samples_X, n_features).
Y [array_like, optional] An array with shape (n_samples_Y, n_features).
sum_over_features [bool, default=True] If True the function returns the pairwise distance ma-
trix else it returns the componentwise L1 pairwise-distances. Not supported for sparse ma-
trix inputs.
size_threshold [int, default=5e8] Unused parameter.
Returns
D [array] If sum_over_features is False shape is (n_samples_X * n_samples_Y, n_features) and
D contains the componentwise L1 pairwise-distances (ie. absolute difference), else shape is
(n_samples_X, n_samples_Y) and D contains the pairwise L1 distances.
Examples
>>> from sklearn.metrics.pairwise import manhattan_distances

>>> manhattan_distances([[3]], [[3]])
array([[0.]])
array([[1.]])
array([[1.]])
>>> manhattan_distances([[1, 2], [3, 4]], [[1, 2], [0, 3]])
array([[0., 2.],
[4., 4.]])
>>> X = np.ones((1, 2))
>>> y = np.full((2, 2), 2.)
>>> manhattan_distances(X, y, sum_over_features=False)
array([[1., 1.],
[1., 1.]])

sklearn.metrics.pairwise.pairwise_kernels
sklearn.metrics.pairwise.pairwise_kernels(X, Y=None, metric=’linear’, fil-

ter_params=False, n_jobs=None, **kwds)
Compute the kernel between arrays X and optional array Y.
This method takes either a vector array or a kernel matrix, and returns a kernel matrix. If the input is a vector
array, the kernels are computed. If the input is a kernel matrix, it is returned instead.
This method provides a safe way to take a kernel matrix as input, while preserving compatibility with many
other algorithms that take a vector array.
If Y is given (default is None), then the returned matrix is the pairwise kernel between the arrays from both X
and Y.
Valid values for metric are:: [‘rbf’, ‘sigmoid’, ‘polynomial’, ‘poly’, ‘linear’, ‘cosine’]
Parameters
n_features] otherwise] Array of pairwise kernels between samples, or a feature array.
Y [array [n_samples_b, n_features]] A second feature array only if X has shape [n_samples_a,
n_features].
metric [string, or callable] The metric to use when calculating kernel between instances
in a feature array. If metric is a string, it must be one of the metrics in pair-
wise.PAIRWISE_KERNEL_FUNCTIONS. If metric is “precomputed”, X is assumed to
be a kernel matrix. Alternatively, if metric is a callable function, it is called on each pair of
instances (rows) and the resulting value recorded. The callable should take two arrays from
X as input and return a value indicating the distance between them.
filter_params [boolean] Whether to filter invalid parameters or not.
them in parallel.
**kwds [optional keyword parameters] Any further parameters are passed directly to the kernel
function.
Returns
K [array [n_samples_a, n_samples_a] or [n_samples_a, n_samples_b]] A kernel matrix K such
that K_{i, j} is the kernel between the ith and jth vectors of the given matrix X, if Y is None.
If Y is not None, then K_{i, j} is the kernel between the ith array from X and the jth array
from Y.
Notes
If metric is ‘precomputed’, Y is ignored and X is returned.

sklearn.metrics.pairwise.polynomial_kernel
sklearn.metrics.pairwise.polynomial_kernel(X, Y=None, degree=3, gamma=None,

coef0=1)
Compute the polynomial kernel between X and Y:
K(X, Y) = (gamma <X, Y> + coef0)^degree

Parameters
X [ndarray of shape (n_samples_1, n_features)]
Y [ndarray of shape (n_samples_2, n_features)]
degree [int, default 3]
gamma [float, default None] if None, defaults to 1.0 / n_features
coef0 [float, default 1]
Returns
sklearn.metrics.pairwise.rbf_kernel
sklearn.metrics.pairwise.rbf_kernel(X, Y=None, gamma=None)

Compute the rbf (gaussian) kernel between X and Y:
K(x, y) = exp(-gamma ||x-y||^2)
for each pair of rows x in X and y in Y.

Parameters
X [array of shape (n_samples_X, n_features)]
Returns
sklearn.metrics.pairwise.sigmoid_kernel
sklearn.metrics.pairwise.sigmoid_kernel(X, Y=None, gamma=None, coef0=1)

Compute the sigmoid kernel between X and Y:
K(X, Y) = tanh(gamma <X, Y> + coef0)

Parameters
X [ndarray of shape (n_samples_1, n_features)]
Y [ndarray of shape (n_samples_2, n_features)]


coef0 [float, default 1]
Returns
sklearn.metrics.pairwise.paired_euclidean_distances
sklearn.metrics.pairwise.paired_euclidean_distances(X, Y)
Computes the paired euclidean distances between X and Y
Parameters
Y [array-like, shape (n_samples, n_features)]
Returns
distances [ndarray (n_samples, )]
sklearn.metrics.pairwise.paired_manhattan_distances
sklearn.metrics.pairwise.paired_manhattan_distances(X, Y)
Compute the L1 distances between the vectors in X and Y.
Parameters
Returns
sklearn.metrics.pairwise.paired_cosine_distances
sklearn.metrics.pairwise.paired_cosine_distances(X, Y)
Computes the paired cosine distances between X and Y
Parameters
Returns
distances [ndarray, shape (n_samples, )]

Notes
The cosine distance is equivalent to the half the squared euclidean distance if each sample is normalized to unit
norm
sklearn.metrics.pairwise.paired_distances
sklearn.metrics.pairwise.paired_distances(X, Y, metric=’euclidean’, **kwds)

Computes the paired distances between X and Y.
Computes the distances between (X[0], Y[0]), (X[1], Y[1]), etc. . .
Parameters
X [ndarray (n_samples, n_features)] Array 1 for distance computation.
Y [ndarray (n_samples, n_features)] Array 2 for distance computation.
metric [string or callable] The metric to use when calculating distance between instances
in a feature array. If metric is a string, it must be one of the options specified in
PAIRED_DISTANCES, including “euclidean”, “manhattan”, or “cosine”. Alternatively, if
metric is a callable function, it is called on each pair of instances (rows) and the resulting
value recorded. The callable should take two arrays from X as input and return a value
indicating the distance between them.
Returns
See also:
pairwise_distances Computes the distance between every pair of samples
Examples
>>> from sklearn.metrics.pairwise import paired_distances

>>> X = [[0, 1], [1, 1]]
>>> Y = [[0, 1], [2, 1]]
>>> paired_distances(X, Y)
array([0., 1.])
sklearn.metrics.pairwise_distances
sklearn.metrics.pairwise_distances(X, Y=None, metric=’euclidean’, n_jobs=None, **kwds)

Compute the distance matrix from a vector array X and optional Y.
This method takes either a vector array or a distance matrix, and returns a distance matrix. If the input is a vector
array, the distances are computed. If the input is a distances matrix, it is returned instead.
This method provides a safe way to take a distance matrix as input, while preserving compatibility with many
other algorithms that take a vector array.
If Y is given (default is None), then the returned matrix is the pairwise distance between the arrays from both X
and Y.

Valid values for metric are:

• From scikit-learn: [‘cityblock’, ‘cosine’, ‘euclidean’, ‘l1’, ‘l2’, ‘manhattan’]. These metrics support sparse
matrix inputs.
• From scipy.spatial.distance: [‘braycurtis’, ‘canberra’, ‘chebyshev’, ‘correlation’, ‘dice’, ‘hamming’, ‘jac-
card’, ‘kulsinski’, ‘mahalanobis’, ‘minkowski’, ‘rogerstanimoto’, ‘russellrao’, ‘seuclidean’, ‘sokalmich-
ener’, ‘sokalsneath’, ‘sqeuclidean’, ‘yule’] See the documentation for scipy.spatial.distance for details on
these metrics. These metrics do not support sparse matrix inputs.
Note that in the case of ‘cityblock’, ‘cosine’ and ‘euclidean’ (which are valid scipy.spatial.distance met-
rics), the scikit-learn implementation will be used, which is faster and has support for sparse matrices (ex-
cept for ‘cityblock’). For a verbose description of the metrics from scikit-learn, see the __doc__ of the
sklearn.pairwise.distance_metrics function.
Parameters
Y [array [n_samples_b, n_features], optional] An optional second feature array. Only allowed
if metric != “precomputed”.
metric [string, or callable] The metric to use when calculating distance between in-
stances in a feature array. If metric is a string, it must be one of the options al-
from X as input and return a value indicating the distance between them.
them in parallel.
**kwds [optional keyword parameters] Any further parameters are passed directly to the dis-
Returns
D [array [n_samples_a, n_samples_a] or [n_samples_a, n_samples_b]] A distance matrix D
such that D_{i, j} is the distance between the ith and jth vectors of the given matrix X, if Y
is None. If Y is not None, then D_{i, j} is the distance between the ith array from X and the
jth array from Y.
See also:
pairwise_distances_chunked performs the same calculation as this function, but returns a generator
of chunks of the distance matrix, in order to limit memory usage.
paired_distances Computes the distances between corresponding elements of two arrays

Examples using sklearn.metrics.pairwise_distances
sklearn.metrics.pairwise_distances_argmin
sklearn.metrics.pairwise_distances_argmin(X, Y, axis=1, metric=’euclidean’,

batch_size=None, metric_kwargs=None)
Compute minimum distances between one point and a set of points.
This function computes for each row in X, the index of the row of Y which is closest (according to the specified
distance).
This is mostly equivalent to calling:
pairwise_distances(X, Y=Y, metric=metric).argmin(axis=axis)
but uses much less memory, and is faster for large arrays.
This function works with dense 2D arrays only.
Parameters
X [array-like] Arrays containing points. Respective shapes (n_samples1, n_features) and
(n_samples2, n_features)
Y [array-like] Arrays containing points. Respective shapes (n_samples1, n_features) and
(n_samples2, n_features)
axis [int, optional, default 1] Axis along which the argmin and distances are to be computed.
metric [string or callable] metric to use for distance computation. Any metric from scikit-learn
or scipy.spatial.distance can be used.
If metric is a callable function, it is called on each pair of instances (rows) and the resulting
value recorded. The callable should take two arrays as input and return one value indicating
the distance between them. This works for Scipy’s metrics, but is less efficient than passing
the metric name as a string.
Distance matrices are not supported.
• from scikit-learn: [‘cityblock’, ‘cosine’, ‘euclidean’, ‘l1’, ‘l2’, ‘manhattan’]
• from scipy.spatial.distance: [‘braycurtis’, ‘canberra’, ‘chebyshev’, ‘correlation’, ‘dice’,
‘hamming’, ‘jaccard’, ‘kulsinski’, ‘mahalanobis’, ‘minkowski’, ‘rogerstanimoto’, ‘rus-
sellrao’, ‘seuclidean’, ‘sokalmichener’, ‘sokalsneath’, ‘sqeuclidean’, ‘yule’]
See the documentation for scipy.spatial.distance for details on these metrics.
batch_size [integer] Deprecated since version 0.20: Deprecated for removal in 0.22. Use
sklearn.set_config(working_memory=. . . ) instead.
metric_kwargs [dict] keyword arguments to pass to specified metric function.
Returns
argmin [numpy.ndarray] Y[argmin[i], :] is the row in Y that is closest to X[i, :].
See also:
sklearn.metrics.pairwise_distances, sklearn.metrics.pairwise_distances_argmin_min

Examples using sklearn.metrics.pairwise_distances_argmin

sklearn.metrics.pairwise_distances_argmin_min
sklearn.metrics.pairwise_distances_argmin_min(X, Y, axis=1, metric=’euclidean’,

batch_size=None, metric_kwargs=None)
Compute minimum distances between one point and a set of points.
This function computes for each row in X, the index of the row of Y which is closest (according to the specified
distance). The minimal distances are also returned.
This is mostly equivalent to calling:
(pairwise_distances(X, Y=Y, metric=metric).argmin(axis=axis), pairwise_distances(X, Y=Y,
metric=metric).min(axis=axis))
but uses much less memory, and is faster for large arrays.
Parameters
X [{array-like, sparse matrix}, shape (n_samples1, n_features)] Array containing points.
Y [{array-like, sparse matrix}, shape (n_samples2, n_features)] Arrays containing points.
axis [int, optional, default 1] Axis along which the argmin and distances are to be computed.
metric [string or callable, default ‘euclidean’] metric to use for distance computation. Any
metric from scikit-learn or scipy.spatial.distance can be used.
batch_size [integer] Deprecated since version 0.20: Deprecated for removal in 0.22. Use
sklearn.set_config(working_memory=. . . ) instead.
metric_kwargs [dict, optional] Keyword arguments to pass to specified metric function.
Returns
argmin [numpy.ndarray] Y[argmin[i], :] is the row in Y that is closest to X[i, :].
distances [numpy.ndarray] distances[i] is the distance between the i-th row in X and the
argmin[i]-th row in Y.
See also:
sklearn.metrics.pairwise_distances, sklearn.metrics.pairwise_distances_argmin

sklearn.metrics.pairwise_distances_chunked
sklearn.metrics.pairwise_distances_chunked(X, Y=None, reduce_func=None, met-

ric=’euclidean’, n_jobs=None, work-
ing_memory=None, **kwds)
Generate a distance matrix chunk by chunk with optional reduction
In cases where not all of a pairwise distance matrix needs to be stored at once, this is used to calculate pairwise
distances in working_memory-sized chunks. If reduce_func is given, it is run on each chunk and its
return values are concatenated into lists, arrays or sparse matrices.
Parameters
X [array [n_samples_a, n_samples_a] if metric == “precomputed”, or,] [n_samples_a,
n_features] otherwise Array of pairwise distances between samples, or a feature array.
Y [array [n_samples_b, n_features], optional] An optional second feature array. Only allowed
if metric != “precomputed”.
reduce_func [callable, optional] The function which is applied on each chunk of the distance
matrix, reducing it to needed values. reduce_func(D_chunk, start) is called re-
peatedly, where D_chunk is a contiguous vertical slice of the pairwise distance matrix,
starting at row start. It should return an array, a list, or a sparse matrix of length
D_chunk.shape[0], or a tuple of such objects.
If None, pairwise_distances_chunked returns a generator of vertical chunks of the distance
matrix.
metric [string, or callable] The metric to use when calculating distance between in-
stances in a feature array. If metric is a string, it must be one of the options al-
from X as input and return a value indicating the distance between them.
them in parallel.
working_memory [int, optional] The sought maximum memory for temporary
distance matrix chunks. When None (default), the value of sklearn.
get_config()['working_memory'] is used.
‘**kwds‘ [optional keyword parameters] Any further parameters are passed directly to the dis-
Yields
D_chunk [array or sparse matrix] A contiguous slice of distance matrix, optionally processed
by reduce_func.
Examples
Without reduce_func:

>>> X = np.random.RandomState(0).rand(5, 3)
>>> D_chunk = next(pairwise_distances_chunked(X))
>>> D_chunk
array([[0. ..., 0.29..., 0.41..., 0.19..., 0.57...],
[0.29..., 0. ..., 0.57..., 0.41..., 0.76...],
[0.41..., 0.57..., 0. ..., 0.44..., 0.90...],
[0.19..., 0.41..., 0.44..., 0. ..., 0.51...],
[0.57..., 0.76..., 0.90..., 0.51..., 0. ...]])
Retrieve all neighbors and average distance within radius r:
>>> r = .2
>>> def reduce_func(D_chunk, start):
... neigh = [np.flatnonzero(d < r) for d in D_chunk]
... avg_dist = (D_chunk * (D_chunk < r)).mean(axis=1)
... return neigh, avg_dist
>>> gen = pairwise_distances_chunked(X, reduce_func=reduce_func)
>>> neigh, avg_dist = next(gen)
>>> neigh
[array([0, 3]), array([1]), array([2]), array([0, 3]), array([4])]
>>> avg_dist
array([0.039..., 0. , 0. , 0.039..., 0. ])
Where r is defined per sample, we need to make use of start:
>>> r = [.2, .4, .4, .3, .1]

>>> def reduce_func(D_chunk, start):
... neigh = [np.flatnonzero(d < r[i])
... for i, d in enumerate(D_chunk, start)]
... return neigh
>>> neigh = next(pairwise_distances_chunked(X, reduce_func=reduce_func))
>>> neigh
[array([0, 3]), array([0, 1]), array([2]), array([0, 3]), array([4])]
Force row-by-row generation by reducing working_memory:
>>> gen = pairwise_distances_chunked(X, reduce_func=reduce_func,

... working_memory=0)
>>> next(gen)
[array([0, 3])]
>>> next(gen)
[array([0, 1])]
6.24 sklearn.mixture: Gaussian Mixture Models
The sklearn.mixture module implements mixture modeling algorithms.

User guide: See the Gaussian mixture models section for further details.
mixture.BayesianGaussianMixture([. . . ]) Variational Bayesian estimation of a Gaussian mixture.

mixture.GaussianMixture([n_components, . . . ]) Gaussian Mixture.
6.24. sklearn.mixture: Gaussian Mixture Models 1951

6.24.1 sklearn.mixture.BayesianGaussianMixture
class sklearn.mixture.BayesianGaussianMixture(n_components=1, covariance_type=’full’,

tol=0.001, reg_covar=1e-06, max_iter=100,
n_init=1, init_params=’kmeans’,
weight_concentration_prior_type=’dirichlet_process’,
weight_concentration_prior=None,
mean_precision_prior=None,
mean_prior=None, de-
grees_of_freedom_prior=None, covari-
ance_prior=None, random_state=None,
warm_start=False, verbose=0, ver-
bose_interval=10)
Variational Bayesian estimation of a Gaussian mixture.
This class allows to infer an approximate posterior distribution over the parameters of a Gaussian mixture
distribution. The effective number of components can be inferred from the data.
This class implements two types of prior for the weights distribution: a finite mixture model with Dirichlet
distribution and an infinite mixture model with the Dirichlet Process. In practice Dirichlet Process inference
algorithm is approximated and uses a truncated distribution with a fixed maximum number of components
(called the Stick-breaking representation). The number of components actually used almost always depends on
the data.
Parameters
n_components [int, defaults to 1.] The number of mixture components. Depending on the data
and the value of the weight_concentration_prior the model can decide to not use all the
components by setting some component weights_ to values very close to zero. The number
of effective components is therefore smaller than n_components.
covariance_type [{‘full’, ‘tied’, ‘diag’, ‘spherical’}, defaults to ‘full’] String describing the
type of covariance parameters to use. Must be one of:
'full' (each component has its own general covariance matrix),

'tied' (all components share the same general covariance matrix),
'diag' (each component has its own diagonal covariance matrix),
'spherical' (each component has its own single variance).
tol [float, defaults to 1e-3.] The convergence threshold. EM iterations will stop when the lower
bound average gain on the likelihood (of the training data with respect to the model) is below
this threshold.
reg_covar [float, defaults to 1e-6.] Non-negative regularization added to the diagonal of co-
variance. Allows to assure that the covariance matrices are all positive.
max_iter [int, defaults to 100.] The number of EM iterations to perform.
n_init [int, defaults to 1.] The number of initializations to perform. The result with the highest
lower bound value on the likelihood is kept.
init_params [{‘kmeans’, ‘random’}, defaults to ‘kmeans’.] The method used to initialize the
weights, the means and the covariances. Must be one of:
'kmeans' : responsibilities are initialized using kmeans.

'random' : responsibilities are initialized randomly.

weight_concentration_prior_type [str, defaults to ‘dirichlet_process’.] String describing the

type of the weight concentration prior. Must be one of:
'dirichlet_process' (using the Stick-breaking representation),
'dirichlet_distribution' (can favor more uniform weights).
weight_concentration_prior [float | None, optional.] The dirichlet concentration of each com-

ponent on the weight distribution (Dirichlet). This is commonly called gamma in the litera-
ture. The higher concentration puts more mass in the center and will lead to more compo-
nents being active, while a lower concentration parameter will lead to more mass at the edge
of the mixture weights simplex. The value of the parameter must be greater than 0. If it is
None, it’s set to 1. / n_components.
mean_precision_prior [float | None, optional.] The precision prior on the mean distribution
(Gaussian). Controls the extend to where means can be placed. Smaller values concentrate
the means of each clusters around mean_prior. The value of the parameter must be greater
than 0. If it is None, it’s set to 1.
mean_prior [array-like, shape (n_features,), optional] The prior on the mean distribution
(Gaussian). If it is None, it’s set to the mean of X.
degrees_of_freedom_prior [float | None, optional.] The prior of the number of degrees of
freedom on the covariance distributions (Wishart). If it is None, it’s set to n_features.
covariance_prior [float or array-like, optional] The prior on the covariance distribution
(Wishart). If it is None, the emiprical covariance prior is initialized using the covariance
of X. The shape depends on covariance_type:
(n_features, n_features) if 'full',
(n_features, n_features) if 'tied',
(n_features) if 'diag',
float if 'spherical'

warm_start [bool, default to False.] If ‘warm_start’ is True, the solution of the last fitting is
used as initialization for the next call of fit(). This can speed up convergence when fit is
called several times on similar problems. See the Glossary.
verbose [int, default to 0.] Enable verbose output. If 1 then it prints the current initialization
and each iteration step. If greater than 1 then it prints also the log probability and the time
needed for each step.
verbose_interval [int, default to 10.] Number of iteration done before the next print.
Attributes
weights_ [array-like, shape (n_components,)] The weights of each mixture components.
means_ [array-like, shape (n_components, n_features)] The mean of each mixture component.
covariances_ [array-like] The covariance of each mixture component. The shape depends on
covariance_type:
(n_components,) if 'spherical',
(n_components, n_features) if 'diag',
(n_components, n_features, n_features) if 'full'

precisions_ [array-like] The precision matrices for each component in the mixture. A preci-
sion matrix is the inverse of a covariance matrix. A covariance matrix is symmetric posi-
tive definite so the mixture of Gaussian can be equivalently parameterized by the precision
matrices. Storing the precision matrices instead of the covariance matrices makes it more
efficient to compute the log-likelihood of new samples at test time. The shape depends on
covariance_type:
precisions_cholesky_ [array-like] The cholesky decomposition of the precision matrices of

each mixture component. A precision matrix is the inverse of a covariance matrix. A covari-
ance matrix is symmetric positive definite so the mixture of Gaussian can be equivalently
parameterized by the precision matrices. Storing the precision matrices instead of the co-
variance matrices makes it more efficient to compute the log-likelihood of new samples at
test time. The shape depends on covariance_type:
converged_ [bool] True when convergence was reached in fit(), False otherwise.
n_iter_ [int] Number of step used by the best fit of inference to reach the convergence.
lower_bound_ [float] Lower bound value on the likelihood (of the training data with respect to
the model) of the best fit of inference.
weight_concentration_prior_ [tuple or float] The dirichlet concentration of each
component on the weight distribution (Dirichlet). The type depends on
weight_concentration_prior_type:
(float, float) if 'dirichlet_process' (Beta parameters),

float if 'dirichlet_distribution' (Dirichlet parameters).
The higher concentration puts more mass in the center and will lead to more components
being active, while a lower concentration parameter will lead to more mass at the edge of
the simplex.
weight_concentration_ [array-like, shape (n_components,)] The dirichlet concentration of
each component on the weight distribution (Dirichlet).
mean_precision_prior [float] The precision prior on the mean distribution (Gaussian). Con-
trols the extend to where means can be placed. Smaller values concentrate the means of
each clusters around mean_prior.
mean_precision_ [array-like, shape (n_components,)] The precision of each components on
the mean distribution (Gaussian).
mean_prior_ [array-like, shape (n_features,)] The prior on the mean distribution (Gaussian).
degrees_of_freedom_prior_ [float] The prior of the number of degrees of freedom on the co-
variance distributions (Wishart).

degrees_of_freedom_ [array-like, shape (n_components,)] The number of degrees of freedom

of each components in the model.
covariance_prior_ [float or array-like] The prior on the covariance distribution (Wishart). The
shape depends on covariance_type:
(n_features, n_features) if 'full',

(n_features) if 'diag',
float if 'spherical'
See also:
GaussianMixture Finite Gaussian mixture fit with EM.
References
[1], [2], [3]
Methods
fit(X[, y]) Estimate model parameters with the EM algorithm.

fit_predict(X[, y]) Estimate model parameters using X and predict the la-
bels for X.
predict(X) Predict the labels for the data samples in X using trained
model.
predict_proba(X) Predict posterior probability of each component given
the data.
sample([n_samples]) Generate random samples from the fitted Gaussian dis-
tribution.
score(X[, y]) Compute the per-sample average log-likelihood of the
given data X.
score_samples(X) Compute the weighted log probabilities for each sam-
ple.
__init__(n_components=1, covariance_type=’full’, tol=0.001, reg_covar=1e-06, max_iter=100,

n_init=1, init_params=’kmeans’, weight_concentration_prior_type=’dirichlet_process’,
weight_concentration_prior=None, mean_precision_prior=None, mean_prior=None,
degrees_of_freedom_prior=None, covariance_prior=None, random_state=None,
warm_start=False, verbose=0, verbose_interval=10)
fit(X, y=None)
Estimate model parameters with the EM algorithm.
The method fits the model n_init times and sets the parameters with which the model has the
largest likelihood or lower bound. Within each trial, the method iterates between E-step and M-step
for max_iter times until the change of likelihood or lower bound is less than tol, otherwise, a
ConvergenceWarning is raised. If warm_start is True, then n_init is ignored and a single
initialization is performed upon the first call. Upon consecutive calls, training starts where it left off.
Parameters

X [array-like, shape (n_samples, n_features)] List of n_features-dimensional data points.

Each row corresponds to a single data point.
Returns
self
Estimate model parameters using X and predict the labels for X.
The method fits the model n_init times and sets the parameters with which the model has the largest
likelihood or lower bound. Within each trial, the method iterates between E-step and M-step for max_iter
times until the change of likelihood or lower bound is less than tol, otherwise, a ConvergenceWarning is
raised. After fitting, it predicts the most probable label for the input data points.
Parameters
Returns
labels [array, shape (n_samples,)] Component labels.
Parameters
Returns
predict(X)
Predict the labels for the data samples in X using trained model.
Parameters
Returns
predict_proba(X)
Predict posterior probability of each component given the data.
Parameters
Returns
resp [array, shape (n_samples, n_components)] Returns the probability each Gaussian
(state) in the model given each sample.
sample(n_samples=1)
Generate random samples from the fitted Gaussian distribution.
Parameters

n_samples [int, optional] Number of samples to generate. Defaults to 1.

Returns
X [array, shape (n_samples, n_features)] Randomly generated sample
y [array, shape (nsamples,)] Component labels
score(X, y=None)
Compute the per-sample average log-likelihood of the given data X.
Parameters
X [array-like, shape (n_samples, n_dimensions)] List of n_features-dimensional data points.
Returns
log_likelihood [float] Log likelihood of the Gaussian mixture given X.
score_samples(X)
Compute the weighted log probabilities for each sample.
Parameters
Returns
log_prob [array, shape (n_samples,)] Log probabilities of each data point in X.
of a nested object.
Returns
self
Examples using sklearn.mixture.BayesianGaussianMixture
• Gaussian Mixture Model Ellipsoids

• Gaussian Mixture Model Sine Curve
• Concentration Prior Type Analysis of Variation Bayesian Gaussian Mixture
6.24.2 sklearn.mixture.GaussianMixture
class sklearn.mixture.GaussianMixture(n_components=1, covariance_type=’full’, tol=0.001,

reg_covar=1e-06, max_iter=100, n_init=1,
init_params=’kmeans’, weights_init=None,
means_init=None, precisions_init=None, ran-
dom_state=None, warm_start=False, verbose=0,
verbose_interval=10)
Gaussian Mixture.
Representation of a Gaussian mixture model probability distribution. This class allows to estimate the parame-
ters of a Gaussian mixture distribution.


Parameters
n_components [int, defaults to 1.] The number of mixture components.
covariance_type [{‘full’ (default), ‘tied’, ‘diag’, ‘spherical’}] String describing the type of
covariance parameters to use. Must be one of:
‘full’ each component has its own general covariance matrix
‘tied’ all components share the same general covariance matrix
‘diag’ each component has its own diagonal covariance matrix
‘spherical’ each component has its own single variance
tol [float, defaults to 1e-3.] The convergence threshold. EM iterations will stop when the lower
bound average gain is below this threshold.
reg_covar [float, defaults to 1e-6.] Non-negative regularization added to the diagonal of co-
variance. Allows to assure that the covariance matrices are all positive.
max_iter [int, defaults to 100.] The number of EM iterations to perform.
n_init [int, defaults to 1.] The number of initializations to perform. The best results are kept.
init_params [{‘kmeans’, ‘random’}, defaults to ‘kmeans’.] The method used to initialize the
weights, the means and the precisions. Must be one of:
'kmeans' : responsibilities are initialized using kmeans.

'random' : responsibilities are initialized randomly.
weights_init [array-like, shape (n_components, ), optional] The user-provided initial weights,

defaults to None. If it None, weights are initialized using the init_params method.
means_init [array-like, shape (n_components, n_features), optional] The user-provided initial
means, defaults to None, If it None, means are initialized using the init_params method.
precisions_init [array-like, optional.] The user-provided initial precisions (inverse of the co-
variance matrices), defaults to None. If it None, precisions are initialized using the
‘init_params’ method. The shape depends on ‘covariance_type’:

warm_start [bool, default to False.] If ‘warm_start’ is True, the solution of the last fitting is
used as initialization for the next call of fit(). This can speed up convergence when fit is
called several times on similar problems. In that case, ‘n_init’ is ignored and only a single
initialization occurs upon the first call. See the Glossary.
verbose [int, default to 0.] Enable verbose output. If 1 then it prints the current initialization
and each iteration step. If greater than 1 then it prints also the log probability and the time
needed for each step.

verbose_interval [int, default to 10.] Number of iteration done before the next print.
Attributes
weights_ [array-like, shape (n_components,)] The weights of each mixture components.
means_ [array-like, shape (n_components, n_features)] The mean of each mixture component.
covariances_ [array-like] The covariance of each mixture component. The shape depends on
covariance_type:
precisions_ [array-like] The precision matrices for each component in the mixture. A precision
matrix is the inverse of a covariance matrix. A covariance matrix is symmetric positive defi-
nite so the mixture of Gaussian can be equivalently parameterized by the precision matrices.
Storing the precision matrices instead of the covariance matrices makes it more efficient
to compute the log-likelihood of new samples at test time. The shape depends on covari-
ance_type:
precisions_cholesky_ [array-like] The cholesky decomposition of the precision matrices of

each mixture component. A precision matrix is the inverse of a covariance matrix. A covari-
ance matrix is symmetric positive definite so the mixture of Gaussian can be equivalently
parameterized by the precision matrices. Storing the precision matrices instead of the co-
variance matrices makes it more efficient to compute the log-likelihood of new samples at
test time. The shape depends on covariance_type:
converged_ [bool] True when convergence was reached in fit(), False otherwise.
n_iter_ [int] Number of step used by the best fit of EM to reach the convergence.
lower_bound_ [float] Lower bound value on the log-likelihood (of the training data with re-
spect to the model) of the best fit of EM.
See also:
BayesianGaussianMixture Gaussian mixture model fit with a variational inference.
Methods
aic(X) Akaike information criterion for the current model on

the input X.


bic(X) Bayesian information criterion for the current model on
the input X.
fit(X[, y]) Estimate model parameters with the EM algorithm.
fit_predict(X[, y]) Estimate model parameters using X and predict the la-
bels for X.
predict(X) Predict the labels for the data samples in X using trained
model.
predict_proba(X) Predict posterior probability of each component given
the data.
sample([n_samples]) Generate random samples from the fitted Gaussian dis-
tribution.
score(X[, y]) Compute the per-sample average log-likelihood of the
given data X.
score_samples(X) Compute the weighted log probabilities for each sam-
ple.
__init__(n_components=1, covariance_type=’full’, tol=0.001, reg_covar=1e-06, max_iter=100,

n_init=1, init_params=’kmeans’, weights_init=None, means_init=None, pre-
cisions_init=None, random_state=None, warm_start=False, verbose=0, ver-
bose_interval=10)
aic(X)
Akaike information criterion for the current model on the input X.
Parameters
X [array of shape (n_samples, n_dimensions)]
Returns
aic [float] The lower the better.
bic(X)
Bayesian information criterion for the current model on the input X.
Parameters
X [array of shape (n_samples, n_dimensions)]
Returns
bic [float] The lower the better.
fit(X, y=None)
Estimate model parameters with the EM algorithm.
The method fits the model n_init times and sets the parameters with which the model has the
largest likelihood or lower bound. Within each trial, the method iterates between E-step and M-step
for max_iter times until the change of likelihood or lower bound is less than tol, otherwise, a
ConvergenceWarning is raised. If warm_start is True, then n_init is ignored and a single
initialization is performed upon the first call. Upon consecutive calls, training starts where it left off.
Parameters
Returns

self
Estimate model parameters using X and predict the labels for X.
The method fits the model n_init times and sets the parameters with which the model has the largest
likelihood or lower bound. Within each trial, the method iterates between E-step and M-step for max_iter
times until the change of likelihood or lower bound is less than tol, otherwise, a ConvergenceWarning is
raised. After fitting, it predicts the most probable label for the input data points.
Parameters
Returns
Parameters
Returns
predict(X)
Predict the labels for the data samples in X using trained model.
Parameters
Returns
predict_proba(X)
Predict posterior probability of each component given the data.
Parameters
Returns
resp [array, shape (n_samples, n_components)] Returns the probability each Gaussian
(state) in the model given each sample.
sample(n_samples=1)
Generate random samples from the fitted Gaussian distribution.
Parameters
Returns

X [array, shape (n_samples, n_features)] Randomly generated sample

y [array, shape (nsamples,)] Component labels
score(X, y=None)
Compute the per-sample average log-likelihood of the given data X.
Parameters
X [array-like, shape (n_samples, n_dimensions)] List of n_features-dimensional data points.
Returns
log_likelihood [float] Log likelihood of the Gaussian mixture given X.
score_samples(X)
Compute the weighted log probabilities for each sample.
Parameters
Returns
log_prob [array, shape (n_samples,)] Log probabilities of each data point in X.
of a nested object.
Returns
self
Examples using sklearn.mixture.GaussianMixture

• Density Estimation for a Gaussian mixture
• Gaussian Mixture Model Ellipsoids
• Gaussian Mixture Model Selection
• GMM covariances
• Gaussian Mixture Model Sine Curve
6.25 sklearn.model_selection: Model Selection
User guide: See the Cross-validation: evaluating estimator performance, Tuning the hyper-parameters of an estima-
tor and Learning curve sections for further details.
6.25.1 Splitter Classes

model_selection.GroupKFold([n_splits]) K-fold iterator variant with non-overlapping groups.

model_selection.GroupShuffleSplit([. . . ]) Shuffle-Group(s)-Out cross-validation iterator
model_selection.KFold([n_splits, shuffle, . . . ]) K-Folds cross-validator
model_selection.LeaveOneGroupOut() Leave One Group Out cross-validator
model_selection.LeavePGroupsOut(n_groups) Leave P Group(s) Out cross-validator
model_selection.LeaveOneOut() Leave-One-Out cross-validator
model_selection.LeavePOut(p) Leave-P-Out cross-validator
model_selection.PredefinedSplit(test_fold) Predefined split cross-validator
model_selection.RepeatedKFold([n_splits, Repeated K-Fold cross validator.
. . . ])
model_selection.RepeatedStratifiedKFold([. .Repeated
. ]) Stratified K-Fold cross validator.
model_selection.ShuffleSplit([n_splits, . . . ]) Random permutation cross-validator
model_selection.StratifiedKFold([n_splits, Stratified K-Folds cross-validator
. . . ])
model_selection.StratifiedShuffleSplit([. . . ]) Stratified ShuffleSplit cross-validator
model_selection.TimeSeriesSplit([n_splits, Time Series cross-validator
. . . ])
sklearn.model_selection.GroupKFold
class sklearn.model_selection.GroupKFold(n_splits=’warn’)
K-fold iterator variant with non-overlapping groups.
The same group will not appear in two different folds (the number of distinct groups has to be at least equal to
the number of folds).
The folds are approximately balanced in the sense that the number of distinct groups is approximately the same
in each fold.
Parameters
n_splits [int, default=3] Number of folds. Must be at least 2.
Changed in version 0.20: n_splits default value will change from 3 to 5 in v0.22.
See also:
LeaveOneGroupOut For splitting the data according to explicit domain-specific stratification of the dataset.
Examples
>>> from sklearn.model_selection import GroupKFold

>>> X = np.array([[1, 2], [3, 4], [5, 6], [7, 8]])
>>> y = np.array([1, 2, 3, 4])
>>> groups = np.array([0, 0, 2, 2])
>>> group_kfold = GroupKFold(n_splits=2)
>>> group_kfold.get_n_splits(X, y, groups)
2
>>> print(group_kfold)
GroupKFold(n_splits=2)
>>> for train_index, test_index in group_kfold.split(X, y, groups):
... print("TRAIN:", train_index, "TEST:", test_index)
... X_train, X_test = X[train_index], X[test_index]
... y_train, y_test = y[train_index], y[test_index]
... print(X_train, X_test, y_train, y_test)
6.25. sklearn.model_selection: Model Selection 1963

...
TRAIN: [0 1] TEST: [2 3]
[[1 2]
[3 4]] [[5 6]
[7 8]] [1 2] [3 4]
TRAIN: [2 3] TEST: [0 1]
[[5 6]
[7 8]] [[1 2]
[3 4]] [3 4] [1 2]
Methods
get_n_splits([X, y, groups]) Returns the number of splitting iterations in the cross-

validator
split(X[, y, groups]) Generate indices to split data into training and test set.
__init__(n_splits=’warn’)
get_n_splits(X=None, y=None, groups=None)
Returns the number of splitting iterations in the cross-validator
Parameters
X [object] Always ignored, exists for compatibility.
y [object] Always ignored, exists for compatibility.
groups [object] Always ignored, exists for compatibility.
Returns
n_splits [int] Returns the number of splitting iterations in the cross-validator.
split(X, y=None, groups=None)
Generate indices to split data into training and test set.
Parameters
y [array-like, shape (n_samples,), optional] The target variable for supervised learning prob-
lems.
groups [array-like, with shape (n_samples,)] Group labels for the samples used while split-
ting the dataset into train/test set.
Yields
train [ndarray] The training set indices for that split.
test [ndarray] The testing set indices for that split.
Examples using sklearn.model_selection.GroupKFold
• Visualizing cross-validation behavior in scikit-learn

sklearn.model_selection.GroupShuffleSplit
class sklearn.model_selection.GroupShuffleSplit(n_splits=5, test_size=’default’,

train_size=None, random_state=None)
Shuffle-Group(s)-Out cross-validation iterator
Provides randomized train/test indices to split data according to a third-party provided group. This group infor-
mation can be used to encode arbitrary domain specific stratifications of the samples as integers.
For instance the groups could be the year of collection of the samples and thus allow for cross-validation against
time-based splits.
The difference between LeavePGroupsOut and GroupShuffleSplit is that the former generates splits using all
subsets of size p unique groups, whereas GroupShuffleSplit generates a user-determined number of random test
splits, each with a user-determined fraction of unique groups.
For example, a less computationally intensive alternative to LeavePGroupsOut(p=10) would be
GroupShuffleSplit(test_size=10, n_splits=100).
Note: The parameters test_size and train_size refer to groups, and not to samples, as in ShuffleSplit.
Parameters
n_splits [int (default 5)] Number of re-shuffling & splitting iterations.
test_size [float, int, None, optional] If float, should be between 0.0 and 1.0 and represent the
proportion of the dataset to include in the test split. If int, represents the absolute number
of test samples. If None, the value is set to the complement of the train size. By default,
the value is set to 0.2. The default will change in version 0.21. It will remain 0.2 only if
train_size is unspecified, otherwise it will complement the specified train_size.
train_size [float, int, or None, default is None] If float, should be between 0.0 and 1.0 and
represent the proportion of the groups to include in the train split. If int, represents the
absolute number of train groups. If None, the value is automatically set to the complement
of the test size.
Methods

validator
__init__(n_splits=5, test_size=’default’, train_size=None, random_state=None)

Parameters

Returns
Parameters
y [array-like, shape (n_samples,), optional] The target variable for supervised learning prob-
lems.
Yields
Notes
Randomized CV splitters may return different results for each call of split. You can make the results
identical by setting random_state to an integer.
Examples using sklearn.model_selection.GroupShuffleSplit
sklearn.model_selection.KFold
class sklearn.model_selection.KFold(n_splits=’warn’, shuffle=False, random_state=None)

K-Folds cross-validator
Provides train/test indices to split data in train/test sets. Split dataset into k consecutive folds (without shuffling
by default).
Each fold is then used once as a validation while the k - 1 remaining folds form the training set.
Parameters
shuffle [boolean, optional] Whether to shuffle the data before splitting into batches.
the RandomState instance used by np.random. Used when shuffle == True.
See also:

StratifiedKFold Takes group information into account to avoid building folds with imbalanced class
distributions (for binary or multiclass classification tasks).
GroupKFold K-fold iterator variant with non-overlapping groups.
RepeatedKFold Repeats K-Fold n times.
Notes
The first n_samples % n_splits folds have size n_samples // n_splits + 1, other folds have
size n_samples // n_splits, where n_samples is the number of samples.
Randomized CV splitters may return different results for each call of split. You can make the results identical
by setting random_state to an integer.
Examples
>>> from sklearn.model_selection import KFold

>>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4]])
>>> y = np.array([1, 2, 3, 4])
>>> kf = KFold(n_splits=2)
>>> kf.get_n_splits(X)
2
>>> print(kf)
KFold(n_splits=2, random_state=None, shuffle=False)
>>> for train_index, test_index in kf.split(X):
TRAIN: [2 3] TEST: [0 1]
TRAIN: [0 1] TEST: [2 3]
Methods

validator
__init__(n_splits=’warn’, shuffle=False, random_state=None)

Parameters
Returns


Parameters
y [array-like, shape (n_samples,)] The target variable for supervised learning problems.
groups [array-like, with shape (n_samples,), optional] Group labels for the samples used
while splitting the dataset into train/test set.
Yields
Examples using sklearn.model_selection.KFold

sklearn.model_selection.LeaveOneGroupOut
class sklearn.model_selection.LeaveOneGroupOut
Leave One Group Out cross-validator
Provides train/test indices to split data according to a third-party provided group. This group information can be
used to encode arbitrary domain specific stratifications of the samples as integers.
time-based splits.
Examples
>>> from sklearn.model_selection import LeaveOneGroupOut

>>> X = np.array([[1, 2], [3, 4], [5, 6], [7, 8]])
>>> y = np.array([1, 2, 1, 2])
>>> groups = np.array([1, 1, 2, 2])
>>> logo = LeaveOneGroupOut()
>>> logo.get_n_splits(X, y, groups)
2
>>> logo.get_n_splits(groups=groups) # 'groups' is always required
2
>>> print(logo)
LeaveOneGroupOut()
>>> for train_index, test_index in logo.split(X, y, groups):


TRAIN: [2 3] TEST: [0 1]
[[5 6]
[7 8]] [[1 2]
[3 4]] [1 2] [1 2]
TRAIN: [0 1] TEST: [2 3]
[[1 2]
[3 4]] [[5 6]
[7 8]] [1 2] [1 2]
Methods

validator
__init__()
Parameters
ting the dataset into train/test set. This ‘groups’ parameter must always be specified to
calculate the number of splits, though the other parameters can be omitted.
Returns
Parameters
y [array-like, of length n_samples, optional] The target variable for supervised learning
problems.
Yields

sklearn.model_selection.LeavePGroupsOut
class sklearn.model_selection.LeavePGroupsOut(n_groups)
Leave P Group(s) Out cross-validator
Provides train/test indices to split data according to a third-party provided group. This group information can be
used to encode arbitrary domain specific stratifications of the samples as integers.
time-based splits.
The difference between LeavePGroupsOut and LeaveOneGroupOut is that the former builds the test sets with
all the samples assigned to p different values of the groups while the latter uses samples all assigned the same
groups.
Parameters
n_groups [int] Number of groups (p) to leave out in the test split.
See also:
Examples
>>> from sklearn.model_selection import LeavePGroupsOut

>>> X = np.array([[1, 2], [3, 4], [5, 6]])
>>> y = np.array([1, 2, 1])
>>> groups = np.array([1, 2, 3])
>>> lpgo = LeavePGroupsOut(n_groups=2)
>>> lpgo.get_n_splits(X, y, groups)
3
>>> lpgo.get_n_splits(groups=groups) # 'groups' is always required
3
>>> print(lpgo)
LeavePGroupsOut(n_groups=2)
>>> for train_index, test_index in lpgo.split(X, y, groups):
TRAIN: [2] TEST: [0 1]
[[5 6]] [[1 2]
[3 4]] [1] [1 2]
[[3 4]] [[1 2]
[5 6]] [2] [1 1]
[[1 2]] [[3 4]
[5 6]] [1] [2 1]
Methods


validator
__init__(n_groups)
Parameters
ting the dataset into train/test set. This ‘groups’ parameter must always be specified to
calculate the number of splits, though the other parameters can be omitted.
Returns
Parameters
y [array-like, of length n_samples, optional] The target variable for supervised learning
problems.
Yields
sklearn.model_selection.LeaveOneOut
class sklearn.model_selection.LeaveOneOut
Leave-One-Out cross-validator
Provides train/test indices to split data in train/test sets. Each sample is used once as a test set (singleton) while
the remaining samples form the training set.
Note: LeaveOneOut() is equivalent to KFold(n_splits=n) and LeavePOut(p=1) where n is the
number of samples.
Due to the high number of test sets (which is the same as the number of samples) this cross-validation method
can be very costly. For large datasets one should favor KFold, ShuffleSplit or StratifiedKFold.
See also:
LeaveOneGroupOut For splitting the data according to explicit, domain-specific stratification of the dataset.

Examples
>>> from sklearn.model_selection import LeaveOneOut

>>> X = np.array([[1, 2], [3, 4]])
>>> y = np.array([1, 2])
>>> loo = LeaveOneOut()
>>> loo.get_n_splits(X)
2
>>> print(loo)
LeaveOneOut()
>>> for train_index, test_index in loo.split(X):
TRAIN: [1] TEST: [0]
[[3 4]] [[1 2]] [2] [1]
[[1 2]] [[3 4]] [1] [2]
Methods
get_n_splits(X[, y, groups]) Returns the number of splitting iterations in the cross-

validator
__init__()
get_n_splits(X, y=None, groups=None)
Parameters
Returns
Parameters
y [array-like, of length n_samples] The target variable for supervised learning problems.

Yields
sklearn.model_selection.LeavePOut
class sklearn.model_selection.LeavePOut(p)
Leave-P-Out cross-validator
Provides train/test indices to split data in train/test sets. This results in testing on all distinct samples of size p,
while the remaining n - p samples form the training set in each iteration.
Note: LeavePOut(p) is NOT equivalent to KFold(n_splits=n_samples // p) which creates non-
overlapping test sets.
Due to the high number of iterations which grows combinatorically with the number of samples this cross-
validation method can be very costly. For large datasets one should favor KFold, StratifiedKFold or
ShuffleSplit.
Parameters
p [int] Size of the test sets.
Examples
>>> from sklearn.model_selection import LeavePOut

>>> X = np.array([[1, 2], [3, 4], [5, 6], [7, 8]])
>>> y = np.array([1, 2, 3, 4])
>>> lpo = LeavePOut(2)
>>> lpo.get_n_splits(X)
6
>>> print(lpo)
LeavePOut(p=2)
>>> for train_index, test_index in lpo.split(X):
TRAIN: [2 3] TEST: [0 1]
TRAIN: [1 3] TEST: [0 2]
TRAIN: [1 2] TEST: [0 3]
TRAIN: [0 3] TEST: [1 2]
TRAIN: [0 2] TEST: [1 3]
TRAIN: [0 1] TEST: [2 3]
Methods

get_n_splits(X[, y, groups]) Returns the number of splitting iterations in the cross-

validator
__init__(p)
get_n_splits(X, y=None, groups=None)
Parameters
Parameters
Yields
sklearn.model_selection.PredefinedSplit
class sklearn.model_selection.PredefinedSplit(test_fold)
Predefined split cross-validator
Provides train/test indices to split data into train/test sets using a predefined scheme specified by the user with
the test_fold parameter.
Parameters
test_fold [array-like, shape (n_samples,)] The entry test_fold[i] represents the index of
the test set that sample i belongs to. It is possible to exclude sample i from any test set (i.e.
include sample i in every training set) by setting test_fold[i] equal to -1.
Examples
>>> from sklearn.model_selection import PredefinedSplit

>>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4]])
>>> y = np.array([0, 0, 1, 1])
>>> test_fold = [0, 1, -1, 1]
>>> ps = PredefinedSplit(test_fold)

>>> ps.get_n_splits()
2
>>> print(ps)
PredefinedSplit(test_fold=array([ 0, 1, -1, 1]))
>>> for train_index, test_index in ps.split():
TRAIN: [1 2 3] TEST: [0]
TRAIN: [0 2] TEST: [1 3]
Methods

validator
split([X, y, groups]) Generate indices to split data into training and test set.
__init__(test_fold)
Parameters
Returns
split(X=None, y=None, groups=None)
Parameters
Yields
sklearn.model_selection.RepeatedKFold
class sklearn.model_selection.RepeatedKFold(n_splits=5, n_repeats=10, ran-

dom_state=None)
Repeated K-Fold cross validator.
Repeats K-Fold n times with different randomization in each repetition.

Parameters
n_repeats [int, default=10] Number of times cross-validator needs to be repeated.
See also:
RepeatedStratifiedKFold Repeats Stratified K-Fold n times.
Notes
Examples
>>> from sklearn.model_selection import RepeatedKFold

>>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4]])
>>> y = np.array([0, 0, 1, 1])
>>> rkf = RepeatedKFold(n_splits=2, n_repeats=2, random_state=2652124)
>>> for train_index, test_index in rkf.split(X):
...
TRAIN: [0 1] TEST: [2 3]
TRAIN: [2 3] TEST: [0 1]
TRAIN: [1 2] TEST: [0 3]
TRAIN: [0 3] TEST: [1 2]
Methods

validator
split(X[, y, groups]) Generates indices to split data into training and test set.
__init__(n_splits=5, n_repeats=10, random_state=None)

Parameters
X [object] Always ignored, exists for compatibility. np.zeros(n_samples) may be
used as a placeholder.
y [object] Always ignored, exists for compatibility. np.zeros(n_samples) may be

Returns
Generates indices to split data into training and test set.
Parameters
Yields
sklearn.model_selection.RepeatedStratifiedKFold
class sklearn.model_selection.RepeatedStratifiedKFold(n_splits=5, n_repeats=10, ran-

dom_state=None)
Repeated Stratified K-Fold cross validator.
Repeats Stratified K-Fold n times with different randomization in each repetition.
Parameters
n_repeats [int, default=10] Number of times cross-validator needs to be repeated.
random_state [None, int or RandomState, default=None] Random state to be used to generate
random state for each repetition.
See also:
RepeatedKFold Repeats K-Fold n times.
Notes
Examples
>>> from sklearn.model_selection import RepeatedStratifiedKFold

>>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4]])
>>> y = np.array([0, 0, 1, 1])
>>> rskf = RepeatedStratifiedKFold(n_splits=2, n_repeats=2,

... random_state=36851234)
>>> for train_index, test_index in rskf.split(X, y):
...
TRAIN: [1 2] TEST: [0 3]
TRAIN: [0 3] TEST: [1 2]
TRAIN: [1 3] TEST: [0 2]
TRAIN: [0 2] TEST: [1 3]
Methods

validator
split(X[, y, groups]) Generates indices to split data into training and test set.
__init__(n_splits=5, n_repeats=10, random_state=None)

Parameters
X [object] Always ignored, exists for compatibility. np.zeros(n_samples) may be
y [object] Always ignored, exists for compatibility. np.zeros(n_samples) may be
Returns
Generates indices to split data into training and test set.
Parameters
Yields

sklearn.model_selection.ShuffleSplit
class sklearn.model_selection.ShuffleSplit(n_splits=10, test_size=’default’,

train_size=None, random_state=None)
Random permutation cross-validator
Yields indices to split data into training and test sets.
Note: contrary to other cross-validation strategies, random splits do not guarantee that all folds will be different,
although this is still very likely for sizeable datasets.
Parameters
n_splits [int, default 10] Number of re-shuffling & splitting iterations.
test_size [float, int, None, default=0.1] If float, should be between 0.0 and 1.0 and represent the
proportion of the dataset to include in the test split. If int, represents the absolute number of
test samples. If None, the value is set to the complement of the train size. By default (the
parameter is unspecified), the value is set to 0.1. The default will change in version 0.21.
It will remain 0.1 only if train_size is unspecified, otherwise it will complement the
specified train_size.
train_size [float, int, or None, default=None] If float, should be between 0.0 and 1.0 and repre-
sent the proportion of the dataset to include in the train split. If int, represents the absolute
number of train samples. If None, the value is automatically set to the complement of the
test size.
Examples

>>> X = np.array([[1, 2], [3, 4], [5, 6], [7, 8], [3, 4], [5, 6]])
>>> y = np.array([1, 2, 1, 2, 1, 2])
>>> rs = ShuffleSplit(n_splits=5, test_size=.25, random_state=0)
>>> rs.get_n_splits(X)
5
>>> print(rs)
ShuffleSplit(n_splits=5, random_state=0, test_size=0.25, train_size=None)
>>> for train_index, test_index in rs.split(X):
...
TRAIN: [1 3 0 4] TEST: [5 2]
TRAIN: [4 0 2 5] TEST: [1 3]
TRAIN: [1 2 4 0] TEST: [3 5]
TRAIN: [3 4 1 0] TEST: [5 2]
TRAIN: [3 5 1 0] TEST: [2 4]
>>> rs = ShuffleSplit(n_splits=5, train_size=0.5, test_size=.25,
... random_state=0)
>>> for train_index, test_index in rs.split(X):
...
TRAIN: [1 3 0] TEST: [5 2]

TRAIN: [4 0 2] TEST: [1 3]
TRAIN: [1 2 4] TEST: [3 5]
TRAIN: [3 4 1] TEST: [5 2]
TRAIN: [3 5 1] TEST: [2 4]
Methods

validator

Parameters
Returns
Parameters
Yields
Notes
Examples using sklearn.model_selection.ShuffleSplit


sklearn.model_selection.StratifiedKFold
class sklearn.model_selection.StratifiedKFold(n_splits=’warn’, shuffle=False, ran-

dom_state=None)
Stratified K-Folds cross-validator
Provides train/test indices to split data in train/test sets.
This cross-validation object is a variation of KFold that returns stratified folds. The folds are made by preserving
the percentage of samples for each class.
Parameters
shuffle [boolean, optional] Whether to shuffle each stratification of the data before splitting into
batches.
the RandomState instance used by np.random. Used when shuffle == True.
See also:
RepeatedStratifiedKFold Repeats Stratified K-Fold n times.
Notes
Train and test sizes may be different in each fold, with a difference of at most n_classes.
Examples
>>> from sklearn.model_selection import StratifiedKFold

>>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4]])
>>> y = np.array([0, 0, 1, 1])
>>> skf = StratifiedKFold(n_splits=2)
>>> skf.get_n_splits(X, y)
2
>>> print(skf)
StratifiedKFold(n_splits=2, random_state=None, shuffle=False)
>>> for train_index, test_index in skf.split(X, y):
TRAIN: [1 3] TEST: [0 2]
TRAIN: [0 2] TEST: [1 3]

Methods

validator
split(X, y[, groups]) Generate indices to split data into training and test set.
__init__(n_splits=’warn’, shuffle=False, random_state=None)

Parameters
Returns
split(X, y, groups=None)
Parameters
Note that providing y is sufficient to generate the splits and hence np.
zeros(n_samples) may be used as a placeholder for X instead of actual training data.
Stratification is done based on the y labels.
Yields
Notes
Examples using sklearn.model_selection.StratifiedKFold

• GMM covariances

sklearn.model_selection.StratifiedShuffleSplit
class sklearn.model_selection.StratifiedShuffleSplit(n_splits=10, test_size=’default’,

train_size=None, ran-
dom_state=None)
Stratified ShuffleSplit cross-validator
Provides train/test indices to split data in train/test sets.
This cross-validation object is a merge of StratifiedKFold and ShuffleSplit, which returns stratified randomized
folds. The folds are made by preserving the percentage of samples for each class.
Note: like the ShuffleSplit strategy, stratified random splits do not guarantee that all folds will be different,
although this is still very likely for sizeable datasets.
Parameters
n_splits [int, default 10] Number of re-shuffling & splitting iterations.
test_size [float, int, None, optional] If float, should be between 0.0 and 1.0 and represent the
proportion of the dataset to include in the test split. If int, represents the absolute number
of test samples. If None, the value is set to the complement of the train size. By default,
the value is set to 0.1. The default will change in version 0.21. It will remain 0.1 only if
train_size is unspecified, otherwise it will complement the specified train_size.
train_size [float, int, or None, default is None] If float, should be between 0.0 and 1.0 and
represent the proportion of the dataset to include in the train split. If int, represents the
absolute number of train samples. If None, the value is automatically set to the complement
of the test size.
Examples
>>> from sklearn.model_selection import StratifiedShuffleSplit

>>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4], [1, 2], [3, 4]])
>>> y = np.array([0, 0, 0, 1, 1, 1])
>>> sss = StratifiedShuffleSplit(n_splits=5, test_size=0.5, random_state=0)
>>> sss.get_n_splits(X, y)
5
>>> print(sss)
StratifiedShuffleSplit(n_splits=5, random_state=0, ...)
>>> for train_index, test_index in sss.split(X, y):
TRAIN: [5 2 3] TEST: [4 1 0]
TRAIN: [5 1 4] TEST: [0 2 3]
TRAIN: [5 0 2] TEST: [4 3 1]
TRAIN: [4 1 0] TEST: [2 3 5]
TRAIN: [0 5 1] TEST: [3 4 2]

Methods

validator
split(X, y[, groups]) Generate indices to split data into training and test set.

Parameters
Returns
split(X, y, groups=None)
Parameters
Note that providing y is sufficient to generate the splits and hence np.
zeros(n_samples) may be used as a placeholder for X instead of actual training data.
Stratification is done based on the y labels.
Yields
Notes
Examples using sklearn.model_selection.StratifiedShuffleSplit


sklearn.model_selection.TimeSeriesSplit
class sklearn.model_selection.TimeSeriesSplit(n_splits=’warn’, max_train_size=None)

Time Series cross-validator
Provides train/test indices to split time series data samples that are observed at fixed time intervals, in train/test
sets. In each split, test indices must be higher than before, and thus shuffling in cross validator is inappropriate.
This cross-validation object is a variation of KFold. In the kth split, it returns first k folds as train set and the
(k+1)th fold as test set.
Note that unlike standard cross-validation methods, successive training sets are supersets of those that come
before them.
Parameters
n_splits [int, default=3] Number of splits. Must be at least 2.
max_train_size [int, optional] Maximum size for a single training set.
Notes
The training set has size i * n_samples // (n_splits + 1) + n_samples % (n_splits

+ 1) in the i``th split, with a test set of size ``n_samples//(n_splits + 1),
where n_samples is the number of samples.
Examples
>>> from sklearn.model_selection import TimeSeriesSplit

>>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4], [1, 2], [3, 4]])
>>> y = np.array([1, 2, 3, 4, 5, 6])
>>> tscv = TimeSeriesSplit(n_splits=5)
>>> print(tscv)
TimeSeriesSplit(max_train_size=None, n_splits=5)
>>> for train_index, test_index in tscv.split(X):
TRAIN: [0 1] TEST: [2]
TRAIN: [0 1 2] TEST: [3]
TRAIN: [0 1 2 3] TEST: [4]
TRAIN: [0 1 2 3 4] TEST: [5]
Methods

validator

__init__(n_splits=’warn’, max_train_size=None)
Parameters
Returns
Parameters
y [array-like, shape (n_samples,)] Always ignored, exists for compatibility.
groups [array-like, with shape (n_samples,)] Always ignored, exists for compatibility.
Yields
Examples using sklearn.model_selection.TimeSeriesSplit
6.25.2 Splitter Functions
model_selection.check_cv([cv, y, classifier]) Input checker utility for building a cross-validator

model_selection.train_test_split(*arrays, Split arrays or matrices into random train and test subsets
...)
sklearn.model_selection.check_cv
sklearn.model_selection.check_cv(cv=’warn’, y=None, classifier=False)

Input checker utility for building a cross-validator
Parameters
• CV splitter,


For integer/None inputs, if classifier is True and y is either binary or multiclass,
StratifiedKFold is used. In all other cases, KFold is used.
Changed in version 0.20: cv default value will change from 3-fold to 5-fold in v0.22.
y [array-like, optional] The target variable for supervised learning problems.
classifier [boolean, optional, default False] Whether the task is a classification task, in which
case stratified KFold will be used.
Returns
checked_cv [a cross-validator instance.] The return value is a cross-validator which generates
the train/test splits via the split method.
sklearn.model_selection.train_test_split
sklearn.model_selection.train_test_split(*arrays, **options)
Split arrays or matrices into random train and test subsets
Quick utility that wraps input validation and next(ShuffleSplit().split(X, y)) and application to
input data into a single call for splitting (and optionally subsampling) data in a oneliner.
Parameters
*arrays [sequence of indexables with same length / shape[0]] Allowed inputs are lists, numpy
arrays, scipy-sparse matrices or pandas dataframes.
test_size [float, int or None, optional (default=0.25)] If float, should be between 0.0 and 1.0
and represent the proportion of the dataset to include in the test split. If int, represents the
absolute number of test samples. If None, the value is set to the complement of the train
size. By default, the value is set to 0.25. The default will change in version 0.21. It will
remain 0.25 only if train_size is unspecified, otherwise it will complement the specified
train_size.
train_size [float, int, or None, (default=None)] If float, should be between 0.0 and 1.0 and
represent the proportion of the dataset to include in the train split. If int, represents the
absolute number of train samples. If None, the value is automatically set to the complement
of the test size.
shuffle [boolean, optional (default=True)] Whether or not to shuffle the data before splitting. If
shuffle=False then stratify must be None.
stratify [array-like or None (default=None)] If not None, data is split in a stratified fashion,
using this as the class labels.
Returns
splitting [list, length=2 * len(arrays)] List containing train-test split of inputs.
New in version 0.16: If the input is sparse, the output will be a scipy.sparse.
csr_matrix. Else, output type is the same as the input type.

Examples

>>> X, y = np.arange(10).reshape((5, 2)), range(5)
>>> X
array([[0, 1],
[2, 3],
[4, 5],
[6, 7],
[8, 9]])
>>> list(y)
[0, 1, 2, 3, 4]

... X, y, test_size=0.33, random_state=42)
...
>>> X_train
array([[4, 5],
[0, 1],
[6, 7]])
>>> y_train
[2, 0, 3]
>>> X_test
array([[2, 3],
[8, 9]])
>>> y_test
[1, 4]
>>> train_test_split(y, shuffle=False)

[[0, 1, 2], [3, 4]]
Examples using sklearn.model_selection.train_test_split



• Map data to a normal distribution
6.25.3 Hyper-parameter optimizers
model_selection.GridSearchCV (estimator, . . . ) Exhaustive search over specified parameter values for an

estimator.
model_selection.ParameterGrid(param_grid) Grid of parameters with a discrete number of values for
each.
model_selection.ParameterSampler(. . . [, . . . ]) Generator on parameters sampled from given distributions.
model_selection.RandomizedSearchCV (. . . [, Randomized search on hyper parameters.
. . . ])
sklearn.model_selection.GridSearchCV
class sklearn.model_selection.GridSearchCV(estimator, param_grid, scoring=None,

fit_params=None, n_jobs=None, iid=’warn’,
refit=True, cv=’warn’, verbose=0,
pre_dispatch=‘2*n_jobs’, error_score=’raise-
deprecating’, return_train_score=’warn’)
Exhaustive search over specified parameter values for an estimator.
Important members are fit, predict.
GridSearchCV implements a “fit” and a “score” method. It also implements “predict”, “predict_proba”, “deci-
sion_function”, “transform” and “inverse_transform” if they are implemented in the estimator used.
The parameters of the estimator used to apply these methods are optimized by cross-validated grid-search over
a parameter grid.
Parameters

estimator [estimator object.] This is assumed to implement the scikit-learn estimator interface.
Either estimator needs to provide a score function, or scoring must be passed.
param_grid [dict or list of dictionaries] Dictionary with parameters names (string) as keys and
lists of parameter settings to try as values, or a list of such dictionaries, in which case the
grids spanned by each dictionary in the list are explored. This enables searching over any
sequence of parameter settings.
scoring [string, callable, list/tuple, dict or None, default: None] A single string (see The scoring
parameter: defining model evaluation rules) or a callable (see Defining your scoring strategy
from metric functions) to evaluate the predictions on the test set.
For evaluating multiple metrics, either give a list of (unique) strings or a dict with names as
keys and callables as values.
NOTE that when using custom scorers, each scorer should return a single value. Metric
functions returning a list/array of values can be wrapped into multiple scorers that return
one value each.
See Specifying multiple metrics for evaluation for an example.
If None, the estimator’s default scorer (if available) is used.
fit_params [dict, optional] Parameters to pass to the fit method.
Deprecated since version 0.19: fit_params as a constructor argument was deprecated in
version 0.19 and will be removed in version 0.21. Pass fit parameters to the fit method
instead.
pre_dispatch [int, or string, optional] Controls the number of jobs that get dispatched during
parallel execution. Reducing this number can be useful to avoid an explosion of memory
consumption when more jobs get dispatched than CPUs can process. This parameter can
be:
• None, in which case all the jobs are immediately created and spawned. Use this for
lightweight and fast-running jobs, to avoid delays due to on-demand spawning of the jobs
• An int, giving the exact number of total jobs that are spawned
• A string, giving an expression as a function of n_jobs, as in ‘2*n_jobs’
iid [boolean, default=’warn’] If True, return the average score across folds, weighted by the
number of samples in each test set. In this case, the data is assumed to be identically dis-
tributed across the folds, and the loss minimized is the total loss per sample, and not the
mean loss across the folds. If False, return the average score across folds. Default is True,
but will change to False in version 0.21, to correspond to the standard definition of cross-
validation.
Changed in version 0.20: Parameter iid will change from True to False by default in
version 0.22, and will be removed in 0.24.
• None, to use the default 3-fold cross validation,
• integer, to specify the number of folds in a (Stratified)KFold,
• CV splitter,


For integer/None inputs, if the estimator is a classifier and y is either binary or multiclass,
v0.22.
refit [boolean, or string, default=True] Refit an estimator using the best found parameters on
the whole dataset.
For multiple metric evaluation, this needs to be a string denoting the scorer is used to find
the best parameters for refitting the estimator at the end.
The refitted estimator is made available at the best_estimator_ attribute and permits
using predict directly on this GridSearchCV instance.
Also for multiple metric evaluation, the attributes best_index_, best_score_ and
best_params_ will only be available if refit is set and all of them will be determined
w.r.t this specific scorer.
See scoring parameter to know more about multiple metric evaluation.
verbose [integer] Controls the verbosity: the higher, the more messages.
error_score [‘raise’ or numeric] Value to assign to the score if an error occurs in estimator
fitting. If set to ‘raise’, the error is raised. If a numeric value is given, FitFailedWarning
is raised. This parameter does not affect the refit step, which will always raise the error.
Default is ‘raise’ but from version 0.22 it will change to np.nan.
return_train_score [boolean, optional] If False, the cv_results_ attribute will not in-
clude training scores.
Current default is 'warn', which behaves as True in addition to raising a warning when
a training score is looked up. That default will be changed to False in 0.21. Comput-
ing training scores is used to get insights on how different parameter settings impact the
overfitting/underfitting trade-off. However computing the scores on the training set can be
computationally expensive and is not strictly required to select the parameters that yield the
best generalization performance.
Attributes
cv_results_ [dict of numpy (masked) ndarrays] A dict with keys as column headers and values
as columns, that can be imported into a pandas DataFrame.
For instance the below given table
param_kernel param_gamma param_degree split0_test_score ... rank_t. . .

‘poly’ – 2 0.80 ... 2
‘poly’ – 3 0.70 ... 4
‘rbf’ 0.1 – 0.80 ... 3
‘rbf’ 0.2 – 0.93 ... 1
will be represented by a cv_results_ dict of:
{
'param_kernel': masked_array(data = ['poly', 'poly', 'rbf', 'rbf'],
mask = [False False False False]...)
'param_gamma': masked_array(data = [-- -- 0.1 0.2],

mask = [ True True False False]...),

'param_degree': masked_array(data = [2.0 3.0 -- --],
mask = [False False True True]...),
'split0_test_score' : [0.80, 0.70, 0.80, 0.93],
'split1_test_score' : [0.82, 0.50, 0.70, 0.78],
'mean_test_score' : [0.81, 0.60, 0.75, 0.85],
'std_test_score' : [0.01, 0.10, 0.05, 0.08],
'rank_test_score' : [2, 4, 3, 1],
'split0_train_score' : [0.80, 0.92, 0.70, 0.93],
'split1_train_score' : [0.82, 0.55, 0.70, 0.87],
'mean_train_score' : [0.81, 0.74, 0.70, 0.90],
'std_train_score' : [0.01, 0.19, 0.00, 0.03],
'mean_fit_time' : [0.73, 0.63, 0.43, 0.49],
'std_fit_time' : [0.01, 0.02, 0.01, 0.01],
'mean_score_time' : [0.01, 0.06, 0.04, 0.04],
'std_score_time' : [0.00, 0.00, 0.00, 0.01],
'params' : [{'kernel': 'poly', 'degree': 2}, ...],
}
NOTE
The key 'params' is used to store a list of parameter settings dicts for all the parameter
candidates.
The mean_fit_time, std_fit_time, mean_score_time and
std_score_time are all in seconds.
For multi-metric evaluation, the scores for all the scorers are available in the
cv_results_ dict at the keys ending with that scorer’s name ('_<scorer_name>')
instead of '_score' shown above. (‘split0_test_precision’, ‘mean_train_precision’ etc.)
best_estimator_ [estimator or dict] Estimator that was chosen by the search, i.e. estimator
which gave highest score (or smallest loss if specified) on the left out data. Not available if
refit=False.
See refit parameter for more information on allowed values.
best_score_ [float] Mean cross-validated score of the best_estimator
For multi-metric evaluation, this is present only if refit is specified.
best_params_ [dict] Parameter setting that gave the best results on the hold out data.
best_index_ [int] The index (of the cv_results_ arrays) which corresponds to the best can-
didate parameter setting.
The dict at search.cv_results_['params'][search.best_index_] gives
the parameter setting for the best model, that gives the highest mean score (search.
best_score_).
scorer_ [function or a dict] Scorer function used on the held out data to choose the best param-
eters for the model.
For multi-metric evaluation, this attribute holds the validated scoring dict which maps
the scorer key to the scorer callable.
n_splits_ [int] The number of cross-validation splits (folds/iterations).

refit_time_ [float] Seconds used for refitting the best model on the whole dataset.
This is present only if refit is not False.
See also:
ParameterGrid generates all the combinations of a hyperparameter grid.

sklearn.model_selection.train_test_split utility function to split the data into a development
set usable for fitting a GridSearchCV instance and an evaluation set for its final evaluation.
sklearn.metrics.make_scorer Make a scorer from a performance metric or loss function.
Notes
The parameters selected are those that maximize the score of the left out data, unless an explicit score is passed
in which case it is used instead.
If n_jobs was set to a value higher than one, the data is copied for each point in the grid (and not n_jobs times).
This is done for efficiency reasons if individual jobs take very little time, but may raise errors if the dataset is
large and not enough memory is available. A workaround in this case is to set pre_dispatch. Then, the memory
is copied only pre_dispatch many times. A reasonable value for pre_dispatch is 2 * n_jobs.
Examples
>>> from sklearn import svm, datasets

>>> parameters = {'kernel':('linear', 'rbf'), 'C':[1, 10]}
>>> svc = svm.SVC(gamma="scale")
>>> clf = GridSearchCV(svc, parameters, cv=5)
>>> clf.fit(iris.data, iris.target)
...
GridSearchCV(cv=5, error_score=...,
estimator=SVC(C=1.0, cache_size=..., class_weight=..., coef0=...,
decision_function_shape='ovr', degree=..., gamma=...,
kernel='rbf', max_iter=-1, probability=False,
random_state=None, shrinking=True, tol=...,
verbose=False),
fit_params=None, iid=..., n_jobs=None,
param_grid=..., pre_dispatch=..., refit=..., return_train_score=...,
scoring=..., verbose=...)
>>> sorted(clf.cv_results_.keys())
...
['mean_fit_time', 'mean_score_time', 'mean_test_score',...
'mean_train_score', 'param_C', 'param_kernel', 'params',...
'rank_test_score', 'split0_test_score',...
'split0_train_score', 'split1_test_score', 'split1_train_score',...
'split2_test_score', 'split2_train_score',...
'std_fit_time', 'std_score_time', 'std_test_score', 'std_train_score'...]
Methods

decision_function(X) Call decision_function on the estimator with the best

found parameters.
fit(X[, y, groups]) Run fit with all sets of parameters.
inverse_transform(Xt) Call inverse_transform on the estimator with the best
found params.
predict(X) Call predict on the estimator with the best found param-
eters.
predict_log_proba(X) Call predict_log_proba on the estimator with the best
found parameters.
predict_proba(X) Call predict_proba on the estimator with the best found
parameters.
score(X[, y]) Returns the score on the given data, if the estimator has
been refit.
transform(X) Call transform on the estimator with the best found pa-
rameters.
__init__(estimator, param_grid, scoring=None, fit_params=None, n_jobs=None, iid=’warn’,

refit=True, cv=’warn’, verbose=0, pre_dispatch=‘2*n_jobs’, error_score=’raise-
deprecating’, return_train_score=’warn’)
Call decision_function on the estimator with the best found parameters.
Only available if refit=True and the underlying estimator supports decision_function.
Parameters
X [indexable, length n_samples] Must fulfill the input assumptions of the underlying esti-
mator.
fit(X, y=None, groups=None, **fit_params)
Run fit with all sets of parameters.
Parameters
y [array-like, shape = [n_samples] or [n_samples, n_output], optional] Target relative to X
for classification or regression; None for unsupervised learning.
**fit_params [dict of string -> object] Parameters passed to the fit method of the estima-
tor
Parameters
Returns

inverse_transform(Xt)
Call inverse_transform on the estimator with the best found params.
Only available if the underlying estimator implements inverse_transform and refit=True.
Parameters
Xt [indexable, length n_samples] Must fulfill the input assumptions of the underlying esti-
mator.
predict(X)
Call predict on the estimator with the best found parameters.
Only available if refit=True and the underlying estimator supports predict.
Parameters
mator.
Call predict_log_proba on the estimator with the best found parameters.
Only available if refit=True and the underlying estimator supports predict_log_proba.
Parameters
mator.
predict_proba(X)
Call predict_proba on the estimator with the best found parameters.
Only available if refit=True and the underlying estimator supports predict_proba.
Parameters
mator.
score(X, y=None)
Returns the score on the given data, if the estimator has been refit.
This uses the score defined by scoring where provided, and the best_estimator_.score method
otherwise.
Parameters
X [array-like, shape = [n_samples, n_features]] Input data, where n_samples is the number
Returns
score [float]
of a nested object.
Returns

self
transform(X)
Call transform on the estimator with the best found parameters.
Only available if the underlying estimator supports transform and refit=True.
Parameters
mator.
Examples using sklearn.model_selection.GridSearchCV

sklearn.model_selection.ParameterGrid
class sklearn.model_selection.ParameterGrid(param_grid)
Grid of parameters with a discrete number of values for each.
Can be used to iterate over parameter value combinations with the Python built-in function iter.
Parameters

param_grid [dict of string to sequence, or sequence of such] The parameter grid to explore, as
a dictionary mapping estimator parameters to sequences of allowed values.
An empty dict signifies default parameters.
A sequence of dicts signifies a sequence of grids to search, and is useful to avoid exploring
parameter combinations that make no sense or have no effect. See the examples below.
See also:
GridSearchCV
Uses class:ParameterGrid to perform a full parallelized parameter search.
Examples
>>> from sklearn.model_selection import ParameterGrid

>>> param_grid = {'a': [1, 2], 'b': [True, False]}
>>> list(ParameterGrid(param_grid)) == (
... [{'a': 1, 'b': True}, {'a': 1, 'b': False},
... {'a': 2, 'b': True}, {'a': 2, 'b': False}])
True
>>> grid = [{'kernel': ['linear']}, {'kernel': ['rbf'], 'gamma': [1, 10]}]

>>> list(ParameterGrid(grid)) == [{'kernel': 'linear'},
... {'kernel': 'rbf', 'gamma': 1},
... {'kernel': 'rbf', 'gamma': 10}]
True
>>> ParameterGrid(grid)[1] == {'kernel': 'rbf', 'gamma': 1}
True
__init__(param_grid)
sklearn.model_selection.ParameterSampler
class sklearn.model_selection.ParameterSampler(param_distributions, n_iter, ran-

dom_state=None)
Generator on parameters sampled from given distributions.
Non-deterministic iterable over random candidate combinations for hyperparameter search. If all parameters
are presented as a list, sampling without replacement is performed. If at least one parameter is given as a
distribution, sampling with replacement is used. It is highly recommended to use continuous distributions for
continuous parameters.
Note that before SciPy 0.16, the scipy.stats.distributions do not accept a custom RNG instance
and always use the singleton RNG from numpy.random. Hence setting random_state will not guarantee
a deterministic iteration whenever scipy.stats distributions are used to define the parameter search space.
Deterministic behavior is however guaranteed from SciPy 0.16 onwards.
Parameters
param_distributions [dict] Dictionary where the keys are parameters and values are distribu-
tions from which a parameter is to be sampled. Distributions either have to provide a rvs
function to sample from them, or can be given as a list of values, where a uniform distribu-
tion is assumed.

n_iter [integer] Number of parameter settings that are produced.

random_state [int, RandomState instance or None, optional (default=None)] Pseudo random
number generator state used for random uniform sampling from lists of possible values
instead of scipy.stats distributions. If int, random_state is the seed used by the random
number generator; If RandomState instance, random_state is the random number generator;
If None, the random number generator is the RandomState instance used by np.random.
Returns
params [dict of string to any] Yields dictionaries mapping each estimator parameter to as sam-
pled value.
Examples
>>> from sklearn.model_selection import ParameterSampler

>>> from scipy.stats.distributions import expon
>>> param_grid = {'a':[1, 2], 'b': expon()}
>>> param_list = list(ParameterSampler(param_grid, n_iter=4))
>>> rounded_list = [dict((k, round(v, 6)) for (k, v) in d.items())
... for d in param_list]
>>> rounded_list == [{'b': 0.89856, 'a': 1},
... {'b': 0.923223, 'a': 1},
... {'b': 1.878964, 'a': 2},
... {'b': 1.038159, 'a': 2}]
True
__init__(param_distributions, n_iter, random_state=None)
sklearn.model_selection.RandomizedSearchCV
class sklearn.model_selection.RandomizedSearchCV(estimator, param_distributions,

n_iter=10, scoring=None,
fit_params=None, n_jobs=None,
iid=’warn’, refit=True, cv=’warn’, ver-
bose=0, pre_dispatch=‘2*n_jobs’,
random_state=None,
error_score=’raise-deprecating’, re-
turn_train_score=’warn’)
Randomized search on hyper parameters.
RandomizedSearchCV implements a “fit” and a “score” method. It also implements “predict”, “predict_proba”,
“decision_function”, “transform” and “inverse_transform” if they are implemented in the estimator used.
The parameters of the estimator used to apply these methods are optimized by cross-validated search over
parameter settings.
In contrast to GridSearchCV, not all parameter values are tried out, but rather a fixed number of parameter
settings is sampled from the specified distributions. The number of parameter settings that are tried is given by
n_iter.
If all parameters are presented as a list, sampling without replacement is performed. If at least one parameter
is given as a distribution, sampling with replacement is used. It is highly recommended to use continuous
distributions for continuous parameters.

Note that before SciPy 0.16, the scipy.stats.distributions do not accept a custom RNG instance
and always use the singleton RNG from numpy.random. Hence setting random_state will not guarantee
a deterministic iteration whenever scipy.stats distributions are used to define the parameter search space.
Parameters
estimator [estimator object.] A object of that type is instantiated for each grid point. This is
assumed to implement the scikit-learn estimator interface. Either estimator needs to provide
a score function, or scoring must be passed.
param_distributions [dict] Dictionary with parameters names (string) as keys and distributions
or lists of parameters to try. Distributions must provide a rvs method for sampling (such
as those from scipy.stats.distributions). If a list is given, it is sampled uniformly.
n_iter [int, default=10] Number of parameter settings that are sampled. n_iter trades off runtime
vs quality of the solution.
one value each.
fit_params [dict, optional] Parameters to pass to the fit method.
Deprecated since version 0.19: fit_params as a constructor argument was deprecated in
version 0.19 and will be removed in version 0.21. Pass fit parameters to the fit method
instead.
be:
iid [boolean, default=’warn’] If True, return the average score across folds, weighted by the
number of samples in each test set. In this case, the data is assumed to be identically dis-
tributed across the folds, and the loss minimized is the total loss per sample, and not the
mean loss across the folds. If False, return the average score across folds. Default is True,
but will change to False in version 0.21, to correspond to the standard definition of cross-
validation.

Changed in version 0.20: Parameter iid will change from True to False by default in
version 0.22, and will be removed in 0.24.
• CV splitter,
v0.22.
refit [boolean, or string default=True] Refit an estimator using the best found parameters on the
whole dataset.
For multiple metric evaluation, this needs to be a string denoting the scorer that would be
used to find the best parameters for refitting the estimator at the end.
The refitted estimator is made available at the best_estimator_ attribute and permits
using predict directly on this RandomizedSearchCV instance.
Also for multiple metric evaluation, the attributes best_index_, best_score_ and
best_params_ will only be available if refit is set and all of them will be determined
w.r.t this specific scorer.
See scoring parameter to know more about multiple metric evaluation.
verbose [integer] Controls the verbosity: the higher, the more messages.
random_state [int, RandomState instance or None, optional, default=None] Pseudo random
number generator state used for random uniform sampling from lists of possible values
instead of scipy.stats distributions. If int, random_state is the seed used by the random
number generator; If RandomState instance, random_state is the random number generator;
If None, the random number generator is the RandomState instance used by np.random.
return_train_score [boolean, optional] If False, the cv_results_ attribute will not in-
clude training scores.
Attributes
cv_results_ [dict of numpy (masked) ndarrays] A dict with keys as column headers and values
as columns, that can be imported into a pandas DataFrame.

For instance the below given table
param_kernel param_gamma split0_test_score ... rank_test_score

‘rbf’ 0.1 0.80 ... 2
‘rbf’ 0.2 0.90 ... 1
‘rbf’ 0.3 0.70 ... 1
will be represented by a cv_results_ dict of:
{
'param_kernel' : masked_array(data = ['rbf', 'rbf', 'rbf'],
mask = False),
'param_gamma' : masked_array(data = [0.1 0.2 0.3], mask = False),
'split0_test_score' : [0.80, 0.90, 0.70],
'split1_test_score' : [0.82, 0.50, 0.70],
'mean_test_score' : [0.81, 0.70, 0.70],
'std_test_score' : [0.01, 0.20, 0.00],
'rank_test_score' : [3, 1, 1],
'split0_train_score' : [0.80, 0.92, 0.70],
'split1_train_score' : [0.82, 0.55, 0.70],
'mean_train_score' : [0.81, 0.74, 0.70],
'std_train_score' : [0.01, 0.19, 0.00],
'mean_fit_time' : [0.73, 0.63, 0.43],
'std_fit_time' : [0.01, 0.02, 0.01],
'mean_score_time' : [0.01, 0.06, 0.04],
'std_score_time' : [0.00, 0.00, 0.00],
'params' : [{'kernel' : 'rbf', 'gamma' : 0.1}, ...],
}
NOTE
The key 'params' is used to store a list of parameter settings dicts for all the parameter
candidates.
The mean_fit_time, std_fit_time, mean_score_time and
std_score_time are all in seconds.
For multi-metric evaluation, the scores for all the scorers are available in the
cv_results_ dict at the keys ending with that scorer’s name ('_<scorer_name>')
instead of '_score' shown above. (‘split0_test_precision’, ‘mean_train_precision’ etc.)
best_estimator_ [estimator or dict] Estimator that was chosen by the search, i.e. estimator
which gave highest score (or smallest loss if specified) on the left out data. Not available if
refit=False.
For multi-metric evaluation, this attribute is present only if refit is specified.
See refit parameter for more information on allowed values.
best_score_ [float] Mean cross-validated score of the best_estimator.
For multi-metric evaluation, this is not available if refit is False. See refit parameter
for more information.
best_params_ [dict] Parameter setting that gave the best results on the hold out data.
best_index_ [int] The index (of the cv_results_ arrays) which corresponds to the best can-
didate parameter setting.

The dict at search.cv_results_['params'][search.best_index_] gives

the parameter setting for the best model, that gives the highest mean score (search.
best_score_).
scorer_ [function or a dict] Scorer function used on the held out data to choose the best param-
eters for the model.
For multi-metric evaluation, this attribute holds the validated scoring dict which maps
the scorer key to the scorer callable.
n_splits_ [int] The number of cross-validation splits (folds/iterations).
refit_time_ [float] Seconds used for refitting the best model on the whole dataset.
This is present only if refit is not False.
See also:
GridSearchCV Does exhaustive search over a grid of parameters.

ParameterSampler A generator over parameter settings, constructed from param_distributions.
Notes
The parameters selected are those that maximize the score of the held-out data, according to the scoring param-
eter.
If n_jobs was set to a value higher than one, the data is copied for each parameter setting(and not n_jobs times).
This is done for efficiency reasons if individual jobs take very little time, but may raise errors if the dataset is
large and not enough memory is available. A workaround in this case is to set pre_dispatch. Then, the memory
is copied only pre_dispatch many times. A reasonable value for pre_dispatch is 2 * n_jobs.
Methods
decision_function(X) Call decision_function on the estimator with the best

found parameters.
fit(X[, y, groups]) Run fit with all sets of parameters.
inverse_transform(Xt) Call inverse_transform on the estimator with the best
found params.
predict(X) Call predict on the estimator with the best found param-
eters.
predict_log_proba(X) Call predict_log_proba on the estimator with the best
found parameters.
predict_proba(X) Call predict_proba on the estimator with the best found
parameters.
score(X[, y]) Returns the score on the given data, if the estimator has
been refit.
transform(X) Call transform on the estimator with the best found pa-
rameters.

__init__(estimator, param_distributions, n_iter=10, scoring=None, fit_params=None, n_jobs=None,

iid=’warn’, refit=True, cv=’warn’, verbose=0, pre_dispatch=‘2*n_jobs’, ran-
dom_state=None, error_score=’raise-deprecating’, return_train_score=’warn’)
Call decision_function on the estimator with the best found parameters.
Only available if refit=True and the underlying estimator supports decision_function.
Parameters
mator.
fit(X, y=None, groups=None, **fit_params)
Run fit with all sets of parameters.
Parameters
**fit_params [dict of string -> object] Parameters passed to the fit method of the estima-
tor
Parameters
Returns
Call inverse_transform on the estimator with the best found params.
Only available if the underlying estimator implements inverse_transform and refit=True.
Parameters
Xt [indexable, length n_samples] Must fulfill the input assumptions of the underlying esti-
mator.
predict(X)
Call predict on the estimator with the best found parameters.
Only available if refit=True and the underlying estimator supports predict.
Parameters
mator.
Call predict_log_proba on the estimator with the best found parameters.
Only available if refit=True and the underlying estimator supports predict_log_proba.

Parameters
mator.
predict_proba(X)
Call predict_proba on the estimator with the best found parameters.
Only available if refit=True and the underlying estimator supports predict_proba.
Parameters
mator.
score(X, y=None)
Returns the score on the given data, if the estimator has been refit.
This uses the score defined by scoring where provided, and the best_estimator_.score method
otherwise.
Parameters
X [array-like, shape = [n_samples, n_features]] Input data, where n_samples is the number
Returns
score [float]
of a nested object.
Returns
self
transform(X)
Call transform on the estimator with the best found parameters.
Only available if the underlying estimator supports transform and refit=True.
Parameters
mator.
Examples using sklearn.model_selection.RandomizedSearchCV
model_selection.fit_grid_point(X, y, . . . [, Run fit on one set of parameters.

. . . ])

sklearn.model_selection.fit_grid_point
sklearn.model_selection.fit_grid_point(X, y, estimator, parameters, train, test, scorer,

verbose, error_score=’raise-deprecating’,
**fit_params)
Run fit on one set of parameters.
Parameters
X [array-like, sparse matrix or list] Input data.
y [array-like or None] Targets for input data.
estimator [estimator object] A object of that type is instantiated for each grid point. This is
assumed to implement the scikit-learn estimator interface. Either estimator needs to provide
a score function, or scoring must be passed.
parameters [dict] Parameters to be set on estimator for this grid point.
train [ndarray, dtype int or bool] Boolean mask or indices for training set.
test [ndarray, dtype int or bool] Boolean mask or indices for test set.
scorer [callable or None] The scorer callable object / function must have its signature as
scorer(estimator, X, y).
If None the estimator’s default scorer is used.
verbose [int] Verbosity level.
**fit_params [kwargs] Additional parameter passed to the fit function of the estimator.
Returns
score [float] Score of this parameter setting on given training / test split.
parameters [dict] The parameters that have been evaluated.
n_samples_test [int] Number of test samples in this split.
6.25.4 Model validation
model_selection.cross_validate(estimator, Evaluate metric(s) by cross-validation and also record

X) fit/score times.
model_selection.cross_val_predict(estimator, Generate cross-validated estimates for each input data point
X)
model_selection.cross_val_score(estimator, Evaluate a score by cross-validation
X)
model_selection.learning_curve(estimator, X, Learning curve.
y)
model_selection.permutation_test_score(. . . )Evaluate the significance of a cross-validated score with
permutations
model_selection.validation_curve(estimator, Validation curve.
...)

sklearn.model_selection.cross_validate
sklearn.model_selection.cross_validate(estimator, X, y=None, groups=None, scor-

ing=None, cv=’warn’, n_jobs=None, verbose=0,
fit_params=None, pre_dispatch=‘2*n_jobs’, re-
turn_train_score=’warn’, return_estimator=False,
error_score=’raise-deprecating’)
Evaluate metric(s) by cross-validation and also record fit/score times.
Parameters
X [array-like] The data to fit. Can be for example a list, or an array.
y [array-like, optional, default: None] The target variable to try to predict in the case of super-
vised learning.
groups [array-like, with shape (n_samples,), optional] Group labels for the samples used while
one value each.
• CV splitter,
v0.22.
n_jobs [int or None, optional (default=None)] The number of CPUs to use to do the computa-
fit_params [dict, optional] Parameters to pass to the fit method of the estimator.

be:
return_train_score [boolean, optional] Whether to include train scores.
return_estimator [boolean, default False] Whether to return the estimators fitted on each split.
error_score [‘raise’ | ‘raise-deprecating’ or numeric] Value to assign to the score if an error
occurs in estimator fitting. If set to ‘raise’, the error is raised. If set to ‘raise-deprecating’,
a FutureWarning is printed before the error is raised. If a numeric value is given, FitFailed-
Warning is raised. This parameter does not affect the refit step, which will always raise the
error. Default is ‘raise-deprecating’ but from version 0.22 it will change to np.nan.
Returns
scores [dict of float arrays of shape=(n_splits,)] Array of scores of the estimator for each run of
the cross validation.
A dict of arrays containing the score/time arrays for each scorer is returned. The possible
keys for this dict are:
test_score The score array for test scores on each cv split.
train_score The score array for train scores on each cv split. This is available
only if return_train_score parameter is True.
fit_time The time for fitting the estimator on the train set for each cv split.
score_time The time for scoring the estimator on the test set for each cv
split. (Note time for scoring on the train set is not included even if
return_train_score is set to True
estimator The estimator objects for each cv split. This is available only if
return_estimator parameter is set to True.
See also:
sklearn.model_selection.cross_val_score Run cross-validation for single metric evaluation.

sklearn.model_selection.cross_val_predict Get predictions from each split of cross-
validation for diagnostic purposes.

Examples
>>> from sklearn import datasets, linear_model

>>> from sklearn.metrics.scorer import make_scorer
>>> X = diabetes.data[:150]
>>> y = diabetes.target[:150]
>>> lasso = linear_model.Lasso()
Single metric evaluation using cross_validate

>>> cv_results = cross_validate(lasso, X, y, cv=3,
... return_train_score=False)
>>> sorted(cv_results.keys())
['fit_time', 'score_time', 'test_score']
>>> cv_results['test_score']
array([0.33150734, 0.08022311, 0.03531764])
Multiple metric evaluation using cross_validate (please refer the scoring parameter doc for more in-
formation)
>>> scores = cross_validate(lasso, X, y, cv=3,
... scoring=('r2', 'neg_mean_squared_error'),
... return_train_score=True)
>>> print(scores['test_neg_mean_squared_error'])
[-3635.5... -3573.3... -6114.7...]
>>> print(scores['train_r2'])
[0.28010158 0.39088426 0.22784852]
sklearn.model_selection.cross_val_predict
sklearn.model_selection.cross_val_predict(estimator, X, y=None, groups=None,

cv=’warn’, n_jobs=None, verbose=0,
fit_params=None, pre_dispatch=‘2*n_jobs’,
method=’predict’)
Generate cross-validated estimates for each input data point
It is not appropriate to pass these predictions into an evaluation metric. Use cross_validate to measure
generalization error.
Parameters
estimator [estimator object implementing ‘fit’ and ‘predict’] The object to use to fit the data.
X [array-like] The data to fit. Can be, for example a list, or an array at least 2d.
vised learning.


• CV splitter,
v0.22.
be:
method [string, optional, default: ‘predict’] Invokes the passed method name of the passed
estimator. For method=’predict_proba’, the columns correspond to the classes in sorted
order.
Returns
predictions [ndarray] This is the result of calling method
See also:
cross_val_score calculate score for each CV split

cross_validate calculate one or more scores and timings for each CV split
Notes
In the case that one or more classes are absent in a training portion, a default score needs to be assigned to
all instances for that class if method produces columns per class, as in {‘decision_function’, ‘predict_proba’,
‘predict_log_proba’}. For predict_proba this value is 0. In order to ensure finite output, we approximate
negative infinity by the minimum finite float value for the dtype in other cases.

Examples

>>> from sklearn.model_selection import cross_val_predict
>>> y_pred = cross_val_predict(lasso, X, y, cv=3)
Examples using sklearn.model_selection.cross_val_predict
sklearn.model_selection.cross_val_score
sklearn.model_selection.cross_val_score(estimator, X, y=None, groups=None, scor-

ing=None, cv=’warn’, n_jobs=None, verbose=0,
fit_params=None, pre_dispatch=‘2*n_jobs’,
Evaluate a score by cross-validation
Parameters
X [array-like] The data to fit. Can be for example a list, or an array.
vised learning.
X, y).
• CV splitter,
v0.22.

be:
Returns
scores [array of float, shape=(len(list(cv)),)] Array of scores of the estimator for each run of the
cross validation.
See also:
sklearn.model_selection.cross_validate To run cross-validation on multiple metrics and also

to return train scores, fit times and score times.
sklearn.model_selection.cross_val_predict Get predictions from each split of cross-
validation for diagnostic purposes.
Examples

>>> print(cross_val_score(lasso, X, y, cv=3))
[0.33150734 0.08022311 0.03531764]
Examples using sklearn.model_selection.cross_val_score



sklearn.model_selection.learning_curve
sklearn.model_selection.learning_curve(estimator, X, y, groups=None,
train_sizes=array([0.1, 0.325, 0.55, 0.775,
1. ]), cv=’warn’, scoring=None, ex-
ploit_incremental_learning=False, n_jobs=None,
pre_dispatch=’all’, verbose=0, shuffle=False, ran-
dom_state=None, error_score=’raise-deprecating’)
Learning curve.
Determines cross-validated training and test scores for different training set sizes.
A cross-validation generator splits the whole dataset k times in training and test data. Subsets of the training set
with varying sizes will be used to train the estimator and a score for each training subset size and the test set
will be computed. Afterwards, the scores will be averaged over all k runs for each training subset size.
Parameters
estimator [object type that implements the “fit” and “predict” methods] An object of that type
which is cloned for each validation.
y [array-like, shape (n_samples) or (n_samples, n_features), optional] Target relative to X for
classification or regression; None for unsupervised learning.
train_sizes [array-like, shape (n_ticks,), dtype float or int] Relative or absolute numbers of
training examples that will be used to generate the learning curve. If the dtype is float,
it is regarded as a fraction of the maximum size of the training set (that is determined by
the selected validation method), i.e. it has to be within (0, 1]. Otherwise it is interpreted
as absolute sizes of the training sets. Note that for classification the number of samples
usually have to be big enough to contain at least one sample from each class. (default:
np.linspace(0.1, 1.0, 5))
• CV splitter,

v0.22.
X, y).
exploit_incremental_learning [boolean, optional, default: False] If the estimator supports in-
cremental learning, this will be used to speed up fitting for different training set sizes.
pre_dispatch [integer or string, optional] Number of predispatched jobs for parallel execution
(default is all). The option can reduce the allocated memory. The string can be an expression
like ‘2*n_jobs’.
verbose [integer, optional] Controls the verbosity: the higher, the more messages.
shuffle [boolean, optional] Whether to shuffle training data before taking prefixes of it based
on‘‘train_sizes‘‘.
the RandomState instance used by np.random. Used when shuffle is True.
Returns
train_sizes_abs [array, shape (n_unique_ticks,), dtype int] Numbers of training examples that
has been used to generate the learning curve. Note that the number of ticks might be less
than n_ticks because duplicate entries will be removed.
train_scores [array, shape (n_ticks, n_cv_folds)] Scores on training sets.
test_scores [array, shape (n_ticks, n_cv_folds)] Scores on test set.
Notes
See examples/model_selection/plot_learning_curve.py
Examples using sklearn.model_selection.learning_curve


sklearn.model_selection.permutation_test_score
sklearn.model_selection.permutation_test_score(estimator, X, y, groups=None,
cv=’warn’, n_permutations=100,
n_jobs=None, random_state=0, ver-
bose=0, scoring=None)
Evaluate the significance of a cross-validated score with permutations
Parameters
X [array-like of shape at least 2D] The data to fit.
y [array-like] The target variable to try to predict in the case of supervised learning.
groups [array-like, with shape (n_samples,), optional] Labels to constrain permutation within
groups, i.e. y values are permuted among samples with the same group identifier. When not
specified, y values are permuted among all samples.
When a grouped cross-validator is used, the group labels are also passed on to the split
method of the cross-validator. The cross-validator uses them for grouping the samples while
scoring [string, callable or None, optional, default: None] A single string (see The scoring
If None the estimator’s default scorer, if available, is used.
• CV splitter,
v0.22.
n_permutations [integer, optional] Number of times to permute y.
random_state [int, RandomState instance or None, optional (default=0)] If int, random_state
Returns

score [float] The true score without permuting targets.

permutation_scores [array, shape (n_permutations,)] The scores obtained for each permuta-
tions.
pvalue [float] The p-value, which approximates the probability that the score would be obtained
by chance. This is calculated as:
(C + 1) / (n_permutations + 1)
Where C is the number of permutations whose score >= the true score.
The best possible p-value is 1/(n_permutations + 1), the worst is 1.0.
Notes
This function implements Test 1 in:

Ojala and Garriga. Permutation Tests for Studying Classifier Performance. The Journal of Machine
Learning Research (2010) vol. 11
Examples using sklearn.model_selection.permutation_test_score
sklearn.model_selection.validation_curve
sklearn.model_selection.validation_curve(estimator, X, y, param_name, param_range,

groups=None, cv=’warn’, scoring=None,
n_jobs=None, pre_dispatch=’all’, verbose=0,
Validation curve.
Determine training and test scores for varying parameter values.
Compute scores for an estimator with different values of a specified parameter. This is similar to grid search
with one parameter. However, this will also compute training scores and is merely a utility for plotting the
results.
Parameters
estimator [object type that implements the “fit” and “predict” methods] An object of that type
which is cloned for each validation.
y [array-like, shape (n_samples) or (n_samples, n_features), optional] Target relative to X for
classification or regression; None for unsupervised learning.
param_name [string] Name of the parameter that will be varied.
param_range [array-like, shape (n_values,)] The values of the parameter that will be evaluated.


• CV splitter,
v0.22.
X, y).
pre_dispatch [integer or string, optional] Number of predispatched jobs for parallel execution
(default is all). The option can reduce the allocated memory. The string can be an expression
like ‘2*n_jobs’.
verbose [integer, optional] Controls the verbosity: the higher, the more messages.
Returns
train_scores [array, shape (n_ticks, n_cv_folds)] Scores on training sets.
test_scores [array, shape (n_ticks, n_cv_folds)] Scores on test set.
Notes
See Plotting Validation Curves
Examples using sklearn.model_selection.validation_curve
6.26 sklearn.multiclass: Multiclass and multilabel classification
6.26.1 Multiclass and multilabel classification strategies
This module implements multiclass learning algorithms:

• one-vs-the-rest / one-vs-all
• one-vs-one
• error correcting output codes
The estimators provided in this module are meta-estimators: they require a base estimator to be provided in their
constructor. For example, it is possible to use these estimators to turn a binary classifier or a regressor into a multiclass
classifier. It is also possible to use these estimators with multiclass estimators in the hope that their accuracy or runtime
performance improves.
All classifiers in scikit-learn implement multiclass classification; you only need to use this module if you want to
experiment with custom multiclass strategies.
The one-vs-the-rest meta-classifier also implements a predict_proba method, so long as such a method is implemented
by the base classifier. This method returns probabilities of class membership in both the single label and multilabel
case. Note that in the multilabel case, probabilities are the marginal probability that a given sample falls in the given
class. As such, in the multilabel case the sum of these probabilities over all possible labels for a given sample will not
sum to unity, as they do in the single label case.
User guide: See the Multiclass and multilabel algorithms section for further details.
multiclass.OneVsRestClassifier(estimator[, One-vs-the-rest (OvR) multiclass/multilabel strategy

. . . ])
multiclass.OneVsOneClassifier(estimator[, One-vs-one multiclass strategy
. . . ])
multiclass.OutputCodeClassifier(estimator[, (Error-Correcting) Output-Code multiclass strategy
. . . ])
6.26.2 sklearn.multiclass.OneVsRestClassifier
class sklearn.multiclass.OneVsRestClassifier(estimator, n_jobs=None)

One-vs-the-rest (OvR) multiclass/multilabel strategy
Also known as one-vs-all, this strategy consists in fitting one classifier per class. For each classifier, the class
is fitted against all the other classes. In addition to its computational efficiency (only n_classes classifiers are
needed), one advantage of this approach is its interpretability. Since each class is represented by one and one
classifier only, it is possible to gain knowledge about the class by inspecting its corresponding classifier. This is
the most commonly used strategy for multiclass classification and is a fair default choice.
This strategy can also be used for multilabel learning, where a classifier is used to predict multiple labels for
instance, by fitting on a 2-d matrix in which cell [i, j] is 1 if sample i has label j and 0 otherwise.
In the multilabel learning literature, OvR is also known as the binary relevance method.
Parameters
estimator [estimator object] An estimator object implementing fit and one of decision_function
or predict_proba.
Attributes
estimators_ [list of n_classes estimators] Estimators used for predictions.
6.26. sklearn.multiclass: Multiclass and multilabel classification 2017

classes_ [array, shape = [n_classes]] Class labels.

label_binarizer_ [LabelBinarizer object] Object used to transform multiclass labels to binary
labels and vice-versa.
multilabel_ [boolean] Whether this is a multilabel classifier
Methods
decision_function(X) Returns the distance of each sample from the decision

boundary for each class.
fit(X, y) Fit underlying estimators.
partial_fit(X, y[, classes]) Partially fit underlying estimators
predict(X) Predict multi-class targets using underlying estimators.
labels.
__init__(estimator, n_jobs=None)
Returns the distance of each sample from the decision boundary for each class. This can only be used with
estimators which implement the decision_function method.
Parameters
Returns
T [array-like, shape = [n_samples, n_classes]]
fit(X, y)
Fit underlying estimators.
Parameters
X [(sparse) array-like, shape = [n_samples, n_features]] Data.
y [(sparse) array-like, shape = [n_samples, ], [n_samples, n_classes]] Multi-class targets.
An indicator matrix turns on multilabel classification.
Returns
self
Parameters
Returns

multilabel_
Whether this is a multilabel classifier
Partially fit underlying estimators
Should be used when memory is inefficient to train all data. Chunks of data can be passed in several
iteration.
Parameters
y [(sparse) array-like, shape = [n_samples, ], [n_samples, n_classes]] Multi-class targets.
An indicator matrix turns on multilabel classification.
classes [array, shape (n_classes, )] Classes across all calls to partial_fit. Can be obtained via
np.unique(y_all), where y_all is the target vector of the entire dataset. This argument is
only required in the first call of partial_fit and can be omitted in the subsequent calls.
Returns
self
predict(X)
Predict multi-class targets using underlying estimators.
Parameters
Returns
y [(sparse) array-like, shape = [n_samples, ], [n_samples, n_classes].] Predicted multi-class
targets.
predict_proba(X)
The returned estimates for all classes are ordered by label of classes.
Note that in the multilabel case, each sample can have any number of labels. This returns the marginal
probability that the given sample has the label in question. For example, it is entirely consistent that two
labels both have a 90% probability of applying to a given sample.
In the single label multiclass case, the rows of the returned matrix sum to 1.
Parameters
Returns
T [(sparse) array-like, shape = [n_samples, n_classes]] Returns the probability of the sample
for each class in the model, where classes are ordered as they are in self.classes_.
Parameters


Returns
of a nested object.
Returns
self
Examples using sklearn.multiclass.OneVsRestClassifier
6.26.3 sklearn.multiclass.OneVsOneClassifier
class sklearn.multiclass.OneVsOneClassifier(estimator, n_jobs=None)

One-vs-one multiclass strategy
This strategy consists in fitting one classifier per class pair. At prediction time, the class which received the
most votes is selected. Since it requires to fit n_classes * (n_classes - 1) / 2 classifiers, this method is usually
slower than one-vs-the-rest, due to its O(n_classes^2) complexity. However, this method may be advantageous
for algorithms such as kernel algorithms which don’t scale well with n_samples. This is because each individual
learning problem only involves a small subset of the data whereas, with one-vs-the-rest, the complete dataset is
used n_classes times.
Parameters
or predict_proba.
Attributes
estimators_ [list of n_classes * (n_classes - 1) / 2 estimators] Estimators used for predictions.
classes_ [numpy array of shape [n_classes]] Array containing labels.
Methods

decision_function(X) Decision function for the OneVsOneClassifier.

partial_fit(X, y[, classes]) Partially fit underlying estimators
predict(X) Estimate the best class label for each sample in X.
labels.
Decision function for the OneVsOneClassifier.
The decision values for the samples are computed by adding the normalized sum of pair-wise classification
confidence levels to the votes in order to disambiguate between the decision values when the votes for all
the classes are equal leading to a tie.
Parameters
Returns
Y [array-like, shape = [n_samples, n_classes]]
fit(X, y)
Parameters
y [array-like, shape = [n_samples]] Multi-class targets.
Returns
self
Parameters
Returns
Partially fit underlying estimators
Should be used when memory is inefficient to train all data. Chunks of data can be passed in several
iteration, where the first call should have an array of all target variables.
Parameters
y [array-like, shape = [n_samples]] Multi-class targets.

classes [array, shape (n_classes, )] Classes across all calls to partial_fit. Can be obtained via
np.unique(y_all), where y_all is the target vector of the entire dataset. This argument is
only required in the first call of partial_fit and can be omitted in the subsequent calls.
Returns
self
predict(X)
Estimate the best class label for each sample in X.
This is implemented as argmax(decision_function(X), axis=1) which will return the label
of the class with most votes by estimators predicting the outcome of a decision for each possible class pair.
Parameters
Returns
y [numpy array of shape [n_samples]] Predicted multi-class targets.
Parameters
Returns
of a nested object.
Returns
self
6.26.4 sklearn.multiclass.OutputCodeClassifier
class sklearn.multiclass.OutputCodeClassifier(estimator, code_size=1.5, ran-

dom_state=None, n_jobs=None)
(Error-Correcting) Output-Code multiclass strategy
Output-code based strategies consist in representing each class with a binary code (an array of 0s and 1s). At
fitting time, one binary classifier per bit in the code book is fitted. At prediction time, the classifiers are used to
project new points in the class space and the class closest to the points is chosen. The main advantage of these
strategies is that the number of classifiers used can be controlled by the user, either for compressing the model
(0 < code_size < 1) or for making the model more robust to errors (code_size > 1). See the documentation for
more details.


Parameters
or predict_proba.
code_size [float] Percentage of the number of classes to be used to create the code book. A num-
ber between 0 and 1 will require fewer classifiers than one-vs-the-rest. A number greater
than 1 will require more classifiers than one-vs-the-rest.
used to initialize the codebook. If int, random_state is the seed used by the random number
Attributes
estimators_ [list of int(n_classes * code_size) estimators] Estimators used for predictions.
classes_ [numpy array of shape [n_classes]] Array containing labels.
code_book_ [numpy array of shape [n_classes, code_size]] Binary array containing the code of
each class.
References
[1], [2], [3]
Methods

predict(X) Predict multi-class targets using underlying estimators.
labels.
__init__(estimator, code_size=1.5, random_state=None, n_jobs=None)

fit(X, y)
Parameters
y [numpy array of shape [n_samples]] Multi-class targets.
Returns
self

Parameters
Returns
predict(X)
Predict multi-class targets using underlying estimators.
Parameters
Returns
y [numpy array of shape [n_samples]] Predicted multi-class targets.
Parameters
Returns
of a nested object.
Returns
self
6.27 sklearn.multioutput: Multioutput regression and classifica-

tion
This module implements multioutput regression and classification.

The estimators provided in this module are meta-estimators: they require a base estimator to be provided in their
constructor. The meta-estimator extends single output estimators to multioutput estimators.
User guide: See the Multiclass and multilabel algorithms section for further details.
multioutput.ClassifierChain(base_estimator) A multi-label model that arranges binary classifiers into a

chain.


multioutput.MultiOutputRegressor(estimator) Multi target regression
multioutput.MultiOutputClassifier(estimator) Multi target classification
multioutput.RegressorChain(base_estimator[, A multi-label model that arranges regressions into a chain.
. . . ])
6.27.1 sklearn.multioutput.ClassifierChain
class sklearn.multioutput.ClassifierChain(base_estimator, order=None, cv=None, ran-

dom_state=None)
A multi-label model that arranges binary classifiers into a chain.
Each model makes a prediction in the order specified by the chain using all of the available features provided to
the model plus the predictions of models that are earlier in the chain.
Parameters
base_estimator [estimator] The base estimator from which the classifier chain is built.
order [array-like, shape=[n_outputs] or ‘random’, optional] By default the order will be deter-
mined by the order of columns in the label matrix Y.:
order = [0, 1, 2, ..., Y.shape[1] - 1]
The order of the chain can be explicitly set by providing a list of integers. For example, for
a chain of length 5.:
order = [1, 3, 2, 4, 0]
means that the first model in the chain will make predictions for column 1 in the Y matrix,
the second model will make predictions for column 3, etc.
If order is ‘random’ a random ordering will be used.
cv [int, cross-validation generator or an iterable, optional (default=None)] Determines whether
to use cross validated predictions or true labels for the results of previous estimators in the
chain. If cv is None the true labels are used when fitting. Otherwise possible inputs for cv
are:
• CV splitter,
The random number generator is used to generate random chain orders.
Attributes
classes_ [list] A list of arrays of length len(estimators_) containing the class labels for
each estimator in the chain.
estimators_ [list] A list of clones of base_estimator.
order_ [list] The order of labels in the classifier chain.
6.27. sklearn.multioutput: Multioutput regression and classification 2025

See also:
RegressorChain Equivalent for regression

MultioutputClassifier Classifies each output independently rather than chaining.
References
Jesse Read, Bernhard Pfahringer, Geoff Holmes, Eibe Frank, “Classifier Chains for Multi-label Classification”,
2009.
Methods
decision_function(X) Evaluate the decision_function of the models in the

chain.
fit(X, Y) Fit the model to data matrix X and targets Y.
predict(X) Predict on the data matrix X using the ClassifierChain
model.
predict_proba(X) Predict probability estimates.
labels.
__init__(base_estimator, order=None, cv=None, random_state=None)

Evaluate the decision_function of the models in the chain.
Parameters
Returns
Y_decision [array-like, shape (n_samples, n_classes )] Returns the decision function of the
sample for each model in the chain.
fit(X, Y)
Fit the model to data matrix X and targets Y.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] The input data.
Y [array-like, shape (n_samples, n_classes)] The target values.
Returns
self [object]
Parameters

Returns
predict(X)
Predict on the data matrix X using the ClassifierChain model.
Parameters
Returns
Y_pred [array-like, shape (n_samples, n_classes)] The predicted values.
predict_proba(X)
Predict probability estimates.
Parameters
Returns
Y_prob [array-like, shape (n_samples, n_classes)]
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.multioutput.ClassifierChain
6.27.2 sklearn.multioutput.MultiOutputRegressor
class sklearn.multioutput.MultiOutputRegressor(estimator, n_jobs=None)

Multi target regression

This strategy consists of fitting one regressor per target. This is a simple strategy for extending regressors that
do not natively support multi-target regression.
Parameters
estimator [estimator object] An estimator object implementing fit and predict.
n_jobs [int or None, optional (default=None)] The number of jobs to run in parallel for fit.
When individual estimators are fast to train or predict using n_jobs>1 can result in slower
performance due to the overhead of spawning processes.
Methods
fit(X, y[, sample_weight]) Fit the model to data.

partial_fit(X, y[, sample_weight]) Incrementally fit the model to data.
predict(X) Predict multi-output variable using a model trained for
each target variable.
diction.
Fit the model to data. Fit a separate model for each output variable.
Parameters
X [(sparse) array-like, shape (n_samples, n_features)] Data.
y [(sparse) array-like, shape (n_samples, n_outputs)] Multi-output targets. An indicator ma-
trix turns on multilabel estimation.
sample_weight [array-like, shape = (n_samples) or None] Sample weights. If None, then
samples are equally weighted. Only supported if the underlying regressor supports sample
weights.
Returns
self [object]
Parameters
Returns
partial_fit(X, y, sample_weight=None)
Incrementally fit the model to data. Fit a separate model for each output variable.
Parameters


y [(sparse) array-like, shape (n_samples, n_outputs)] Multi-output targets.
weights.
Returns
self [object]
predict(X)
Predict multi-output variable using a model trained for each target variable.
Parameters
Returns
y [(sparse) array-like, shape (n_samples, n_outputs)] Multi-output targets predicted across
multiple predictors. Note: Separate models are generated for each predictor.
2).sum() and v is the regression sum of squares ((y_true - y_true.mean()) ** 2).sum(). Best possible score
Parameters
X [array-like, shape (n_samples, n_features)] Test samples.
y [array-like, shape (n_samples) or (n_samples, n_outputs)] True values for X.
sample_weight [array-like, shape [n_samples], optional] Sample weights.
Returns
Notes
R^2 is calculated by weighting all the targets equally using multioutput=’uniform_average’.

of a nested object.
Returns
self

Examples using sklearn.multioutput.MultiOutputRegressor
6.27.3 sklearn.multioutput.MultiOutputClassifier
class sklearn.multioutput.MultiOutputClassifier(estimator, n_jobs=None)

Multi target classification
This strategy consists of fitting one classifier per target. This is a simple strategy for extending classifiers that
do not natively support multi-target classification
Parameters
estimator [estimator object] An estimator object implementing fit, score and predict_proba.
n_jobs [int or None, optional (default=None)] The number of jobs to use for the computa-
tion. It does each target variable in y in parallel. None means 1 unless in a joblib.
details.
Attributes
estimators_ [list of n_output estimators] Estimators used for predictions.
Methods
fit(X, y[, sample_weight]) Fit the model to data.

partial_fit(X, y[, classes, sample_weight]) Incrementally fit the model to data.
predict(X) Predict multi-output variable using a model trained for
each target variable.
score(X, y) “Returns the mean accuracy on the given test data and
labels.
Fit the model to data. Fit a separate model for each output variable.
Parameters
y [(sparse) array-like, shape (n_samples, n_outputs)] Multi-output targets. An indicator ma-
trix turns on multilabel estimation.
weights.
Returns
self [object]

Parameters
Returns
Incrementally fit the model to data. Fit a separate model for each output variable.
Parameters
y [(sparse) array-like, shape (n_samples, n_outputs)] Multi-output targets.
classes [list of numpy arrays, shape (n_outputs)] Each array is unique classes for one output
in str/int Can be obtained by via [np.unique(y[:, i]) for i in range(y.
shape[1])], where y is the target matrix of the entire dataset. This argument is required
for the first call to partial_fit and can be omitted in the subsequent calls. Note that y doesn’t
need to contain all labels in classes.
weights.
Returns
self [object]
predict(X)
Predict multi-output variable using a model trained for each target variable.
Parameters
Returns
y [(sparse) array-like, shape (n_samples, n_outputs)] Multi-output targets predicted across
multiple predictors. Note: Separate models are generated for each predictor.
predict_proba(X)
Probability estimates. Returns prediction probabilities for each class of each output.
Parameters
X [array-like, shape (n_samples, n_features)] Data
Returns
p [array of shape = [n_samples, n_classes], or a list of n_outputs such arrays if n_outputs >
1.] The class probabilities of the input samples. The order of the classes corresponds to
score(X, y)
“Returns the mean accuracy on the given test data and labels.
Parameters

X [array-like, shape [n_samples, n_features]] Test samples

y [array-like, shape [n_samples, n_outputs]] True values for X
Returns
scores [float] accuracy_score of self.predict(X) versus y
of a nested object.
Returns
self
6.27.4 sklearn.multioutput.RegressorChain
class sklearn.multioutput.RegressorChain(base_estimator, order=None, cv=None, ran-

dom_state=None)
A multi-label model that arranges regressions into a chain.
Each model makes a prediction in the order specified by the chain using all of the available features provided to
the model plus the predictions of models that are earlier in the chain.
Parameters
base_estimator [estimator] The base estimator from which the classifier chain is built.
order [array-like, shape=[n_outputs] or ‘random’, optional] By default the order will be deter-
mined by the order of columns in the label matrix Y.:
order = [0, 1, 2, ..., Y.shape[1] - 1]
The order of the chain can be explicitly set by providing a list of integers. For example, for
a chain of length 5.:
order = [1, 3, 2, 4, 0]
means that the first model in the chain will make predictions for column 1 in the Y matrix,
the second model will make predictions for column 3, etc.
If order is ‘random’ a random ordering will be used.
cv [int, cross-validation generator or an iterable, optional (default=None)] Determines whether
to use cross validated predictions or true labels for the results of previous estimators in the
chain. If cv is None the true labels are used when fitting. Otherwise possible inputs for cv
are:
• CV splitter,

The random number generator is used to generate random chain orders.
Attributes
estimators_ [list] A list of clones of base_estimator.
order_ [list] The order of labels in the classifier chain.
See also:
ClassifierChain Equivalent for classification

MultioutputRegressor Learns each output independently rather than chaining.
Methods
fit(X, Y) Fit the model to data matrix X and targets Y.

predict(X) Predict on the data matrix X using the ClassifierChain
model.
diction.
__init__(base_estimator, order=None, cv=None, random_state=None)

fit(X, Y)
Fit the model to data matrix X and targets Y.
Parameters
Y [array-like, shape (n_samples, n_classes)] The target values.
Returns
self [object]
Parameters
Returns
predict(X)
Predict on the data matrix X using the ClassifierChain model.
Parameters
Returns

Y_pred [array-like, shape (n_samples, n_classes)] The predicted values.

Parameters
Returns
of a nested object.
Returns
self
6.28 sklearn.naive_bayes: Naive Bayes
The sklearn.naive_bayes module implements Naive Bayes algorithms. These are supervised learning methods
based on applying Bayes’ theorem with strong (naive) feature independence assumptions.
User guide: See the Naive Bayes section for further details.
naive_bayes.BernoulliNB([alpha, binarize, . . . ]) Naive Bayes classifier for multivariate Bernoulli models.

naive_bayes.GaussianNB([priors, var_smoothing]) Gaussian Naive Bayes (GaussianNB)
naive_bayes.MultinomialNB([alpha, . . . ]) Naive Bayes classifier for multinomial models
naive_bayes.ComplementNB([alpha, fit_prior, . . . ]) The Complement Naive Bayes classifier described in Ren-
nie et al.
6.28.1 sklearn.naive_bayes.BernoulliNB
class sklearn.naive_bayes.BernoulliNB(alpha=1.0, binarize=0.0, fit_prior=True,

class_prior=None)
Naive Bayes classifier for multivariate Bernoulli models.
Like MultinomialNB, this classifier is suitable for discrete data. The difference is that while MultinomialNB
works with occurrence counts, BernoulliNB is designed for binary/boolean features.

Parameters
alpha [float, optional (default=1.0)] Additive (Laplace/Lidstone) smoothing parameter (0 for
no smoothing).
binarize [float or None, optional (default=0.0)] Threshold for binarizing (mapping to booleans)
of sample features. If None, input is presumed to already consist of binary vectors.
fit_prior [boolean, optional (default=True)] Whether to learn class prior probabilities or not. If
false, a uniform prior will be used.
class_prior [array-like, size=[n_classes,], optional (default=None)] Prior probabilities of the
classes. If specified the priors are not adjusted according to the data.
Attributes
class_log_prior_ [array, shape = [n_classes]] Log probability of each class (smoothed).
feature_log_prob_ [array, shape = [n_classes, n_features]] Empirical log probability of fea-
tures given a class, P(x_i|y).
class_count_ [array, shape = [n_classes]] Number of samples encountered for each class during
fitting. This value is weighted by the sample weight when provided.
feature_count_ [array, shape = [n_classes, n_features]] Number of samples encountered for
each (class, feature) during fitting. This value is weighted by the sample weight when
provided.
References
C.D. Manning, P. Raghavan and H. Schuetze (2008). Introduction to Information Retrieval. Cambridge Univer-
sity Press, pp. 234-265. http://nlp.stanford.edu/IR-book/html/htmledition/the-bernoulli-model-1.html
A. McCallum and K. Nigam (1998). A comparison of event models for naive Bayes text classification. Proc.
AAAI/ICML-98 Workshop on Learning for Text Categorization, pp. 41-48.
V. Metsis, I. Androutsopoulos and G. Paliouras (2006). Spam filtering with naive Bayes – Which naive Bayes?
3rd Conf. on Email and Anti-Spam (CEAS).
Examples

>>> X = np.random.randint(2, size=(6, 100))
>>> Y = np.array([1, 2, 3, 4, 4, 5])
>>> from sklearn.naive_bayes import BernoulliNB
>>> clf = BernoulliNB()
>>> clf.fit(X, Y)
BernoulliNB(alpha=1.0, binarize=0.0, class_prior=None, fit_prior=True)
>>> print(clf.predict(X[2:3]))
[3]
Methods
fit(X, y[, sample_weight]) Fit Naive Bayes classifier according to X, y

6.28. sklearn.naive_bayes: Naive Bayes 2035


partial_fit(X, y[, classes, sample_weight]) Incremental fit on a batch of samples.
predict_log_proba(X) Return log-probability estimates for the test vector X.
labels.
__init__(alpha=1.0, binarize=0.0, fit_prior=True, class_prior=None)

Fit Naive Bayes classifier according to X, y
Parameters
sample_weight [array-like, shape = [n_samples], (default=None)] Weights applied to indi-
vidual samples (1. for unweighted).
Returns
self [object]
Parameters
Returns
Incremental fit on a batch of samples.
This method is expected to be called several times consecutively on different chunks of a dataset so as to
implement out-of-core or online learning.
This is especially useful when the whole dataset is too big to fit in memory at once.
This method has some performance overhead hence it is better to call partial_fit on chunks of data that are
as large as possible (as long as fitting in the memory budget) to hide the overhead.
Parameters
classes [array-like, shape = [n_classes] (default=None)] List of all the classes that can pos-
sibly appear in the y vector.
Must be provided at the first call to partial_fit, can be omitted in subsequent calls.
sample_weight [array-like, shape = [n_samples] (default=None)] Weights applied to indi-

Returns
self [object]
predict(X)
Parameters
Returns
C [array, shape = [n_samples]] Predicted target values for X
Return log-probability estimates for the test vector X.
Parameters
Returns
C [array-like, shape = [n_samples, n_classes]] Returns the log-probability of the samples
for each class in the model. The columns correspond to the classes in sorted order, as they
predict_proba(X)
Parameters
Returns
C [array-like, shape = [n_samples, n_classes]] Returns the probability of the samples for
Parameters
Returns
of a nested object.
Returns

self
Examples using sklearn.naive_bayes.BernoulliNB

6.28.2 sklearn.naive_bayes.GaussianNB
class sklearn.naive_bayes.GaussianNB(priors=None, var_smoothing=1e-09)

Gaussian Naive Bayes (GaussianNB)
Can perform online updates to model parameters via partial_fit method. For details on algorithm used to update
feature means and variance online, see Stanford CS tech report STAN-CS-79-773 by Chan, Golub, and LeVeque:
http://i.stanford.edu/pub/cstr/reports/cs/tr/79/773/CS-TR-79-773.pdf
Parameters
priors [array-like, shape (n_classes,)] Prior probabilities of the classes. If specified the priors
are not adjusted according to the data.
var_smoothing [float, optional (default=1e-9)] Portion of the largest variance of all features
that is added to variances for calculation stability.
Attributes
class_prior_ [array, shape (n_classes,)] probability of each class.
class_count_ [array, shape (n_classes,)] number of training samples observed in each class.
theta_ [array, shape (n_classes, n_features)] mean of each feature per class
sigma_ [array, shape (n_classes, n_features)] variance of each feature per class
epsilon_ [float] absolute additive value to variances
Examples

>>> X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]])
>>> Y = np.array([1, 1, 1, 2, 2, 2])
>>> clf = GaussianNB()
>>> clf.fit(X, Y)
GaussianNB(priors=None, var_smoothing=1e-09)
[1]
>>> clf_pf = GaussianNB()
>>> clf_pf.partial_fit(X, Y, np.unique(Y))
GaussianNB(priors=None, var_smoothing=1e-09)
>>> print(clf_pf.predict([[-0.8, -1]]))
[1]

Methods
fit(X, y[, sample_weight]) Fit Gaussian Naive Bayes according to X, y

labels.
__init__(priors=None, var_smoothing=1e-09)
Fit Gaussian Naive Bayes according to X, y
Parameters
sample_weight [array-like, shape (n_samples,), optional (default=None)] Weights applied
to individual samples (1. for unweighted).
New in version 0.17: Gaussian Naive Bayes supports fitting with sample_weight.
Returns
self [object]
Parameters
Returns
This method has some performance and numerical stability overhead, hence it is better to call partial_fit on
chunks of data that are as large as possible (as long as fitting in the memory budget) to hide the overhead.
Parameters

classes [array-like, shape (n_classes,), optional (default=None)] List of all the classes that
can possibly appear in the y vector.
sample_weight [array-like, shape (n_samples,), optional (default=None)] Weights applied
to individual samples (1. for unweighted).
Returns
self [object]
predict(X)
Parameters
Returns
Parameters
Returns
predict_proba(X)
Parameters
Returns
Parameters
Returns

of a nested object.
Returns
self
Examples using sklearn.naive_bayes.GaussianNB

6.28.3 sklearn.naive_bayes.MultinomialNB
class sklearn.naive_bayes.MultinomialNB(alpha=1.0, fit_prior=True, class_prior=None)

Naive Bayes classifier for multinomial models
The multinomial Naive Bayes classifier is suitable for classification with discrete features (e.g., word counts for
text classification). The multinomial distribution normally requires integer feature counts. However, in practice,
fractional counts such as tf-idf may also work.
Parameters
no smoothing).
fit_prior [boolean, optional (default=True)] Whether to learn class prior probabilities or not. If
false, a uniform prior will be used.
class_prior [array-like, size (n_classes,), optional (default=None)] Prior probabilities of the
classes. If specified the priors are not adjusted according to the data.
Attributes
class_log_prior_ [array, shape (n_classes, )] Smoothed empirical log probability for each class.
intercept_ [array, shape (n_classes, )] Mirrors class_log_prior_ for interpreting Multi-
nomialNB as a linear model.
feature_log_prob_ [array, shape (n_classes, n_features)] Empirical log probability of features
given a class, P(x_i|y).
coef_ [array, shape (n_classes, n_features)] Mirrors feature_log_prob_ for interpreting
MultinomialNB as a linear model.

class_count_ [array, shape (n_classes,)] Number of samples encountered for each class during
feature_count_ [array, shape (n_classes, n_features)] Number of samples encountered for each
(class, feature) during fitting. This value is weighted by the sample weight when provided.
Notes
For the rationale behind the names coef_ and intercept_, i.e. naive Bayes as a linear classifier, see J. Rennie et
al. (2003), Tackling the poor assumptions of naive Bayes text classifiers, ICML.
References
C.D. Manning, P. Raghavan and H. Schuetze (2008). Introduction to Information Retrieval. Cambridge Uni-
versity Press, pp. 234-265. http://nlp.stanford.edu/IR-book/html/htmledition/naive-bayes-text-classification-1.
html
Examples

>>> y = np.array([1, 2, 3, 4, 5, 6])
>>> clf = MultinomialNB()
>>> clf.fit(X, y)
[3]
Methods

labels.
__init__(alpha=1.0, fit_prior=True, class_prior=None)

Parameters


Returns
self [object]
Parameters
Returns
Parameters
Returns
self [object]
predict(X)
Parameters
Returns
Parameters

Returns
predict_proba(X)
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.naive_bayes.MultinomialNB

6.28.4 sklearn.naive_bayes.ComplementNB
class sklearn.naive_bayes.ComplementNB(alpha=1.0, fit_prior=True, class_prior=None,

norm=False)
The Complement Naive Bayes classifier described in Rennie et al. (2003).
The Complement Naive Bayes classifier was designed to correct the “severe assumptions” made by the standard
Multinomial Naive Bayes classifier. It is particularly suited for imbalanced data sets.


Parameters
no smoothing).
fit_prior [boolean, optional (default=True)] Only used in edge case with a single class in the
training set.
class_prior [array-like, size (n_classes,), optional (default=None)] Prior probabilities of the
classes. Not used.
norm [boolean, optional (default=False)] Whether or not a second normalization of the weights
is performed. The default behavior mirrors the implementations found in Mahout and Weka,
which do not follow the full algorithm described in Table 9 of the paper.
Attributes
class_log_prior_ [array, shape (n_classes, )] Smoothed empirical log probability for each class.
Only used in edge case with a single class in the training set.
feature_log_prob_ [array, shape (n_classes, n_features)] Empirical weights for class comple-
ments.
class_count_ [array, shape (n_classes,)] Number of samples encountered for each class during
feature_count_ [array, shape (n_classes, n_features)] Number of samples encountered for each
(class, feature) during fitting. This value is weighted by the sample weight when provided.
feature_all_ [array, shape (n_features,)] Number of samples encountered for each feature dur-
ing fitting. This value is weighted by the sample weight when provided.
References
Rennie, J. D., Shih, L., Teevan, J., & Karger, D. R. (2003). Tackling the poor assumptions of naive bayes text
classifiers. In ICML (Vol. 3, pp. 616-623). https://people.csail.mit.edu/jrennie/papers/icml03-nb.pdf
Examples

>>> y = np.array([1, 2, 3, 4, 5, 6])
>>> from sklearn.naive_bayes import ComplementNB
>>> clf = ComplementNB()
>>> clf.fit(X, y)
ComplementNB(alpha=1.0, class_prior=None, fit_prior=True, norm=False)
[3]
Methods



labels.
__init__(alpha=1.0, fit_prior=True, class_prior=None, norm=False)

Parameters
Returns
self [object]
Parameters
Returns
Parameters

Returns
self [object]
predict(X)
Parameters
Returns
Parameters
Returns
predict_proba(X)
Parameters
Returns
Parameters
Returns
of a nested object.
Returns

self
Examples using sklearn.naive_bayes.ComplementNB
6.29 sklearn.neighbors: Nearest Neighbors
The sklearn.neighbors module implements the k-nearest neighbors algorithm.

User guide: See the Nearest Neighbors section for further details.
neighbors.BallTree BallTree for fast generalized N-point problems

neighbors.DistanceMetric DistanceMetric class
neighbors.KDTree KDTree for fast generalized N-point problems
neighbors.KernelDensity([bandwidth, . . . ]) Kernel Density Estimation
neighbors.KNeighborsClassifier([. . . ]) Classifier implementing the k-nearest neighbors vote.
neighbors.KNeighborsRegressor([n_neighbors, Regression based on k-nearest neighbors.
. . . ])
neighbors.LocalOutlierFactor([n_neighbors, Unsupervised Outlier Detection using Local Outlier Factor
. . . ]) (LOF)
neighbors.RadiusNeighborsClassifier([. . . ])
Classifier implementing a vote among neighbors within a
given radius
neighbors.RadiusNeighborsRegressor([radius, Regression based on neighbors within a fixed radius.
. . . ])
neighbors.NearestCentroid([metric, . . . ]) Nearest centroid classifier.
neighbors.NearestNeighbors([n_neighbors, Unsupervised learner for implementing neighbor searches.
. . . ])
6.29.1 sklearn.neighbors.BallTree
class sklearn.neighbors.BallTree
BallTree for fast generalized N-point problems
BallTree(X, leaf_size=40, metric=’minkowski’, **kwargs)
Parameters
X [array-like, shape = [n_samples, n_features]] n_samples is the number of points in the data
set, and n_features is the dimension of the parameter space. Note: if X is a C-contiguous
array of doubles then data will not be copied. Otherwise, an internal copy will be made.
leaf_size [positive integer (default = 40)] Number of points at which to switch to brute-force.
Changing leaf_size will not affect the results of a query, but can significantly impact the
speed of a query and the memory required to store the constructed tree. The amount of
memory needed to store the tree scales as approximately n_samples / leaf_size. For a spec-
ified leaf_size, a leaf node is guaranteed to satisfy leaf_size <= n_points <=
2 * leaf_size, except in the case that n_samples < leaf_size.
metric [string or DistanceMetric object] the distance metric to use for the tree. De-
fault=’minkowski’ with p=2 (that is, a euclidean metric). See the documentation of the
DistanceMetric class for a list of available metrics. ball_tree.valid_metrics gives a list of the
metrics which are valid for BallTree.

Additional keywords are passed to the distance metric class.

Attributes
data [memory view] The training data
Examples
Query for k-nearest neighbors

>>> X = np.random.random((10, 3)) # 10 points in 3 dimensions
>>> tree = BallTree(X, leaf_size=2)
>>> dist, ind = tree.query(X[:1], k=3)
>>> print(ind) # indices of 3 closest neighbors
[0 3 1]
>>> print(dist) # distances to 3 closest neighbors
[ 0. 0.19662693 0.29473397]
Pickle and Unpickle a tree. Note that the state of the tree is saved in the pickle operation: the tree needs not be
rebuilt upon unpickling.
>>> import pickle
>>> s = pickle.dumps(tree)
>>> tree_copy = pickle.loads(s)
>>> dist, ind = tree_copy.query(X[:1], k=3)
[0 3 1]
[ 0. 0.19662693 0.29473397]
Query for neighbors within a given radius

>>> print(tree.query_radius(X[:1], r=0.3, count_only=True))
3
>>> ind = tree.query_radius(X[:1], r=0.3)
>>> print(ind) # indices of neighbors within distance 0.3
[3 0 1]
Compute a gaussian kernel density estimate:

>>> X = np.random.random((100, 3))
>>> tree = BallTree(X)
>>> tree.kernel_density(X[:3], h=0.1, kernel='gaussian')
array([ 6.94114649, 7.83281226, 7.2071716 ])
Compute a two-point auto-correlation function
6.29. sklearn.neighbors: Nearest Neighbors 2049


>>> r = np.linspace(0, 1, 5)
>>> tree = BallTree(X)
>>> tree.two_point_correlation(X, r)
array([ 30, 62, 278, 580, 820])
Methods
kernel_density(self, X, h[, kernel, atol, . . . ]) Compute the kernel density estimate at points X with
the given kernel, using the distance metric specified at
tree creation.
query(X[, k, return_distance, dualtree, . . . ]) query the tree for the k nearest neighbors
query_radius query_radius(self, X, r, count_only = False):
two_point_correlation Compute the two-point correlation function
get_arrays
get_n_calls
get_tree_stats
reset_n_calls

kernel_density(self, X, h, kernel=’gaussian’, atol=0, rtol=1E-8, breadth_first=True, re-
turn_log=False)
Compute the kernel density estimate at points X with the given kernel, using the distance metric specified
at tree creation.
Parameters
X [array-like, shape = [n_samples, n_features]] An array of points to query. Last dimension
should match dimension of training data.
h [float] the bandwidth of the kernel
kernel [string] specify the kernel to use. Options are - ‘gaussian’ - ‘tophat’ - ‘epanechnikov’
- ‘exponential’ - ‘linear’ - ‘cosine’ Default is kernel = ‘gaussian’
atol, rtol [float (default = 0)] Specify the desired relative and absolute tolerance of the re-
sult. If the true result is K_true, then the returned result K_ret satisfies abs(K_true -
K_ret) < atol + rtol * K_ret The default is zero (i.e. machine precision) for
both.
breadth_first [boolean (default = False)] if True, use a breadth-first search. If False (default)
use a depth-first search. Breadth-first is generally faster for compact kernels and/or high
tolerances.
return_log [boolean (default = False)] return the logarithm of the result. This can be more
accurate than returning the result itself for narrow kernels.
Returns
density [ndarray] The array of (log)-density evaluations, shape = X.shape[:-1]

query(X, k=1, return_distance=True, dualtree=False, breadth_first=False)

query the tree for the k nearest neighbors
Parameters
X [array-like, shape = [n_samples, n_features]] An array of points to query
k [integer (default = 1)] The number of nearest neighbors to return
return_distance [boolean (default = True)] if True, return a tuple (d, i) of distances and
indices if False, return array i
dualtree [boolean (default = False)] if True, use the dual tree formalism for the query: a tree
is built for the query points, and the pair of trees is used to efficiently search this space.
This can lead to better performance as the number of points grows large.
breadth_first [boolean (default = False)] if True, then query the nodes in a breadth-first
manner. Otherwise, query the nodes in a depth-first manner.
sort_results [boolean (default = True)] if True, then distances and indices of each point are
sorted on return, so that the first column contains the closest points. Otherwise, neighbors
are returned in an arbitrary order.
Returns
i [if return_distance == False]
(d,i) [if return_distance == True]
d [array of doubles - shape: x.shape[:-1] + (k,)] each entry gives the list of distances to the
neighbors of the corresponding point
i [array of integers - shape: x.shape[:-1] + (k,)] each entry gives the list of indices of neigh-
bors of the corresponding point
query_radius()
query_radius(self, X, r, count_only = False):
query the tree for neighbors within a radius r
Parameters
r [distance within which neighbors are returned] r can be a single value, or an array of values
of shape x.shape[:-1] if different radii are desired for each point.
return_distance [boolean (default = False)] if True, return distances to neighbors of each
point if False, return only neighbors Note that unlike the query() method, setting re-
turn_distance=True here adds to the computation time. Not all distances need to be
calculated explicitly for return_distance=False. Results are not sorted by default: see
sort_results keyword.
count_only [boolean (default = False)] if True, return only the count of points within dis-
tance r if False, return the indices of all points within distance r If return_distance==True,
setting count_only=True will result in an error.
sort_results [boolean (default = False)] if True, the distances and indices will be sorted
before being returned. If False, the results will not be sorted. If return_distance == False,
setting sort_results = True will result in an error.
Returns
count [if count_only == True]

ind [if count_only == False and return_distance == False]

(ind, dist) [if count_only == False and return_distance == True]
count [array of integers, shape = X.shape[:-1]] each entry gives the number of neighbors
within a distance r of the corresponding point.
ind [array of objects, shape = X.shape[:-1]] each element is a numpy integer array listing
the indices of neighbors of the corresponding point. Note that unlike the results of a k-
neighbors query, the returned neighbors are not sorted by distance by default.
dist [array of objects, shape = X.shape[:-1]] each element is a numpy double array listing
the distances corresponding to indices in i.
two_point_correlation()
Compute the two-point correlation function
Parameters
r [array_like] A one-dimensional array of distances
dualtree [boolean (default = False)] If true, use a dualtree algorithm. Otherwise, use a
single-tree algorithm. Dual tree algorithms can have better scaling for large N.
Returns
counts [ndarray] counts[i] contains the number of pairs of points with distance less than or
equal to r[i]
6.29.2 sklearn.neighbors.DistanceMetric
class sklearn.neighbors.DistanceMetric
DistanceMetric class
This class provides a uniform interface to fast distance metric functions. The various metrics can be accessed
via the get_metric class method and the metric string identifier (see below). For example, to use the Euclidean
distance:
>>> dist = DistanceMetric.get_metric('euclidean')

>>> X = [[0, 1, 2],
[3, 4, 5]]
>>> dist.pairwise(X)
array([[ 0. , 5.19615242],
[ 5.19615242, 0. ]])
Available Metrics The following lists the string metric identifiers and the associated distance metric classes:
Metrics intended for real-valued vector spaces:

identifier class name args distance function

“euclidean” EuclideanDistance sqrt(sum((x -
• y)^2))
“manhattan” ManhattanDistance sum(|x - y|)

•
“chebyshev” ChebyshevDistance max(|x - y|)

•
“minkowski” MinkowskiDistance p sum(|x -

y|^p)^(1/p)
“wminkowski” WMinkowskiDistance p, w sum(|w * (x -
y)|^p)^(1/p)
“seuclidean” SEuclideanDistance V sqrt(sum((x -
y)^2 / V))
“mahalanobis” MahalanobisDistance V or VI sqrt((x - y)'
V^-1 (x - y))
Metrics intended for two-dimensional vector spaces: Note that the haversine distance metric requires data in
the form of [latitude, longitude] and both inputs and outputs are in units of radians.
identifier class name distance function

“haver- HaversineDis- 2 arcsin(sqrt(sin^2(0.5*dx) +
sine” tance cos(x1)cos(x2)sin^2(0.5*dy)))
Metrics intended for integer-valued vector spaces: Though intended for integer-valued vectors, these are also
valid metrics in the case of real-valued vectors.

“hamming” HammingDistance N_unequal(x, y) / N_tot
“canberra” CanberraDistance sum(|x - y| / (|x| + |y|))
“braycurtis” BrayCurtisDistance sum(|x - y|) / (sum(|x|) + sum(|y|))
Metrics intended for boolean-valued vector spaces: Any nonzero entry is evaluated to “True”. In the listings
below, the following abbreviations are used:
• N : number of dimensions
• NTT : number of dims in which both values are True
• NTF : number of dims in which the first value is True, second is False
• NFT : number of dims in which the first value is False, second is True
• NFF : number of dims in which both values are False
• NNEQ : number of non-equal dimensions, NNEQ = NTF + NFT
• NNZ : number of nonzero dimensions, NNZ = NTF + NFT + NTT


“jaccard” JaccardDistance NNEQ / NNZ
“matching” MatchingDistance NNEQ / N
“dice” DiceDistance NNEQ / (NTT + NNZ)
“kulsinski” KulsinskiDistance (NNEQ + N - NTT) / (NNEQ + N)
“rogerstanimoto” RogersTanimotoDistance 2 * NNEQ / (N + NNEQ)
“russellrao” RussellRaoDistance NNZ / N
“sokalmichener” SokalMichenerDistance 2 * NNEQ / (N + NNEQ)
“sokalsneath” SokalSneathDistance NNEQ / (NNEQ + 0.5 * NTT)
User-defined distance:
identifier class name args

“pyfunc” PyFuncDistance func
Here func is a function which takes two one-dimensional numpy arrays, and returns a distance. Note that
in order to be used within the BallTree, the distance must be a true metric: i.e. it must satisfy the following
properties
1. Non-negativity: d(x, y) >= 0
2. Identity: d(x, y) = 0 if and only if x == y
3. Symmetry: d(x, y) = d(y, x)
4. Triangle Inequality: d(x, y) + d(y, z) >= d(x, z)
Because of the Python object overhead involved in calling the python function, this will be fairly slow, but it
will have the same scaling as other distances.
Methods
dist_to_rdist Convert the true distance to the reduced distance.

get_metric Get the given distance metric from the string identifier.
pairwise Compute the pairwise distances between X and Y
rdist_to_dist Convert the Reduced distance to the true distance.

dist_to_rdist()
Convert the true distance to the reduced distance.
The reduced distance, defined for some metrics, is a computationally more efficient measure which pre-
serves the rank of the true distance. For example, in the Euclidean distance metric, the reduced distance is
the squared-euclidean distance.
get_metric()
Get the given distance metric from the string identifier.
See the docstring of DistanceMetric for a list of available metrics.
Parameters
metric [string or class name] The distance metric to use

**kwargs additional arguments will be passed to the requested metric

pairwise()
Compute the pairwise distances between X and Y
This is a convenience routine for the sake of testing. For many metrics, the utilities in
scipy.spatial.distance.cdist and scipy.spatial.distance.pdist will be faster.
Parameters
X [array_like] Array of shape (Nx, D), representing Nx points in D dimensions.
Y [array_like (optional)] Array of shape (Ny, D), representing Ny points in D dimensions.
If not specified, then Y=X.
Returns
——-
dist [ndarray] The shape (Nx, Ny) array of pairwise distances between points in X and Y.
rdist_to_dist()
Convert the Reduced distance to the true distance.
The reduced distance, defined for some metrics, is a computationally more efficient measure which pre-
serves the rank of the true distance. For example, in the Euclidean distance metric, the reduced distance is
the squared-euclidean distance.
6.29.3 sklearn.neighbors.KDTree
class sklearn.neighbors.KDTree
KDTree for fast generalized N-point problems
KDTree(X, leaf_size=40, metric=’minkowski’, **kwargs)
Parameters
X [array-like, shape = [n_samples, n_features]] n_samples is the number of points in the data
set, and n_features is the dimension of the parameter space. Note: if X is a C-contiguous
array of doubles then data will not be copied. Otherwise, an internal copy will be made.
leaf_size [positive integer (default = 40)] Number of points at which to switch to brute-force.
Changing leaf_size will not affect the results of a query, but can significantly impact the
speed of a query and the memory required to store the constructed tree. The amount of
memory needed to store the tree scales as approximately n_samples / leaf_size. For a spec-
ified leaf_size, a leaf node is guaranteed to satisfy leaf_size <= n_points <=
2 * leaf_size, except in the case that n_samples < leaf_size.
metric [string or DistanceMetric object] the distance metric to use for the tree. De-
fault=’minkowski’ with p=2 (that is, a euclidean metric). See the documentation of the
DistanceMetric class for a list of available metrics. kd_tree.valid_metrics gives a list of the
metrics which are valid for KDTree.
Additional keywords are passed to the distance metric class.
Attributes
data [memory view] The training data

Examples
Query for k-nearest neighbors

>>> tree = KDTree(X, leaf_size=2)
>>> dist, ind = tree.query(X[:1], k=3)
[0 3 1]
[ 0. 0.19662693 0.29473397]
Pickle and Unpickle a tree. Note that the state of the tree is saved in the pickle operation: the tree needs not be
rebuilt upon unpickling.
>>> import pickle
>>> s = pickle.dumps(tree)
>>> tree_copy = pickle.loads(s)
>>> dist, ind = tree_copy.query(X[:1], k=3)
[0 3 1]
[ 0. 0.19662693 0.29473397]
Query for neighbors within a given radius

>>> print(tree.query_radius(X[:1], r=0.3, count_only=True))
3
>>> ind = tree.query_radius(X[:1], r=0.3)
>>> print(ind) # indices of neighbors within distance 0.3
[3 0 1]
Compute a gaussian kernel density estimate:

>>> tree = KDTree(X)
>>> tree.kernel_density(X[:3], h=0.1, kernel='gaussian')
array([ 6.94114649, 7.83281226, 7.2071716 ])
Compute a two-point auto-correlation function

>>> r = np.linspace(0, 1, 5)
>>> tree = KDTree(X)

>>> tree.two_point_correlation(X, r)
array([ 30, 62, 278, 580, 820])
Methods
kernel_density(self, X, h[, kernel, atol, . . . ]) Compute the kernel density estimate at points X with
the given kernel, using the distance metric specified at
tree creation.
query(X[, k, return_distance, dualtree, . . . ]) query the tree for the k nearest neighbors
query_radius query_radius(self, X, r, count_only = False):
two_point_correlation Compute the two-point correlation function
get_arrays
get_n_calls
get_tree_stats
reset_n_calls

kernel_density(self, X, h, kernel=’gaussian’, atol=0, rtol=1E-8, breadth_first=True, re-
turn_log=False)
Compute the kernel density estimate at points X with the given kernel, using the distance metric specified
at tree creation.
Parameters
h [float] the bandwidth of the kernel
kernel [string] specify the kernel to use. Options are - ‘gaussian’ - ‘tophat’ - ‘epanechnikov’
- ‘exponential’ - ‘linear’ - ‘cosine’ Default is kernel = ‘gaussian’
atol, rtol [float (default = 0)] Specify the desired relative and absolute tolerance of the re-
sult. If the true result is K_true, then the returned result K_ret satisfies abs(K_true -
K_ret) < atol + rtol * K_ret The default is zero (i.e. machine precision) for
both.
breadth_first [boolean (default = False)] if True, use a breadth-first search. If False (default)
use a depth-first search. Breadth-first is generally faster for compact kernels and/or high
tolerances.
return_log [boolean (default = False)] return the logarithm of the result. This can be more
accurate than returning the result itself for narrow kernels.
Returns
density [ndarray] The array of (log)-density evaluations, shape = X.shape[:-1]
query(X, k=1, return_distance=True, dualtree=False, breadth_first=False)
query the tree for the k nearest neighbors
Parameters


k [integer (default = 1)] The number of nearest neighbors to return
return_distance [boolean (default = True)] if True, return a tuple (d, i) of distances and
indices if False, return array i
dualtree [boolean (default = False)] if True, use the dual tree formalism for the query: a tree
is built for the query points, and the pair of trees is used to efficiently search this space.
This can lead to better performance as the number of points grows large.
breadth_first [boolean (default = False)] if True, then query the nodes in a breadth-first
manner. Otherwise, query the nodes in a depth-first manner.
sort_results [boolean (default = True)] if True, then distances and indices of each point are
sorted on return, so that the first column contains the closest points. Otherwise, neighbors
are returned in an arbitrary order.
Returns
i [if return_distance == False]
(d,i) [if return_distance == True]
d [array of doubles - shape: x.shape[:-1] + (k,)] each entry gives the list of distances to the
neighbors of the corresponding point
i [array of integers - shape: x.shape[:-1] + (k,)] each entry gives the list of indices of neigh-
bors of the corresponding point
query_radius()
query_radius(self, X, r, count_only = False):
query the tree for neighbors within a radius r
Parameters
r [distance within which neighbors are returned] r can be a single value, or an array of values
of shape x.shape[:-1] if different radii are desired for each point.
return_distance [boolean (default = False)] if True, return distances to neighbors of each
point if False, return only neighbors Note that unlike the query() method, setting re-
turn_distance=True here adds to the computation time. Not all distances need to be
calculated explicitly for return_distance=False. Results are not sorted by default: see
sort_results keyword.
count_only [boolean (default = False)] if True, return only the count of points within dis-
tance r if False, return the indices of all points within distance r If return_distance==True,
setting count_only=True will result in an error.
sort_results [boolean (default = False)] if True, the distances and indices will be sorted
before being returned. If False, the results will not be sorted. If return_distance == False,
setting sort_results = True will result in an error.
Returns
count [if count_only == True]
ind [if count_only == False and return_distance == False]
(ind, dist) [if count_only == False and return_distance == True]

count [array of integers, shape = X.shape[:-1]] each entry gives the number of neighbors
within a distance r of the corresponding point.
ind [array of objects, shape = X.shape[:-1]] each element is a numpy integer array listing
the indices of neighbors of the corresponding point. Note that unlike the results of a k-
neighbors query, the returned neighbors are not sorted by distance by default.
dist [array of objects, shape = X.shape[:-1]] each element is a numpy double array listing
the distances corresponding to indices in i.
two_point_correlation()
Compute the two-point correlation function
Parameters
r [array_like] A one-dimensional array of distances
dualtree [boolean (default = False)] If true, use a dualtree algorithm. Otherwise, use a
single-tree algorithm. Dual tree algorithms can have better scaling for large N.
Returns
counts [ndarray] counts[i] contains the number of pairs of points with distance less than or
equal to r[i]
6.29.4 sklearn.neighbors.KernelDensity
class sklearn.neighbors.KernelDensity(bandwidth=1.0, algorithm=’auto’, kernel=’gaussian’,

metric=’euclidean’, atol=0, rtol=0, breadth_first=True,
leaf_size=40, metric_params=None)
Kernel Density Estimation
Parameters
bandwidth [float] The bandwidth of the kernel.
algorithm [string] The tree algorithm to use. Valid options are [‘kd_tree’|’ball_tree’|’auto’].
Default is ‘auto’.
kernel [string] The kernel to use. Valid kernels are [‘gaus-
sian’|’tophat’|’epanechnikov’|’exponential’|’linear’|’cosine’] Default is ‘gaussian’.
metric [string] The distance metric to use. Note that not all metrics are valid with all algorithms.
Refer to the documentation of BallTree and KDTree for a description of available algo-
rithms. Note that the normalization of the density output is correct only for the Euclidean
distance metric. Default is ‘euclidean’.
atol [float] The desired absolute tolerance of the result. A larger tolerance will generally lead
to faster execution. Default is 0.
rtol [float] The desired relative tolerance of the result. A larger tolerance will generally lead to
faster execution. Default is 1E-8.
breadth_first [boolean] If true (default), use a breadth-first approach to the problem. Otherwise
use a depth-first approach.
leaf_size [int] Specify the leaf size of the underlying tree. See BallTree or KDTree for
details. Default is 40.

metric_params [dict] Additional parameters to be passed to the tree for use with the metric.
For more information, see the documentation of BallTree or KDTree.
Methods
fit(X[, y, sample_weight]) Fit the Kernel Density model on the data.

sample([n_samples, random_state]) Generate random samples from the model.
score(X[, y]) Compute the total log probability density under the
model.
score_samples(X) Evaluate the density model on the data.
__init__(bandwidth=1.0, algorithm=’auto’, kernel=’gaussian’, metric=’euclidean’, atol=0, rtol=0,

breadth_first=True, leaf_size=40, metric_params=None)
Fit the Kernel Density model on the data.
Parameters
X [array_like, shape (n_samples, n_features)] List of n_features-dimensional data points.
sample_weight [array_like, shape (n_samples,), optional] List of sample weights attached
to the data X.
Parameters
Returns
sample(n_samples=1, random_state=None)
Generate random samples from the model.
Currently, this is implemented only for gaussian and tophat kernels.
Parameters
random_state [int, RandomState instance or None. default to None] If int, random_state is
Returns
X [array_like, shape (n_samples, n_features)] List of samples.
score(X, y=None)
Compute the total log probability density under the model.
Parameters

X [array_like, shape (n_samples, n_features)] List of n_features-dimensional data points.

Returns
logprob [float] Total log-likelihood of the data in X. This is normalized to be a probability
density, so the value will be low for high-dimensional data.
score_samples(X)
Evaluate the density model on the data.
Parameters
X [array_like, shape (n_samples, n_features)] An array of points to query. Last dimension
should match dimension of training data (n_features).
Returns
density [ndarray, shape (n_samples,)] The array of log(density) evaluations. These are nor-
malized to be probability densities, so values will be low for high-dimensional data.
of a nested object.
Returns
self
Examples using sklearn.neighbors.KernelDensity

• Kernel Density Estimate of Species Distributions
• Simple 1D Kernel Density Estimation
6.29.5 sklearn.neighbors.KNeighborsClassifier
class sklearn.neighbors.KNeighborsClassifier(n_neighbors=5, weights=’uniform’, al-

gorithm=’auto’, leaf_size=30, p=2, met-
ric=’minkowski’, metric_params=None,
n_jobs=None, **kwargs)
Classifier implementing the k-nearest neighbors vote.
Parameters
n_neighbors [int, optional (default = 5)] Number of neighbors to use by default for
kneighbors queries.
weights [str or callable, optional (default = ‘uniform’)] weight function used in prediction. Pos-
sible values:
• ‘uniform’ : uniform weights. All points in each neighborhood are weighted equally.
• ‘distance’ : weight points by the inverse of their distance. in this case, closer neighbors
of a query point will have a greater influence than neighbors which are further away.

• [callable] : a user-defined function which accepts an array of distances, and returns an

array of the same shape containing the weights.
algorithm [{‘auto’, ‘ball_tree’, ‘kd_tree’, ‘brute’}, optional] Algorithm used to compute the
nearest neighbors:
• ‘ball_tree’ will use BallTree
• ‘kd_tree’ will use KDTree
• ‘brute’ will use a brute-force search.
• ‘auto’ will attempt to decide the most appropriate algorithm based on the values passed
to fit method.
Note: fitting on sparse input will override the setting of this parameter, using brute force.
leaf_size [int, optional (default = 30)] Leaf size passed to BallTree or KDTree. This can affect
p [integer, optional (default = 2)] Power parameter for the Minkowski metric. When p = 1, this
is equivalent to using manhattan_distance (l1), and euclidean_distance (l2) for p = 2. For
arbitrary p, minkowski_distance (l_p) is used.
metric [string or callable, default ‘minkowski’] the distance metric to use for the tree. The
default metric is minkowski, and with p=2 is equivalent to the standard Euclidean metric.
See the documentation of the DistanceMetric class for a list of available metrics.
metric_params [dict, optional (default = None)] Additional keyword arguments for the metric
function.
using all processors. See Glossary for more details. Doesn’t affect fit method.
See also:
RadiusNeighborsClassifier, KNeighborsRegressor, RadiusNeighborsRegressor,
NearestNeighbors
Notes
See Nearest Neighbors in the online documentation for a discussion of the choice of algorithm and
leaf_size.
Warning: Regarding the Nearest Neighbors algorithms, if it is found that two neighbors, neighbor k+1 and
k, have identical distances but different labels, the results will depend on the ordering of the training data.
https://en.wikipedia.org/wiki/K-nearest_neighbor_algorithm
Examples
>>> X = [[0], [1], [2], [3]]

>>> y = [0, 0, 1, 1]
>>> neigh = KNeighborsClassifier(n_neighbors=3)

>>> neigh.fit(X, y)
KNeighborsClassifier(...)
>>> print(neigh.predict([[1.1]]))
[0]
>>> print(neigh.predict_proba([[0.9]]))
[[0.66666667 0.33333333]]
Methods
fit(X, y) Fit the model using X as training data and y as target

values
kneighbors([X, n_neighbors, return_distance]) Finds the K-neighbors of a point.
kneighbors_graph([X, n_neighbors, mode]) Computes the (weighted) graph of k-Neighbors for
points in X
predict(X) Predict the class labels for the provided data
predict_proba(X) Return probability estimates for the test data X.
labels.
__init__(n_neighbors=5, weights=’uniform’, algorithm=’auto’, leaf_size=30, p=2, met-

ric=’minkowski’, metric_params=None, n_jobs=None, **kwargs)
fit(X, y)
Fit the model using X as training data and y as target values
Parameters
X [{array-like, sparse matrix, BallTree, KDTree}] Training data. If array or matrix, shape
[n_samples, n_features], or [n_samples, n_samples] if metric=’precomputed’.
y [{array-like, sparse matrix}] Target values of shape = [n_samples] or [n_samples,
n_outputs]
Parameters
Returns
kneighbors(X=None, n_neighbors=None, return_distance=True)
Finds the K-neighbors of a point. Returns indices of and distances to the neighbors of each point.
Parameters
X [array-like, shape (n_query, n_features), or (n_query, n_indexed) if metric == ‘precom-
puted’] The query point or points. If not provided, neighbors of each indexed point are
returned. In this case, the query point is not considered its own neighbor.
n_neighbors [int] Number of neighbors to get (default is the value passed to the construc-
tor).

return_distance [boolean, optional. Defaults to True.] If False, distances will not be re-
turned
Returns
dist [array] Array representing the lengths to points, only present if return_distance=True
ind [array] Indices of the nearest points in the population matrix.
Examples
In the following example, we construct a NeighborsClassifier class from an array representing our data set
and ask who’s the closest point to [1,1,1]
>>> samples = [[0., 0., 0.], [0., .5, 0.], [1., 1., .5]]
>>> neigh = NearestNeighbors(n_neighbors=1)
>>> neigh.fit(samples)
NearestNeighbors(algorithm='auto', leaf_size=30, ...)
>>> print(neigh.kneighbors([[1., 1., 1.]]))
(array([[0.5]]), array([[2]]))
As you can see, it returns [[0.5]], and [[2]], which means that the element is at distance 0.5 and is the third
element of samples (indexes start at 0). You can also query for multiple points:
>>> X = [[0., 1., 0.], [1., 0., 1.]]

>>> neigh.kneighbors(X, return_distance=False)
array([[1],
[2]]...)
kneighbors_graph(X=None, n_neighbors=None, mode=’connectivity’)

Computes the (weighted) graph of k-Neighbors for points in X
Parameters
n_neighbors [int] Number of neighbors for each sample. (default is value passed to the
constructor).
mode [{‘connectivity’, ‘distance’}, optional] Type of returned matrix: ‘connectivity’ will
return the connectivity matrix with ones and zeros, in ‘distance’ the edges are Euclidean
distance between points.
Returns
A [sparse matrix in CSR format, shape = [n_samples, n_samples_fit]] n_samples_fit is the
number of samples in the fitted data A[i, j] is assigned the weight of edge that connects i
to j.
See also:
NearestNeighbors.radius_neighbors_graph

Examples
>>> X = [[0], [3], [1]]

>>> neigh.fit(X)
>>> A = neigh.kneighbors_graph(X)
>>> A.toarray()
array([[1., 0., 1.],
[0., 1., 1.],
[1., 0., 1.]])
predict(X)
Predict the class labels for the provided data
Parameters
puted’] Test samples.
Returns
y [array of shape [n_samples] or [n_samples, n_outputs]] Class labels for each data sample.
predict_proba(X)
Return probability estimates for the test data X.
Parameters
Returns
p [array of shape = [n_samples, n_classes], or a list of n_outputs] of such arrays if n_outputs
> 1. The class probabilities of the input samples. Classes are ordered by lexicographic
order.
Parameters
Returns
of a nested object.

Returns
self
Examples using sklearn.neighbors.KNeighborsClassifier
• Nearest Neighbors Classification
6.29.6 sklearn.neighbors.KNeighborsRegressor
class sklearn.neighbors.KNeighborsRegressor(n_neighbors=5, weights=’uniform’, algo-

rithm=’auto’, leaf_size=30, p=2, met-
Regression based on k-nearest neighbors.
The target is predicted by local interpolation of the targets associated of the nearest neighbors in the training set.
Parameters
kneighbors queries.
weights [str or callable] weight function used in prediction. Possible values:
Uniform weights are used by default.
nearest neighbors:
to fit method.

function.
using all processors. See Glossary for more details. Doesn’t affect fit method.
See also:
NearestNeighbors, RadiusNeighborsRegressor, KNeighborsClassifier,
RadiusNeighborsClassifier
Notes
leaf_size.
Warning: Regarding the Nearest Neighbors algorithms, if it is found that two neighbors, neighbor k+1 and
k, have identical distances but different labels, the results will depend on the ordering of the training data.
Examples
>>> X = [[0], [1], [2], [3]]

>>> y = [0, 0, 1, 1]
>>> from sklearn.neighbors import KNeighborsRegressor
>>> neigh = KNeighborsRegressor(n_neighbors=2)
>>> neigh.fit(X, y)
KNeighborsRegressor(...)
[0.5]
Methods

values
points in X
predict(X) Predict the target for the provided data


diction.
__init__(n_neighbors=5, weights=’uniform’, algorithm=’auto’, leaf_size=30, p=2, met-

ric=’minkowski’, metric_params=None, n_jobs=None, **kwargs)
fit(X, y)
Parameters
y [{array-like, sparse matrix}]
Target values, array of float values, shape = [n_samples] or [n_samples, n_outputs]
Parameters
Returns
Parameters
tor).
turned
Returns
Examples
>>> samples = [[0., 0., 0.], [0., .5, 0.], [1., 1., .5]]


(array([[0.5]]), array([[2]]))
>>> X = [[0., 1., 0.], [1., 0., 1.]]

array([[1],
[2]]...)

Parameters
constructor).
Returns
to j.
See also:
Examples
>>> X = [[0], [3], [1]]

>>> neigh.fit(X)
>>> A.toarray()
array([[1., 0., 1.],
[0., 1., 1.],
[1., 0., 1.]])
predict(X)
Predict the target for the provided data
Parameters

Returns
y [array of int, shape = [n_samples] or [n_samples, n_outputs]] Target values
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.neighbors.KNeighborsRegressor

• Nearest Neighbors regression
6.29.7 sklearn.neighbors.LocalOutlierFactor
class sklearn.neighbors.LocalOutlierFactor(n_neighbors=20, algorithm=’auto’,

leaf_size=30, metric=’minkowski’, p=2, met-
ric_params=None, contamination=’legacy’,
novelty=False, n_jobs=None)
Unsupervised Outlier Detection using Local Outlier Factor (LOF)
The anomaly score of each sample is called Local Outlier Factor. It measures the local deviation of density of
a given sample with respect to its neighbors. It is local in that the anomaly score depends on how isolated the
object is with respect to the surrounding neighborhood. More precisely, locality is given by k-nearest neighbors,
whose distance is used to estimate the local density. By comparing the local density of a sample to the local
densities of its neighbors, one can identify samples that have a substantially lower density than their neighbors.
These are considered outliers.
Parameters

n_neighbors [int, optional (default=20)] Number of neighbors to use by default for

kneighbors queries. If n_neighbors is larger than the number of samples provided, all
samples will be used.
nearest neighbors:
to fit method.
leaf_size [int, optional (default=30)] Leaf size passed to BallTree or KDTree. This can
affect the speed of the construction and query, as well as the memory required to store the
tree. The optimal value depends on the nature of the problem.
metric [string or callable, default ‘minkowski’] metric used for the distance computation. Any
If ‘precomputed’, the training input X is expected to be a distance matrix.
See the documentation for scipy.spatial.distance for details on these metrics: https://docs.
scipy.org/doc/scipy/reference/spatial.distance.html
p [integer, optional (default=2)] Parameter for the Minkowski metric from sklearn.
metrics.pairwise.pairwise_distances. When p = 1, this is equivalent to
using manhattan_distance (l1), and euclidean_distance (l2) for p = 2. For arbitrary p,
minkowski_distance (l_p) is used.
metric_params [dict, optional (default=None)] Additional keyword arguments for the metric
function.
data set, i.e. the proportion of outliers in the data set. When fitting this is used to define the
threshold on the decision function. If “auto”, the decision function threshold is determined
as in the original paper.
Changed in version 0.20: The default value of contamination will change from 0.1 in
0.20 to 'auto' in 0.22.
novelty [boolean, default False] By default, LocalOutlierFactor is only meant to be used for
outlier detection (novelty=False). Set novelty to True if you want to use LocalOutlierFactor
for novelty detection. In this case be aware that that you should only use predict, deci-
sion_function and score_samples on new unseen data and not on the training set.

n_jobs [int or None, optional (default=None)] The number of parallel jobs to run for neigh-
bors search. None means 1 unless in a joblib.parallel_backend context. -1
means using all processors. See Glossary for more details. Affects only kneighbors and
kneighbors_graph methods.
Attributes
negative_outlier_factor_ [numpy array, shape (n_samples,)] The opposite LOF of the train-
ing samples. The higher, the more normal. Inliers tend to have a LOF score close to 1
(negative_outlier_factor_ close to -1), while outliers tend to have a larger LOF
score.
The local outlier factor (LOF) of a sample captures its supposed ‘degree of abnormality’.
It is the average of the ratio of the local reachability density of a sample and those of its
k-nearest neighbors.
n_neighbors_ [integer] The actual number of neighbors used for kneighbors queries.
offset_ [float] Offset used to obtain binary labels from the raw scores. Observations having a
negative_outlier_factor smaller than offset_ are detected as abnormal. The offset is set to
-1.5 (inliers score around -1), except when a contamination parameter different than “auto”
is provided. In that case, the offset is defined in such a way we obtain the expected number
of outliers in training.
References
[1]
Methods
fit(X[, y]) Fit the model using X as training data.

points in X
__init__(n_neighbors=20, algorithm=’auto’, leaf_size=30, metric=’minkowski’, p=2, met-

ric_params=None, contamination=’legacy’, novelty=False, n_jobs=None)
decision_function
Shifted opposite of the Local Outlier Factor of X.
Bigger is better, i.e. large values correspond to inliers.
The shift offset allows a zero threshold for being an outlier. Only available for novelty detection (when
novelty is set to True). The argument X is supposed to contain new data: if X contains a point from
training, it considers the later in its own neighborhood. Also, the samples in X are not considered in the
neighborhood of any point.
Parameters
X [array-like, shape (n_samples, n_features)] The query sample or samples to compute the
Local Outlier Factor w.r.t. the training samples.
Returns

shifted_opposite_lof_scores [array, shape (n_samples,)] The shifted opposite of the Local

Outlier Factor of each input samples. The lower, the more abnormal. Negative scores
represent outliers, positive scores represent inliers.
fit(X, y=None)
Fit the model using X as training data.
Parameters
Returns
self [object]
fit_predict
“Fits the model to the training set X and returns the labels.
Label is 1 for an inlier and -1 for an outlier according to the LOF score and the contamination parameter.
Parameters
X [array-like, shape (n_samples, n_features), default=None] The query sample or samples
to compute the Local Outlier Factor w.r.t. to the training samples.
Returns
is_inlier [array, shape (n_samples,)] Returns -1 for anomalies/outliers and 1 for inliers.
Parameters
Returns
Parameters
tor).
turned
Returns

Examples
>>> samples = [[0., 0., 0.], [0., .5, 0.], [1., 1., .5]]
(array([[0.5]]), array([[2]]))
>>> X = [[0., 1., 0.], [1., 0., 1.]]

array([[1],
[2]]...)

Parameters
constructor).
Returns
to j.
See also:
Examples
>>> X = [[0], [3], [1]]

>>> neigh.fit(X)
>>> A.toarray()
array([[1., 0., 1.],

[0., 1., 1.],

[1., 0., 1.]])
predict
Predict the labels (1 inlier, -1 outlier) of X according to LOF.
This method allows to generalize prediction to new observations (not in the training set). Only available
for novelty detection (when novelty is set to True).
Parameters
Local Outlier Factor w.r.t. to the training samples.
Returns
is_inlier [array, shape (n_samples,)] Returns -1 for anomalies/outliers and +1 for inliers.
score_samples
Opposite of the Local Outlier Factor of X.
It is the opposite as as bigger is better, i.e. large values correspond to inliers.
Only available for novelty detection (when novelty is set to True). The argument X is supposed to contain
new data: if X contains a point from training, it considers the later in its own neighborhood. Also, the
samples in X are not considered in the neighborhood of any point. The score_samples on training data is
available by considering the the negative_outlier_factor_ attribute.
Parameters
Local Outlier Factor w.r.t. the training samples.
Returns
opposite_lof_scores [array, shape (n_samples,)] The opposite of the Local Outlier Factor of
each input samples. The lower, the more abnormal.
of a nested object.
Returns
self
Examples using sklearn.neighbors.LocalOutlierFactor

• Outlier detection with Local Outlier Factor (LOF)
• Novelty detection with Local Outlier Factor (LOF)

6.29.8 sklearn.neighbors.RadiusNeighborsClassifier
class sklearn.neighbors.RadiusNeighborsClassifier(radius=1.0, weights=’uniform’, algo-

ric=’minkowski’, outlier_label=None,
metric_params=None, n_jobs=None,
**kwargs)
Classifier implementing a vote among neighbors within a given radius
Parameters
radius [float, optional (default = 1.0)] Range of parameter space to use by default for
radius_neighbors queries.
nearest neighbors:
to fit method.
outlier_label [int, optional (default = None)] Label, which is given for outlier samples (samples
with no neighbors on given radius). If set to None, ValueError is raised, when outlier is
detected.
function.

See also:
KNeighborsClassifier, RadiusNeighborsRegressor, KNeighborsRegressor,
NearestNeighbors
Notes
leaf_size.
Examples
>>> X = [[0], [1], [2], [3]]

>>> y = [0, 0, 1, 1]
>>> from sklearn.neighbors import RadiusNeighborsClassifier
>>> neigh = RadiusNeighborsClassifier(radius=1.0)
>>> neigh.fit(X, y)
RadiusNeighborsClassifier(...)
[0]
Methods

values
predict(X) Predict the class labels for the provided data
radius_neighbors([X, radius, return_distance]) Finds the neighbors within a given radius of a point or
points.
radius_neighbors_graph([X, radius, mode]) Computes the (weighted) graph of Neighbors for points
in X
labels.
__init__(radius=1.0, weights=’uniform’, algorithm=’auto’, leaf_size=30, p=2, metric=’minkowski’,

outlier_label=None, metric_params=None, n_jobs=None, **kwargs)
fit(X, y)
Parameters
y [{array-like, sparse matrix}] Target values of shape = [n_samples] or [n_samples,
n_outputs]

Parameters
Returns
predict(X)
Predict the class labels for the provided data
Parameters
Returns
y [array of shape [n_samples] or [n_samples, n_outputs]] Class labels for each data sample.
radius_neighbors(X=None, radius=None, return_distance=True)
Finds the neighbors within a given radius of a point or points.
Return the indices and distances of each point from the dataset lying in a ball with size radius around
the points of the query array. Points lying on the boundary are included in the results.
The result points are not necessarily sorted by distance to their query point.
Parameters
X [array-like, (n_samples, n_features), optional] The query point or points. If not provided,
neighbors of each indexed point are returned. In this case, the query point is not considered
its own neighbor.
radius [float] Limiting distance of neighbors to return. (default is the value passed to the
constructor).
turned
Returns
dist [array, shape (n_samples,) of arrays] Array representing the distances to each point,
only present if return_distance=True. The distance values are computed according to the
metric constructor parameter.
ind [array, shape (n_samples,) of arrays] An array of arrays of indices of the approximate
nearest points from the population matrix that lie within a ball of size radius around the
query points.
Notes
Because the number of neighbors of each point is not necessarily equal, the results for multiple query
points cannot be fit in a standard data array. For efficiency, radius_neighbors returns arrays of objects,
where each object is a 1D array of indices or distances.
Examples
and ask who’s the closest point to [1, 1, 1]:


>>> samples = [[0., 0., 0.], [0., .5, 0.], [1., 1., .5]]
>>> neigh = NearestNeighbors(radius=1.6)
>>> rng = neigh.radius_neighbors([[1., 1., 1.]])
>>> print(np.asarray(rng[0][0]))
[1.5 0.5]
[1 2]
The first array returned contains the distances to all points which are closer than 1.6, while the second
array returned contains their indices. In general, multiple points can be queried at the same time.
radius_neighbors_graph(X=None, radius=None, mode=’connectivity’)
Computes the (weighted) graph of Neighbors for points in X
Neighborhoods are restricted the points at a distance lower than radius.
Parameters
X [array-like, shape = [n_samples, n_features], optional] The query point or points. If not
provided, neighbors of each indexed point are returned. In this case, the query point is not
considered its own neighbor.
radius [float] Radius of neighborhoods. (default is the value passed to the constructor).
Returns
A [sparse matrix in CSR format, shape = [n_samples, n_samples]] A[i, j] is assigned the
weight of edge that connects i to j.
See also:
kneighbors_graph
Examples
>>> X = [[0], [3], [1]]

>>> neigh.fit(X)
>>> A = neigh.radius_neighbors_graph(X)
>>> A.toarray()
array([[1., 0., 1.],
[0., 1., 0.],
[1., 0., 1.]])

Parameters
Returns
of a nested object.
Returns
self
6.29.9 sklearn.neighbors.RadiusNeighborsRegressor
class sklearn.neighbors.RadiusNeighborsRegressor(radius=1.0, weights=’uniform’, algo-

Regression based on neighbors within a fixed radius.
The target is predicted by local interpolation of the targets associated of the nearest neighbors in the training set.
Parameters
nearest neighbors:
to fit method.

function.
n_jobs [int or None, optional (default=None)]
The number of parallel jobs to run for neighbors search. None means 1 unless in a
joblib.parallel_backend context.
-1 means using all processors. See Glossary for more details.
See also:
NearestNeighbors, KNeighborsRegressor, KNeighborsClassifier,
RadiusNeighborsClassifier
Notes
leaf_size.
Examples
>>> X = [[0], [1], [2], [3]]

>>> y = [0, 0, 1, 1]
>>> from sklearn.neighbors import RadiusNeighborsRegressor
>>> neigh = RadiusNeighborsRegressor(radius=1.0)
>>> neigh.fit(X, y)
RadiusNeighborsRegressor(...)
[0.5]
Methods

values
predict(X) Predict the target for the provided data
points.


in X
diction.
__init__(radius=1.0, weights=’uniform’, algorithm=’auto’, leaf_size=30, p=2, metric=’minkowski’,

metric_params=None, n_jobs=None, **kwargs)
fit(X, y)
Parameters
y [{array-like, sparse matrix}]
Target values, array of float values, shape = [n_samples] or [n_samples, n_outputs]
Parameters
Returns
predict(X)
Predict the target for the provided data
Parameters
Returns
y [array of float, shape = [n_samples] or [n_samples, n_outputs]] Target values
Parameters
its own neighbor.
constructor).
turned

Returns
query points.
Notes
Examples

>>> samples = [[0., 0., 0.], [0., .5, 0.], [1., 1., .5]]
[1.5 0.5]
[1 2]
Parameters
Returns

See also:
kneighbors_graph
Examples
>>> X = [[0], [3], [1]]

>>> neigh.fit(X)
>>> A.toarray()
array([[1., 0., 1.],
[0., 1., 0.],
[1., 0., 1.]])
Parameters
Returns
of a nested object.
Returns
self
6.29.10 sklearn.neighbors.NearestCentroid
class sklearn.neighbors.NearestCentroid(metric=’euclidean’, shrink_threshold=None)

Nearest centroid classifier.
Each class is represented by its centroid, with test samples classified to the class with the nearest centroid.
Parameters

a feature array. If metric is a string or callable, it must be one of the options allowed by
metrics.pairwise.pairwise_distances for its metric parameter. The centroids for the samples
corresponding to each class is the point from which the sum of the distances (according to
the metric) of all samples that belong to that particular class are minimized. If the “manhat-
tan” metric is provided, this centroid is the median and for all other metrics, the centroid is
now set to be the mean.
shrink_threshold [float, optional (default = None)] Threshold for shrinking centroids to re-
move features.
Attributes
centroids_ [array-like, shape = [n_classes, n_features]] Centroid of each class
See also:
sklearn.neighbors.KNeighborsClassifier nearest neighbors classifier
Notes
When used for text classification with tf-idf vectors, this classifier is also known as the Rocchio classifier.
References
Tibshirani, R., Hastie, T., Narasimhan, B., & Chu, G. (2002). Diagnosis of multiple cancer types by shrunken
centroids of gene expression. Proceedings of the National Academy of Sciences of the United States of America,
99(10), 6567-6572. The National Academy of Sciences.
Examples
>>> from sklearn.neighbors.nearest_centroid import NearestCentroid

>>> X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]])
>>> y = np.array([1, 1, 1, 2, 2, 2])
>>> clf = NearestCentroid()
>>> clf.fit(X, y)
[1]
Methods
fit(X, y) Fit the NearestCentroid model according to the given

training data.
labels.

__init__(metric=’euclidean’, shrink_threshold=None)
fit(X, y)
Fit the NearestCentroid model according to the given training data.
Parameters
n_samples is the number of samples and n_features is the number of features. Note that
centroid shrinking cannot be used with sparse matrices.
y [array, shape = [n_samples]] Target values (integers)
Parameters
Returns
predict(X)
The predicted class C for each sample in X is returned.
Parameters
Returns
C [array, shape = [n_samples]]
Notes
If the metric constructor parameter is “precomputed”, X is assumed to be the distance matrix between the
data to be predicted and self.centroids_.
Parameters
Returns

of a nested object.
Returns
self
Examples using sklearn.neighbors.NearestCentroid
• Nearest Centroid Classification

6.29.11 sklearn.neighbors.NearestNeighbors
class sklearn.neighbors.NearestNeighbors(n_neighbors=5, radius=1.0, algorithm=’auto’,

leaf_size=30, metric=’minkowski’, p=2, met-
ric_params=None, n_jobs=None, **kwargs)
Unsupervised learner for implementing neighbor searches.
Parameters
kneighbors queries.
nearest neighbors:
to fit method.
metric [string or callable, default ‘minkowski’] metric to use for distance computation. Any


p [integer, optional (default = 2)] Parameter for the Minkowski metric from
sklearn.metrics.pairwise.pairwise_distances. When p = 1, this is equivalent to us-
ing manhattan_distance (l1), and euclidean_distance (l2) for p = 2. For arbitrary p,
function.
See also:
KNeighborsClassifier, RadiusNeighborsClassifier, KNeighborsRegressor,
RadiusNeighborsRegressor, BallTree
Notes
leaf_size.
Examples

>>> samples = [[0, 0, 2], [1, 0, 0], [0, 0, 1]]
>>> neigh = NearestNeighbors(2, 0.4)

NearestNeighbors(...)
>>> neigh.kneighbors([[0, 0, 1.3]], 2, return_distance=False)

...
array([[2, 0]]...)
>>> nbrs = neigh.radius_neighbors([[0, 0, 1.3]], 0.4, return_distance=False)

>>> np.asarray(nbrs[0][0])
array(2)
Methods
fit(X[, y]) Fit the model using X as training data



points in X
points.
in X
__init__(n_neighbors=5, radius=1.0, algorithm=’auto’, leaf_size=30, metric=’minkowski’, p=2,

metric_params=None, n_jobs=None, **kwargs)
fit(X, y=None)
Fit the model using X as training data
Parameters
Parameters
Returns
Parameters
tor).
turned
Returns
Examples

>>> samples = [[0., 0., 0.], [0., .5, 0.], [1., 1., .5]]
(array([[0.5]]), array([[2]]))
>>> X = [[0., 1., 0.], [1., 0., 1.]]

array([[1],
[2]]...)

Parameters
constructor).
Returns
to j.
See also:
Examples
>>> X = [[0], [3], [1]]

>>> neigh.fit(X)
>>> A.toarray()
array([[1., 0., 1.],
[0., 1., 1.],
[1., 0., 1.]])


Parameters
its own neighbor.
constructor).
turned
Returns
query points.
Notes
Examples
>>> samples = [[0., 0., 0.], [0., .5, 0.], [1., 1., .5]]
[1.5 0.5]
[1 2]
Parameters

Returns
See also:
kneighbors_graph
Examples
>>> X = [[0], [3], [1]]

>>> neigh.fit(X)
>>> A.toarray()
array([[1., 0., 1.],
[0., 1., 0.],
[1., 0., 1.]])
of a nested object.
Returns
self
neighbors.kneighbors_graph(X, n_neighbors[, Computes the (weighted) graph of k-Neighbors for points

. . . ]) in X
neighbors.radius_neighbors_graph(X, radius) Computes the (weighted) graph of Neighbors for points in
X
6.29.12 sklearn.neighbors.kneighbors_graph
sklearn.neighbors.kneighbors_graph(X, n_neighbors, mode=’connectivity’, met-

ric=’minkowski’, p=2, metric_params=None, in-
clude_self=False, n_jobs=None)
Parameters

X [array-like or BallTree, shape = [n_samples, n_features]] Sample data, in the form of a numpy
array or a precomputed BallTree.
n_neighbors [int] Number of neighbors for each sample.
mode [{‘connectivity’, ‘distance’}, optional] Type of returned matrix: ‘connectivity’ will return
the connectivity matrix with ones and zeros, and ‘distance’ will return the distances between
neighbors according to the given metric.
metric [string, default ‘minkowski’] The distance metric used to calculate the k-Neighbors for
each sample point. The DistanceMetric class gives a list of available metrics. The default
distance is ‘euclidean’ (‘minkowski’ metric with the p param equal to 2.)
p [int, default 2] Power parameter for the Minkowski metric. When p = 1, this is equivalent
to using manhattan_distance (l1), and euclidean_distance (l2) for p = 2. For arbitrary p,
metric_params [dict, optional] additional keyword arguments for the metric function.
include_self [bool, default=False.] Whether or not to mark each sample as the first near-
est neighbor to itself. If None, then True is used for mode=’connectivity’ and False for
mode=’distance’ as this will preserve backwards compatibility.
Returns
A [sparse matrix in CSR format, shape = [n_samples, n_samples]] A[i, j] is assigned the weight
of edge that connects i to j.
See also:
radius_neighbors_graph
Examples
>>> X = [[0], [3], [1]]

>>> from sklearn.neighbors import kneighbors_graph
>>> A = kneighbors_graph(X, 2, mode='connectivity', include_self=True)
>>> A.toarray()
array([[1., 0., 1.],
[0., 1., 1.],
[1., 0., 1.]])
Examples using sklearn.neighbors.kneighbors_graph


6.29.13 sklearn.neighbors.radius_neighbors_graph
sklearn.neighbors.radius_neighbors_graph(X, radius, mode=’connectivity’, met-

ric=’minkowski’, p=2, metric_params=None,
include_self=False, n_jobs=None)
Parameters
X [array-like or BallTree, shape = [n_samples, n_features]] Sample data, in the form of a numpy
array or a precomputed BallTree.
radius [float] Radius of neighborhoods.
mode [{‘connectivity’, ‘distance’}, optional] Type of returned matrix: ‘connectivity’ will return
the connectivity matrix with ones and zeros, and ‘distance’ will return the distances between
neighbors according to the given metric.
metric [string, default ‘minkowski’] The distance metric used to calculate the neighbors within
a given radius for each sample point. The DistanceMetric class gives a list of available
metrics. The default distance is ‘euclidean’ (‘minkowski’ metric with the param equal to 2.)
p [int, default 2] Power parameter for the Minkowski metric. When p = 1, this is equivalent
to using manhattan_distance (l1), and euclidean_distance (l2) for p = 2. For arbitrary p,
metric_params [dict, optional] additional keyword arguments for the metric function.
include_self [bool, default=False] Whether or not to mark each sample as the first nearest
neighbor to itself. If None, then True is used for mode=’connectivity’ and False for
mode=’distance’ as this will preserve backwards compatibility.
Returns
A [sparse matrix in CSR format, shape = [n_samples, n_samples]] A[i, j] is assigned the weight
of edge that connects i to j.
See also:
kneighbors_graph
Examples
>>> X = [[0], [3], [1]]

>>> from sklearn.neighbors import radius_neighbors_graph
>>> A = radius_neighbors_graph(X, 1.5, mode='connectivity',
... include_self=True)
>>> A.toarray()
array([[1., 0., 1.],
[0., 1., 0.],
[1., 0., 1.]])

6.30 sklearn.neural_network: Neural network models
The sklearn.neural_network module includes models based on neural networks.

User guide: See the Neural network models (supervised) and Neural network models (unsupervised) sections for
further details.
neural_network.BernoulliRBM ([n_components, Bernoulli Restricted Boltzmann Machine (RBM).

. . . ])
neural_network.MLPClassifier([. . . ]) Multi-layer Perceptron classifier.
neural_network.MLPRegressor([. . . ]) Multi-layer Perceptron regressor.
6.30.1 sklearn.neural_network.BernoulliRBM
class sklearn.neural_network.BernoulliRBM(n_components=256, learning_rate=0.1,

batch_size=10, n_iter=10, verbose=0, ran-
dom_state=None)
Bernoulli Restricted Boltzmann Machine (RBM).
A Restricted Boltzmann Machine with binary visible units and binary hidden units. Parameters are estimated
using Stochastic Maximum Likelihood (SML), also known as Persistent Contrastive Divergence (PCD) [2].
The time complexity of this implementation is O(d ** 2) assuming d ~ n_features ~ n_components.
Parameters
n_components [int, optional] Number of binary hidden units.
learning_rate [float, optional] The learning rate for weight updates. It is highly recommended
to tune this hyper-parameter. Reasonable values are in the 10**[0., -3.] range.
batch_size [int, optional] Number of examples per minibatch.
n_iter [int, optional] Number of iterations/sweeps over the training dataset to perform during
training.
verbose [int, optional] The verbosity level. The default, zero, means silent mode.
random_state [integer or RandomState, optional] A random number generator instance to de-
fine the state of the random permutations generator. If an integer is given, it fixes the seed.
Defaults to the global numpy random number generator.
Attributes
intercept_hidden_ [array-like, shape (n_components,)] Biases of the hidden units.
intercept_visible_ [array-like, shape (n_features,)] Biases of the visible units.
components_ [array-like, shape (n_components, n_features)] Weight matrix, where n_features
in the number of visible units and n_components is the number of hidden units.
References
[1] Hinton, G. E., Osindero, S. and Teh, Y. A fast learning algorithm for deep belief nets. Neural Compu-
tation 18, pp 1527-1554. https://www.cs.toronto.edu/~hinton/absps/fastnc.pdf
6.30. sklearn.neural_network: Neural network models 2095

[2] Tieleman, T. Training Restricted Boltzmann Machines using Approximations to the Likelihood Gradi-
ent. International Conference on Machine Learning (ICML) 2008
Examples

>>> from sklearn.neural_network import BernoulliRBM
>>> X = np.array([[0, 0, 0], [0, 1, 1], [1, 0, 1], [1, 1, 1]])
>>> model = BernoulliRBM(n_components=2)
>>> model.fit(X)
BernoulliRBM(batch_size=10, learning_rate=0.1, n_components=2, n_iter=10,
Methods
fit(X[, y]) Fit the model to the data X.

gibbs(v) Perform one Gibbs sampling step.
partial_fit(X[, y]) Fit the model to the data X which should contain a par-
tial segment of the data.
score_samples(X) Compute the pseudo-likelihood of X.
transform(X) Compute the hidden layer activation probabilities,
P(h=1|v=X).
__init__(n_components=256, learning_rate=0.1, batch_size=10, n_iter=10, verbose=0, ran-

dom_state=None)
fit(X, y=None)
Fit the model to the data X.
Parameters
X [{array-like, sparse matrix} shape (n_samples, n_features)] Training data.
Returns
self [BernoulliRBM] The fitted model.
Parameters
Returns

Parameters
Returns
gibbs(v)
Perform one Gibbs sampling step.
Parameters
v [array-like, shape (n_samples, n_features)] Values of the visible layer to start from.
Returns
v_new [array-like, shape (n_samples, n_features)] Values of the visible layer after one Gibbs
step.
Fit the model to the data X which should contain a partial segment of the data.
Parameters
Returns
self [BernoulliRBM] The fitted model.
score_samples(X)
Compute the pseudo-likelihood of X.
Parameters
X [{array-like, sparse matrix} shape (n_samples, n_features)] Values of the visible layer.
Must be all-boolean (not checked).
Returns
pseudo_likelihood [array-like, shape (n_samples,)] Value of the pseudo-likelihood (proxy
for likelihood).
Notes
This method is not deterministic: it computes a quantity called the free energy on X, then on a randomly
corrupted version of X, and returns the log of the logistic function of the difference.
of a nested object.
Returns
self
transform(X)
Compute the hidden layer activation probabilities, P(h=1|v=X).
Parameters

X [{array-like, sparse matrix} shape (n_samples, n_features)] The data to be transformed.

Returns
h [array, shape (n_samples, n_components)] Latent representations of the data.
Examples using sklearn.neural_network.BernoulliRBM
6.30.2 sklearn.neural_network.MLPClassifier
class sklearn.neural_network.MLPClassifier(hidden_layer_sizes=(100, ), activa-

tion=’relu’, solver=’adam’, alpha=0.0001,
batch_size=’auto’, learning_rate=’constant’,
learning_rate_init=0.001, power_t=0.5,
max_iter=200, shuffle=True, ran-
dom_state=None, tol=0.0001, ver-
bose=False, warm_start=False, momen-
tum=0.9, nesterovs_momentum=True,
beta_1=0.9, beta_2=0.999, epsilon=1e-08,
n_iter_no_change=10)
Multi-layer Perceptron classifier.
This model optimizes the log-loss function using LBFGS or stochastic gradient descent.
Parameters
hidden_layer_sizes [tuple, length = n_layers - 2, default (100,)] The ith element represents the
number of neurons in the ith hidden layer.
activation [{‘identity’, ‘logistic’, ‘tanh’, ‘relu’}, default ‘relu’] Activation function for the hid-
den layer.
• ‘identity’, no-op activation, useful to implement linear bottleneck, returns f(x) = x
• ‘logistic’, the logistic sigmoid function, returns f(x) = 1 / (1 + exp(-x)).
• ‘tanh’, the hyperbolic tan function, returns f(x) = tanh(x).
• ‘relu’, the rectified linear unit function, returns f(x) = max(0, x)
solver [{‘lbfgs’, ‘sgd’, ‘adam’}, default ‘adam’] The solver for weight optimization.
• ‘lbfgs’ is an optimizer in the family of quasi-Newton methods.
• ‘sgd’ refers to stochastic gradient descent.
• ‘adam’ refers to a stochastic gradient-based optimizer proposed by Kingma, Diederik,
and Jimmy Ba
Note: The default solver ‘adam’ works pretty well on relatively large datasets (with thou-
sands of training samples or more) in terms of both training time and validation score. For
small datasets, however, ‘lbfgs’ can converge faster and perform better.
alpha [float, optional, default 0.0001] L2 penalty (regularization term) parameter.

batch_size [int, optional, default ‘auto’] Size of minibatches for stochastic optimizers. If
the solver is ‘lbfgs’, the classifier will not use minibatch. When set to “auto”,
batch_size=min(200, n_samples)
learning_rate [{‘constant’, ‘invscaling’, ‘adaptive’}, default ‘constant’] Learning rate schedule
for weight updates.
• ‘constant’ is a constant learning rate given by ‘learning_rate_init’.
• ‘invscaling’ gradually decreases the learning rate at each time step ‘t’ using an in-
verse scaling exponent of ‘power_t’. effective_learning_rate = learning_rate_init / pow(t,
power_t)
• ‘adaptive’ keeps the learning rate constant to ‘learning_rate_init’ as long as training loss
keeps decreasing. Each time two consecutive epochs fail to decrease training loss by at
least tol, or fail to increase validation score by at least tol if ‘early_stopping’ is on, the
current learning rate is divided by 5.
Only used when solver='sgd'.
learning_rate_init [double, optional, default 0.001] The initial learning rate used. It controls
the step-size in updating the weights. Only used when solver=’sgd’ or ‘adam’.
power_t [double, optional, default 0.5] The exponent for inverse scaling learning rate. It is used
in updating effective learning rate when the learning_rate is set to ‘invscaling’. Only used
when solver=’sgd’.
max_iter [int, optional, default 200] Maximum number of iterations. The solver iterates until
convergence (determined by ‘tol’) or this number of iterations. For stochastic solvers (‘sgd’,
‘adam’), note that this determines the number of epochs (how many times each data point
will be used), not the number of gradient steps.
shuffle [bool, optional, default True] Whether to shuffle samples in each iteration. Only used
when solver=’sgd’ or ‘adam’.
tol [float, optional, default 1e-4] Tolerance for the optimization. When the loss or score is
not improving by at least tol for n_iter_no_change consecutive iterations, unless
learning_rate is set to ‘adaptive’, convergence is considered to be reached and training
stops.
verbose [bool, optional, default False] Whether to print progress messages to stdout.
warm_start [bool, optional, default False] When set to True, reuse the solution of the previous
call to fit as initialization, otherwise, just erase the previous solution. See the Glossary.
momentum [float, default 0.9] Momentum for gradient descent update. Should be between 0
and 1. Only used when solver=’sgd’.
nesterovs_momentum [boolean, default True] Whether to use Nesterov’s momentum. Only
used when solver=’sgd’ and momentum > 0.
early_stopping [bool, default False] Whether to use early stopping to terminate training when
validation score is not improving. If set to true, it will automatically set aside 10% of training
data as validation and terminate training when validation score is not improving by at least
tol for n_iter_no_change consecutive epochs. Only effective when solver=’sgd’ or
‘adam’

validation_fraction [float, optional, default 0.1] The proportion of training data to set aside as
validation set for early stopping. Must be between 0 and 1. Only used if early_stopping is
True
beta_1 [float, optional, default 0.9] Exponential decay rate for estimates of first moment vector
in adam, should be in [0, 1). Only used when solver=’adam’
beta_2 [float, optional, default 0.999] Exponential decay rate for estimates of second moment
vector in adam, should be in [0, 1). Only used when solver=’adam’
epsilon [float, optional, default 1e-8] Value for numerical stability in adam. Only used when
solver=’adam’
n_iter_no_change [int, optional, default 10] Maximum number of epochs to not meet tol
improvement. Only effective when solver=’sgd’ or ‘adam’
Attributes
classes_ [array or list of array of shape (n_classes,)] Class labels for each output.
loss_ [float] The current loss computed with the loss function.
coefs_ [list, length n_layers - 1] The ith element in the list represents the weight matrix corre-
sponding to layer i.
intercepts_ [list, length n_layers - 1] The ith element in the list represents the bias vector cor-
responding to layer i + 1.
n_iter_ [int,] The number of iterations the solver has ran.
n_layers_ [int] Number of layers.
n_outputs_ [int] Number of outputs.
out_activation_ [string] Name of the output activation function.
Notes
MLPClassifier trains iteratively since at each time step the partial derivatives of the loss function with respect to
the model parameters are computed to update the parameters.
It can also have a regularization term added to the loss function that shrinks model parameters to prevent over-
fitting.
This implementation works with data represented as dense numpy arrays or sparse scipy arrays of floating point
values.
References
Hinton, Geoffrey E. “Connectionist learning procedures.” Artificial intelligence 40.1 (1989): 185-234.
Glorot, Xavier, and Yoshua Bengio. “Understanding the difficulty of training deep feedforward neural net-
works.” International Conference on Artificial Intelligence and Statistics. 2010.
He, Kaiming, et al. “Delving deep into rectifiers: Surpassing human-level performance on imagenet classi-
fication.” arXiv preprint arXiv:1502.01852 (2015).
Kingma, Diederik, and Jimmy Ba. “Adam: A method for stochastic optimization.” arXiv preprint
arXiv:1412.6980 (2014).

Methods
fit(X, y) Fit the model to data matrix X and target(s) y.

predict(X) Predict using the multi-layer perceptron classifier
predict_log_proba(X) Return the log of probability estimates.
labels.
__init__(hidden_layer_sizes=(100, ), activation=’relu’, solver=’adam’, alpha=0.0001,

batch_size=’auto’, learning_rate=’constant’, learning_rate_init=0.001,
power_t=0.5, max_iter=200, shuffle=True, random_state=None, tol=0.0001, ver-
bose=False, warm_start=False, momentum=0.9, nesterovs_momentum=True,
early_stopping=False, validation_fraction=0.1, beta_1=0.9, beta_2=0.999, epsilon=1e-08,
fit(X, y)
Fit the model to data matrix X and target(s) y.
Parameters
X [array-like or sparse matrix, shape (n_samples, n_features)] The input data.
y [array-like, shape (n_samples,) or (n_samples, n_outputs)] The target values (class labels
Returns
self [returns a trained MLP model.]
Parameters
Returns
partial_fit
Update the model with a single iteration over the given data.
Parameters
y [array-like, shape (n_samples,)] The target values.
classes [array, shape (n_classes), default None] Classes across all calls to partial_fit. Can be
obtained via np.unique(y_all), where y_all is the target vector of the entire dataset. This
argument is required for the first call to partial_fit and can be omitted in the subsequent
calls. Note that y doesn’t need to contain all labels in classes.
Returns

predict(X)
Predict using the multi-layer perceptron classifier
Parameters
Returns
y [array-like, shape (n_samples,) or (n_samples, n_classes)] The predicted classes.
Return the log of probability estimates.
Parameters
X [array-like, shape (n_samples, n_features)] The input data.
Returns
log_y_prob [array-like, shape (n_samples, n_classes)] The predicted log-probability of the
sample for each class in the model, where classes are ordered as they are in self.classes_.
Equivalent to log(predict_proba(X))
predict_proba(X)
Parameters
Returns
y_prob [array-like, shape (n_samples, n_classes)] The predicted probability of the sample
for each class in the model, where classes are ordered as they are in self.classes_.
Parameters
Returns
of a nested object.
Returns
self

Examples using sklearn.neural_network.MLPClassifier
6.30.3 sklearn.neural_network.MLPRegressor
class sklearn.neural_network.MLPRegressor(hidden_layer_sizes=(100, ), activa-

tion=’relu’, solver=’adam’, alpha=0.0001,
batch_size=’auto’, learning_rate=’constant’,
learning_rate_init=0.001, power_t=0.5,
max_iter=200, shuffle=True, random_state=None,
tol=0.0001, verbose=False, warm_start=False,
momentum=0.9, nesterovs_momentum=True,
beta_1=0.9, beta_2=0.999, epsilon=1e-08,
Multi-layer Perceptron regressor.
This model optimizes the squared-loss using LBFGS or stochastic gradient descent.
Parameters
hidden_layer_sizes [tuple, length = n_layers - 2, default (100,)] The ith element represents the
number of neurons in the ith hidden layer.
activation [{‘identity’, ‘logistic’, ‘tanh’, ‘relu’}, default ‘relu’] Activation function for the hid-
den layer.
• ‘identity’, no-op activation, useful to implement linear bottleneck, returns f(x) = x
• ‘logistic’, the logistic sigmoid function, returns f(x) = 1 / (1 + exp(-x)).
• ‘tanh’, the hyperbolic tan function, returns f(x) = tanh(x).
• ‘relu’, the rectified linear unit function, returns f(x) = max(0, x)
solver [{‘lbfgs’, ‘sgd’, ‘adam’}, default ‘adam’] The solver for weight optimization.
• ‘lbfgs’ is an optimizer in the family of quasi-Newton methods.
• ‘sgd’ refers to stochastic gradient descent.
• ‘adam’ refers to a stochastic gradient-based optimizer proposed by Kingma, Diederik,
and Jimmy Ba
Note: The default solver ‘adam’ works pretty well on relatively large datasets (with thou-
sands of training samples or more) in terms of both training time and validation score. For
small datasets, however, ‘lbfgs’ can converge faster and perform better.
alpha [float, optional, default 0.0001] L2 penalty (regularization term) parameter.
batch_size [int, optional, default ‘auto’] Size of minibatches for stochastic optimizers. If
the solver is ‘lbfgs’, the classifier will not use minibatch. When set to “auto”,
batch_size=min(200, n_samples)

learning_rate [{‘constant’, ‘invscaling’, ‘adaptive’}, default ‘constant’] Learning rate schedule

for weight updates.
• ‘constant’ is a constant learning rate given by ‘learning_rate_init’.
• ‘invscaling’ gradually decreases the learning rate learning_rate_ at each time step
‘t’ using an inverse scaling exponent of ‘power_t’. effective_learning_rate = learn-
ing_rate_init / pow(t, power_t)
• ‘adaptive’ keeps the learning rate constant to ‘learning_rate_init’ as long as training loss
keeps decreasing. Each time two consecutive epochs fail to decrease training loss by at
least tol, or fail to increase validation score by at least tol if ‘early_stopping’ is on, the
current learning rate is divided by 5.
Only used when solver=’sgd’.
learning_rate_init [double, optional, default 0.001] The initial learning rate used. It controls
the step-size in updating the weights. Only used when solver=’sgd’ or ‘adam’.
power_t [double, optional, default 0.5] The exponent for inverse scaling learning rate. It is used
in updating effective learning rate when the learning_rate is set to ‘invscaling’. Only used
when solver=’sgd’.
max_iter [int, optional, default 200] Maximum number of iterations. The solver iterates until
convergence (determined by ‘tol’) or this number of iterations. For stochastic solvers (‘sgd’,
‘adam’), note that this determines the number of epochs (how many times each data point
will be used), not the number of gradient steps.
shuffle [bool, optional, default True] Whether to shuffle samples in each iteration. Only used
when solver=’sgd’ or ‘adam’.
tol [float, optional, default 1e-4] Tolerance for the optimization. When the loss or score is
not improving by at least tol for n_iter_no_change consecutive iterations, unless
learning_rate is set to ‘adaptive’, convergence is considered to be reached and training
stops.
verbose [bool, optional, default False] Whether to print progress messages to stdout.
warm_start [bool, optional, default False] When set to True, reuse the solution of the previous
call to fit as initialization, otherwise, just erase the previous solution. See the Glossary.
momentum [float, default 0.9] Momentum for gradient descent update. Should be between 0
and 1. Only used when solver=’sgd’.
nesterovs_momentum [boolean, default True] Whether to use Nesterov’s momentum. Only
used when solver=’sgd’ and momentum > 0.
early_stopping [bool, default False] Whether to use early stopping to terminate training when
validation score is not improving. If set to true, it will automatically set aside 10% of training
data as validation and terminate training when validation score is not improving by at least
tol for n_iter_no_change consecutive epochs. Only effective when solver=’sgd’ or
‘adam’
validation_fraction [float, optional, default 0.1] The proportion of training data to set aside as
validation set for early stopping. Must be between 0 and 1. Only used if early_stopping is
True

beta_1 [float, optional, default 0.9] Exponential decay rate for estimates of first moment vector
in adam, should be in [0, 1). Only used when solver=’adam’
beta_2 [float, optional, default 0.999] Exponential decay rate for estimates of second moment
vector in adam, should be in [0, 1). Only used when solver=’adam’
epsilon [float, optional, default 1e-8] Value for numerical stability in adam. Only used when
solver=’adam’
n_iter_no_change [int, optional, default 10] Maximum number of epochs to not meet tol
improvement. Only effective when solver=’sgd’ or ‘adam’
Attributes
loss_ [float] The current loss computed with the loss function.
coefs_ [list, length n_layers - 1] The ith element in the list represents the weight matrix corre-
sponding to layer i.
intercepts_ [list, length n_layers - 1] The ith element in the list represents the bias vector cor-
responding to layer i + 1.
n_iter_ [int,] The number of iterations the solver has ran.
n_layers_ [int] Number of layers.
n_outputs_ [int] Number of outputs.
out_activation_ [string] Name of the output activation function.
Notes
MLPRegressor trains iteratively since at each time step the partial derivatives of the loss function with respect
to the model parameters are computed to update the parameters.
It can also have a regularization term added to the loss function that shrinks model parameters to prevent over-
fitting.
This implementation works with data represented as dense and sparse numpy arrays of floating point values.
References
Hinton, Geoffrey E. “Connectionist learning procedures.” Artificial intelligence 40.1 (1989): 185-234.
Glorot, Xavier, and Yoshua Bengio. “Understanding the difficulty of training deep feedforward neural net-
works.” International Conference on Artificial Intelligence and Statistics. 2010.
He, Kaiming, et al. “Delving deep into rectifiers: Surpassing human-level performance on imagenet classi-
fication.” arXiv preprint arXiv:1502.01852 (2015).
Kingma, Diederik, and Jimmy Ba. “Adam: A method for stochastic optimization.” arXiv preprint
arXiv:1412.6980 (2014).
Methods

fit(X, y) Fit the model to data matrix X and target(s) y.

predict(X) Predict using the multi-layer perceptron model.
diction.
__init__(hidden_layer_sizes=(100, ), activation=’relu’, solver=’adam’, alpha=0.0001,

batch_size=’auto’, learning_rate=’constant’, learning_rate_init=0.001,
power_t=0.5, max_iter=200, shuffle=True, random_state=None, tol=0.0001, ver-
bose=False, warm_start=False, momentum=0.9, nesterovs_momentum=True,
early_stopping=False, validation_fraction=0.1, beta_1=0.9, beta_2=0.999, epsilon=1e-08,
fit(X, y)
Fit the model to data matrix X and target(s) y.
Parameters
X [array-like or sparse matrix, shape (n_samples, n_features)] The input data.
y [array-like, shape (n_samples,) or (n_samples, n_outputs)] The target values (class labels
Returns
Parameters
Returns
partial_fit
Update the model with a single iteration over the given data.
Parameters
y [array-like, shape (n_samples,)] The target values.
Returns
predict(X)
Predict using the multi-layer perceptron model.
Parameters
Returns
y [array-like, shape (n_samples, n_outputs)] The predicted values.

Parameters
Returns
of a nested object.
Returns
self
6.31 sklearn.pipeline: Pipeline
The sklearn.pipeline module implements utilities to build a composite estimator, as a chain of transforms and
estimators.
pipeline.FeatureUnion(transformer_list[, . . . ]) Concatenates results of multiple transformer objects.

pipeline.Pipeline(steps[, memory]) Pipeline of transforms with a final estimator.
6.31.1 sklearn.pipeline.FeatureUnion
class sklearn.pipeline.FeatureUnion(transformer_list, n_jobs=None, trans-

Concatenates results of multiple transformer objects.
This estimator applies a list of transformer objects in parallel to the input data, then concatenates the results.
This is useful to combine several feature extraction mechanisms into a single transformer.
Parameters of the transformers may be set using its name and the parameter name separated by a ‘__’. A
transformer may be replaced entirely by setting the parameter with its name to another transformer, or removed
by setting to ‘drop’ or None.
Parameters
6.31. sklearn.pipeline: Pipeline 2107

transformer_list [list of (string, transformer) tuples] List of transformer objects to be applied

to the data. The first half of each tuple is the name of the transformer.
transformer_weights [dict, optional] Multiplicative weights for features per transformer. Keys
are transformer names, values the weights.
See also:
sklearn.pipeline.make_union convenience function for simplified feature union construction.
Examples
>>> from sklearn.pipeline import FeatureUnion

>>> from sklearn.decomposition import PCA, TruncatedSVD
>>> union = FeatureUnion([("pca", PCA(n_components=1)),
... ("svd", TruncatedSVD(n_components=2))])
>>> X = [[0., 1., 3], [2., 2., 5]]
>>> union.fit_transform(X)
array([[ 1.5 , 3.0..., 0.8...],
[-1.5 , 5.7..., -0.4...]])
Methods
fit(X[, y]) Fit all transformers using X.

fit_transform(X[, y]) Fit all transformers, transform the data and concatenate
results.
get_feature_names() Get feature names from all transformers.
transform(X) Transform X separately by each transformer, concate-
nate results.
__init__(transformer_list, n_jobs=None, transformer_weights=None)

fit(X, y=None)
Fit all transformers using X.
Parameters
X [iterable or array-like, depending on transformers] Input data, used to fit transformers.
Returns
self [FeatureUnion] This estimator
Fit all transformers, transform the data and concatenate results.
Parameters

X [iterable or array-like, depending on transformers] Input data to be transformed.

Returns
transformers.
get_feature_names()
Get feature names from all transformers.
Returns
feature_names [list of strings] Names of the features produced by transform.
Parameters
Returns
Returns
self
transform(X)
Transform X separately by each transformer, concatenate results.
Parameters
X [iterable or array-like, depending on transformers] Input data to be transformed.
Returns
transformers.
Examples using sklearn.pipeline.FeatureUnion
6.31.2 sklearn.pipeline.Pipeline
class sklearn.pipeline.Pipeline(steps, memory=None)

Pipeline of transforms with a final estimator.
Sequentially apply a list of transforms and a final estimator. Intermediate steps of the pipeline must be ‘trans-
forms’, that is, they must implement fit and transform methods. The final estimator only needs to implement fit.
The transformers in the pipeline can be cached using memory argument.

The purpose of the pipeline is to assemble several steps that can be cross-validated together while setting differ-
ent parameters. For this, it enables setting parameters of the various steps using their names and the parameter
name separated by a ‘__’, as in the example below. A step’s estimator may be replaced entirely by setting the
parameter with its name to another estimator, or a transformer removed by setting to None.
Parameters
steps [list] List of (name, transform) tuples (implementing fit/transform) that are chained, in the
order in which they are chained, with the last object an estimator.
memory [None, str or object with the joblib.Memory interface, optional] Used to cache the fit-
ted transformers of the pipeline. By default, no caching is performed. If a string is given,
it is the path to the caching directory. Enabling caching triggers a clone of the transform-
ers before fitting. Therefore, the transformer instance given to the pipeline cannot be in-
spected directly. Use the attribute named_steps or steps to inspect estimators within
the pipeline. Caching the transformers is advantageous when fitting is time consuming.
Attributes
named_steps [bunch object, a dictionary with attribute access] Read-only attribute to access
any step parameter by user given name. Keys are step names and values are steps parameters.
See also:
sklearn.pipeline.make_pipeline convenience function for simplified pipeline construction.
Examples

>>> from sklearn.datasets import samples_generator
>>> from sklearn.feature_selection import SelectKBest
>>> from sklearn.feature_selection import f_regression
>>> # generate some data to play with
>>> X, y = samples_generator.make_classification(
... n_informative=5, n_redundant=0, random_state=42)
>>> # ANOVA SVM-C
>>> anova_filter = SelectKBest(f_regression, k=5)
>>> clf = svm.SVC(kernel='linear')
>>> anova_svm = Pipeline([('anova', anova_filter), ('svc', clf)])
>>> # You can set the parameters using the names issued
>>> # For instance, fit using a k of 10 in the SelectKBest
>>> # and a parameter 'C' of the svm
>>> anova_svm.set_params(anova__k=10, svc__C=.1).fit(X, y)
...
steps=[('anova', SelectKBest(...)),
('svc', SVC(...))])
>>> prediction = anova_svm.predict(X)
>>> anova_svm.score(X, y)
0.83
>>> # getting the selected features chosen by anova_filter
>>> anova_svm.named_steps['anova'].get_support()
...
array([False, False, True, True, False, False, True, True, False,
True, False, True, True, False, True, False, True, True,

False, False])
>>> # Another way to get selected features chosen by anova_filter
>>> anova_svm.named_steps.anova.get_support()
...
array([False, False, True, True, False, False, True, True, False,
True, False, True, True, False, True, False, True, True,
False, False])
Methods
decision_function(X) Apply transforms, and decision_function of the final es-

timator
fit(X[, y]) Fit the model
fit_predict(X[, y]) Applies fit_predict of last step in pipeline after trans-
forms.
fit_transform(X[, y]) Fit the model and transform with the final estimator
predict(X, **predict_params) Apply transforms to the data, and predict with the final
estimator
predict_log_proba(X) Apply transforms, and predict_log_proba of the final es-
timator
predict_proba(X) Apply transforms, and predict_proba of the final estima-
tor
score(X[, y, sample_weight]) Apply transforms, and score with the final estimator
__init__(steps, memory=None)
Apply transforms, and decision_function of the final estimator
Parameters
X [iterable] Data to predict on. Must fulfill input requirements of first step of the pipeline.
Returns
y_score [array-like, shape = [n_samples, n_classes]]
fit(X, y=None, **fit_params)
Fit the model
Fit all the transforms one after the other and transform the data, then fit the transformed data using the final
estimator.
Parameters
X [iterable] Training data. Must fulfill input requirements of first step of the pipeline.
y [iterable, default=None] Training targets. Must fulfill label requirements for all steps of
the pipeline.
**fit_params [dict of string -> object] Parameters passed to the fit method of each step,
where each parameter name is prefixed such that parameter p for step s has key s__p.
Returns

self [Pipeline] This estimator

fit_predict(X, y=None, **fit_params)
Applies fit_predict of last step in pipeline after transforms.
Applies fit_transforms of a pipeline to the data, followed by the fit_predict method of the final estimator in
the pipeline. Valid only if the final estimator implements fit_predict.
Parameters
the pipeline.
Returns
y_pred [array-like]
Fit the model and transform with the final estimator
Fits all the transforms one after the other and transforms the data, then uses fit_transform on transformed
data with the final estimator.
Parameters
the pipeline.
Returns
Xt [array-like, shape = [n_samples, n_transformed_features]] Transformed samples
Parameters
Returns
inverse_transform
Apply inverse transformations in reverse order
All estimators in the pipeline must support inverse_transform.
Parameters
Xt [array-like, shape = [n_samples, n_transformed_features]] Data samples, where
n_samples is the number of samples and n_features is the number of features. Must
fulfill input requirements of last step of pipeline’s inverse_transform method.
Returns

Xt [array-like, shape = [n_samples, n_features]]

predict(X, **predict_params)
Apply transforms to the data, and predict with the final estimator
Parameters
**predict_params [dict of string -> object] Parameters to the predict called at the end of
all transformations in the pipeline. Note that while this may be used to return uncertainties
from some models with return_std or return_cov, uncertainties that are generated by the
transformations in the pipeline are not propagated to the final estimator.
Returns
y_pred [array-like]
Apply transforms, and predict_log_proba of the final estimator
Parameters
Returns
y_score [array-like, shape = [n_samples, n_classes]]
predict_proba(X)
Apply transforms, and predict_proba of the final estimator
Parameters
Returns
y_proba [array-like, shape = [n_samples, n_classes]]
Apply transforms, and score with the final estimator
Parameters
y [iterable, default=None] Targets used for scoring. Must fulfill label requirements for all
steps of the pipeline.
sample_weight [array-like, default=None] If not None, this argument is passed as
sample_weight keyword argument to the score method of the final estimator.
Returns
score [float]
Returns
self

transform
Apply transforms, and transform with the final estimator
This also works where final estimator is None: all prior transformations are applied.
Parameters
X [iterable] Data to transform. Must fulfill input requirements of first step of the pipeline.
Returns
Xt [array-like, shape = [n_samples, n_transformed_features]]
Examples using sklearn.pipeline.Pipeline

pipeline.make_pipeline(*steps, **kwargs) Construct a Pipeline from the given estimators.

pipeline.make_union(*transformers, **kwargs) Construct a FeatureUnion from the given transformers.
6.31.3 sklearn.pipeline.make_pipeline
sklearn.pipeline.make_pipeline(*steps, **kwargs)
Construct a Pipeline from the given estimators.
This is a shorthand for the Pipeline constructor; it does not require, and does not permit, naming the estimators.
Instead, their names will be set to the lowercase of their types automatically.
Parameters
*steps [list of estimators.]
memory [None, str or object with the joblib.Memory interface, optional] Used to cache the fit-
ted transformers of the pipeline. By default, no caching is performed. If a string is given,
it is the path to the caching directory. Enabling caching triggers a clone of the transform-
ers before fitting. Therefore, the transformer instance given to the pipeline cannot be in-
spected directly. Use the attribute named_steps or steps to inspect estimators within
the pipeline. Caching the transformers is advantageous when fitting is time consuming.
Returns

p [Pipeline]
See also:
sklearn.pipeline.Pipeline Class for creating a pipeline of transforms with a final estimator.
Examples

>>> make_pipeline(StandardScaler(), GaussianNB(priors=None))
...
steps=[('standardscaler',
StandardScaler(copy=True, with_mean=True, with_std=True)),
('gaussiannb',
GaussianNB(priors=None, var_smoothing=1e-09))])
Examples using sklearn.pipeline.make_pipeline

6.31.4 sklearn.pipeline.make_union
sklearn.pipeline.make_union(*transformers, **kwargs)
Construct a FeatureUnion from the given transformers.
This is a shorthand for the FeatureUnion constructor; it does not require, and does not permit, naming the
transformers. Instead, they will be given names automatically based on their types. It also does not allow
weighting.
Parameters
*transformers [list of estimators]
Returns
f [FeatureUnion]

See also:
sklearn.pipeline.FeatureUnion Class for concatenating the results of multiple transformer objects.
Examples
>>> from sklearn.decomposition import PCA, TruncatedSVD

>>> from sklearn.pipeline import make_union
>>> make_union(PCA(), TruncatedSVD())
transformer_list=[('pca',
PCA(copy=True, iterated_power='auto',
n_components=None, random_state=None,
svd_solver='auto', tol=0.0, whiten=False)),
('truncatedsvd',
TruncatedSVD(algorithm='randomized',
n_components=2, n_iter=5,
random_state=None, tol=0.0))],
Examples using sklearn.pipeline.make_union
6.32 sklearn.preprocessing: Preprocessing and Normalization
The sklearn.preprocessing module includes scaling, centering, normalization, binarization and imputation
methods.
User guide: See the Preprocessing data section for further details.
preprocessing.Binarizer([threshold, copy]) Binarize data (set feature values to 0 or 1) according to a

threshold
preprocessing.FunctionTransformer([func, Constructs a transformer from an arbitrary callable.
. . . ])
preprocessing.KBinsDiscretizer([n_bins, Bin continuous data into intervals.
. . . ])
preprocessing.KernelCenterer() Center a kernel matrix
preprocessing.LabelBinarizer([neg_label, Binarize labels in a one-vs-all fashion
. . . ])
preprocessing.LabelEncoder Encode labels with value between 0 and n_classes-1.
preprocessing.MultiLabelBinarizer([classes, Transform between iterable of iterables and a multilabel
. . . ]) format
preprocessing.MaxAbsScaler([copy]) Scale each feature by its maximum absolute value.
preprocessing.MinMaxScaler([feature_range, Transforms features by scaling each feature to a given
copy]) range.
preprocessing.Normalizer([norm, copy]) Normalize samples individually to unit norm.
preprocessing.OneHotEncoder([n_values, . . . ]) Encode categorical integer features as a one-hot numeric
array.


preprocessing.OrdinalEncoder([categories, Encode categorical features as an integer array.
dtype])
preprocessing.PolynomialFeatures([degree, Generate polynomial and interaction features.
. . . ])
preprocessing.PowerTransformer([method, Apply a power transform featurewise to make data more
. . . ]) Gaussian-like.
preprocessing.QuantileTransformer([. . . ]) Transform features using quantiles information.
preprocessing.RobustScaler([with_centering, Scale features using statistics that are robust to outliers.
. . . ])
preprocessing.StandardScaler([copy, . . . ]) Standardize features by removing the mean and scaling to
unit variance
6.32.1 sklearn.preprocessing.Binarizer
class sklearn.preprocessing.Binarizer(threshold=0.0, copy=True)

Binarize data (set feature values to 0 or 1) according to a threshold
Values greater than the threshold map to 1, while values less than or equal to the threshold map to 0. With the
default threshold of 0, only positive values map to 1.
Binarization is a common operation on text count data where the analyst can decide to only consider the presence
or absence of a feature rather than a quantified number of occurrences for instance.
It can also be used as a pre-processing step for estimators that consider boolean random variables (e.g. modelled
using the Bernoulli distribution in a Bayesian setting).
Parameters
threshold [float, optional (0.0 by default)] Feature values below or equal to this are replaced by
0, above it by 1. Threshold may not be less than 0 for operations on sparse matrices.
copy [boolean, optional, default True] set to False to perform inplace binarization and avoid a
copy (if the input is already a numpy array or a scipy.sparse CSR matrix).
See also:
binarize Equivalent function without the estimator API.
Notes
If the input is a sparse matrix, only the non-zero values are subject to update by the Binarizer class.
This estimator is stateless (besides constructor parameters), the fit method does nothing but is useful when used
in a pipeline.
Examples
>>> from sklearn.preprocessing import Binarizer

>>> X = [[ 1., -1., 2.],
... [ 2., 0., 0.],
... [ 0., 1., -1.]]
>>> transformer = Binarizer().fit(X) # fit does nothing.
6.32. sklearn.preprocessing: Preprocessing and Normalization 2117

>>> transformer
Binarizer(copy=True, threshold=0.0)
array([[1., 0., 1.],
[1., 0., 0.],
[0., 1., 0.]])
Methods

transform(X[, y, copy]) Binarize each element of X
__init__(threshold=0.0, copy=True)
fit(X, y=None)
Parameters
X [array-like]
Parameters
Returns
Parameters
Returns
of a nested object.
Returns

self
transform(X, y=’deprecated’, copy=None)
Binarize each element of X
Parameters
X [{array-like, sparse matrix}, shape [n_samples, n_features]] The data to binarize, element
by element. scipy.sparse matrices should be in CSR format to avoid an un-necessary copy.
copy [bool] Copy the input X or not.
6.32.2 sklearn.preprocessing.FunctionTransformer
class sklearn.preprocessing.FunctionTransformer(func=None, inverse_func=None, val-

idate=None, accept_sparse=False,
pass_y=’deprecated’,
check_inverse=True, kw_args=None,
inv_kw_args=None)
Constructs a transformer from an arbitrary callable.
A FunctionTransformer forwards its X (and optionally y) arguments to a user-defined function or function
object and returns the result of this function. This is useful for stateless transformations such as taking the log
of frequencies, doing custom scaling, etc.
Note: If a lambda is used as the function, then the resulting transformer will not be pickleable.
Parameters
func [callable, optional default=None] The callable to use for the transformation. This will be
passed the same arguments as transform, with args and kwargs forwarded. If func is None,
then func will be the identity function.
inverse_func [callable, optional default=None] The callable to use for the inverse transforma-
tion. This will be passed the same arguments as inverse transform, with args and kwargs
forwarded. If inverse_func is None, then inverse_func will be the identity function.
validate [bool, optional default=True] Indicate that the input X array should be checked before
calling func. The possibilities are:
• If False, there is no input validation.
• If True, then X will be converted to a 2-dimensional NumPy array or sparse matrix. If the
conversion is not possible an exception is raised.
Deprecated since version 0.20: validate=True as default will be replaced by
validate=False in 0.22.
accept_sparse [boolean, optional] Indicate that func accepts a sparse matrix as input. If validate
is False, this has no effect. Otherwise, if accept_sparse is false, sparse matrix inputs will
cause an exception to be raised.
pass_y [bool, optional default=False] Indicate that transform should forward the y argument to
the inner callable.

check_inverse [bool, default=True] Whether to check that or func followed by

inverse_func leads to the original inputs. It can be used for a sanity check, raising
a warning when the condition is not fulfilled.
kw_args [dict, optional] Dictionary of additional keyword arguments to pass to func.
inv_kw_args [dict, optional] Dictionary of additional keyword arguments to pass to in-
verse_func.
Methods
fit(X[, y]) Fit transformer by checking X.

inverse_transform(X[, y]) Transform X using the inverse function.
transform(X[, y]) Transform X using the forward function.
__init__(func=None, inverse_func=None, validate=None, accept_sparse=False,

pass_y=’deprecated’, check_inverse=True, kw_args=None, inv_kw_args=None)
fit(X, y=None)
Fit transformer by checking X.
If validate is True, X will be checked.
Parameters
X [array-like, shape (n_samples, n_features)] Input array.
Returns
self
Parameters
Returns
Parameters
Returns

inverse_transform(X, y=’deprecated’)
Transform X using the inverse function.
Parameters
y [(ignored)]
Returns
X_out [array-like, shape (n_samples, n_features)] Transformed input.
of a nested object.
Returns
self
transform(X, y=’deprecated’)
Transform X using the forward function.
Parameters
y [(ignored)]
Returns
X_out [array-like, shape (n_samples, n_features)] Transformed input.
Examples using sklearn.preprocessing.FunctionTransformer
6.32.3 sklearn.preprocessing.KBinsDiscretizer
class sklearn.preprocessing.KBinsDiscretizer(n_bins=5, encode=’onehot’, strat-

egy=’quantile’)
Bin continuous data into intervals.
Parameters
n_bins [int or array-like, shape (n_features,) (default=5)] The number of bins to produce.
Raises ValueError if n_bins < 2.
encode [{‘onehot’, ‘onehot-dense’, ‘ordinal’}, (default=’onehot’)] Method used to encode the
transformed result.
onehot Encode the transformed result with one-hot encoding and return a sparse matrix.
Ignored features are always stacked to the right.
onehot-dense Encode the transformed result with one-hot encoding and return a dense ar-
ray. Ignored features are always stacked to the right.
ordinal Return the bin identifier encoded as an integer value.

strategy [{‘uniform’, ‘quantile’, ‘kmeans’}, (default=’quantile’)] Strategy used to define the

widths of the bins.
uniform All bins in each feature have identical widths.
quantile All bins in each feature have the same number of points.
kmeans Values in each bin have the same nearest center of a 1D k-means cluster.
Attributes
n_bins_ [int array, shape (n_features,)] Number of bins per feature. Bins whose width are too
small (i.e., <= 1e-8) are removed with a warning.
bin_edges_ [array of arrays, shape (n_features, )] The edges of each bin. Contain arrays of
varying shapes (n_bins_, ) Ignored features will have empty arrays.
See also:
sklearn.preprocessing.Binarizer class used to bin values as 0 or 1 based on a parameter

threshold.
Notes
In bin edges for feature i, the first and last values are used only for inverse_transform. During transform,
bin edges are extended to:
np.concatenate([-np.inf, bin_edges_[i][1:-1], np.inf])
You can combine KBinsDiscretizer with sklearn.compose.ColumnTransformer if you only

want to preprocess part of the features.
KBinsDiscretizer might produce constant features (e.g., when encode = 'onehot' and certain bins
do not contain any data). These features can be removed with feature selection algorithms (e.g., sklearn.
feature_selection.VarianceThreshold).
Examples
>>> X = [[-2, 1, -4, -1],

... [-1, 2, -3, -0.5],
... [ 0, 3, -2, 0.5],
... [ 1, 4, -1, 2]]
>>> est = KBinsDiscretizer(n_bins=3, encode='ordinal', strategy='uniform')
>>> est.fit(X)
KBinsDiscretizer(...)
>>> Xt = est.transform(X)
>>> Xt
array([[ 0., 0., 0., 0.],
[ 1., 1., 1., 0.],
[ 2., 2., 2., 1.],
[ 2., 2., 2., 2.]])
Sometimes it may be useful to convert the data back into the original feature space. The
inverse_transform function converts the binned data into the original feature space. Each value will
be equal to the mean of the two bin edges.

>>> est.bin_edges_[0]
array([-2., -1., 0., 1.])
>>> est.inverse_transform(Xt)
array([[-1.5, 1.5, -3.5, -0.5],
[-0.5, 2.5, -2.5, -0.5],
[ 0.5, 3.5, -1.5, 0.5],
[ 0.5, 3.5, -1.5, 1.5]])
Methods
fit(X[, y]) Fits the estimator.

inverse_transform(Xt) Transforms discretized data back to original feature
space.
transform(X) Discretizes the data.
__init__(n_bins=5, encode=’onehot’, strategy=’quantile’)

fit(X, y=None)
Fits the estimator.
Parameters
X [numeric array-like, shape (n_samples, n_features)] Data to be discretized.
y [ignored]
Returns
self
Parameters
Returns
Parameters
Returns

Transforms discretized data back to original feature space.
Note that this function does not regenerate the original data due to discretization rounding.
Parameters
Xt [numeric array-like, shape (n_sample, n_features)] Transformed data in the binned space.
Returns
Xinv [numeric array-like] Data in the original feature space.
of a nested object.
Returns
self
transform(X)
Discretizes the data.
Parameters
X [numeric array-like, shape (n_samples, n_features)] Data to be discretized.
Returns
Xt [numeric array-like or sparse matrix] Data in the binned space.
Examples using sklearn.preprocessing.KBinsDiscretizer

6.32.4 sklearn.preprocessing.KernelCenterer
class sklearn.preprocessing.KernelCenterer
Center a kernel matrix
Let K(x, z) be a kernel defined by phi(x)^T phi(z), where phi is a function mapping x to a Hilbert space.
KernelCenterer centers (i.e., normalize to have zero mean) the data without explicitly computing phi(x). It is
equivalent to centering phi(x) with sklearn.preprocessing.StandardScaler(with_std=False).
Examples
>>> from sklearn.preprocessing import KernelCenterer

>>> from sklearn.metrics.pairwise import pairwise_kernels
>>> X = [[ 1., -2., 2.],
... [ -2., 1., 3.],

... [ 4., 1., -2.]]

>>> K = pairwise_kernels(X, metric='linear')
>>> K
array([[ 9., 2., -2.],
[ 2., 14., -13.],
[ -2., -13., 21.]])
>>> transformer = KernelCenterer().fit(K)
>>> transformer
KernelCenterer()
>>> transformer.transform(K)
array([[ 5., 0., -5.],
[ 0., 14., -14.],
[ -5., -14., 19.]])
Methods
fit(K[, y]) Fit KernelCenterer

transform(K[, y, copy]) Center kernel matrix.
__init__()
fit(K, y=None)
Fit KernelCenterer
Parameters
K [numpy array of shape [n_samples, n_samples]] Kernel matrix.
Returns
Parameters
Returns
Parameters
Returns


of a nested object.
Returns
self
transform(K, y=’deprecated’, copy=True)
Center kernel matrix.
Parameters
K [numpy array of shape [n_samples1, n_samples2]] Kernel matrix.
copy [boolean, optional, default True] Set to False to perform inplace computation.
Returns
K_new [numpy array of shape [n_samples1, n_samples2]]
6.32.5 sklearn.preprocessing.LabelBinarizer
class sklearn.preprocessing.LabelBinarizer(neg_label=0, pos_label=1,

sparse_output=False)
Binarize labels in a one-vs-all fashion
Several regression and binary classification algorithms are available in scikit-learn. A simple way to extend
these algorithms to the multi-class classification case is to use the so-called one-vs-all scheme.
At learning time, this simply consists in learning one regressor or binary classifier per class. In doing so, one
needs to convert multi-class labels to binary labels (belong or does not belong to the class). LabelBinarizer
makes this process easy with the transform method.
At prediction time, one assigns the class for which the corresponding model gave the greatest confidence. La-
belBinarizer makes this easy with the inverse_transform method.
Parameters
neg_label [int (default: 0)] Value with which negative labels must be encoded.
pos_label [int (default: 1)] Value with which positive labels must be encoded.
sparse_output [boolean (default: False)] True if the returned array from transform is desired to
be in sparse CSR format.
Attributes
classes_ [array of shape [n_class]] Holds the label for each class.
y_type_ [str,] Represents the type of the target data as evaluated by
utils.multiclass.type_of_target. Possible type are ‘continuous’, ‘continuous-multioutput’,
‘binary’, ‘multiclass’, ‘multiclass-multioutput’, ‘multilabel-indicator’, and ‘unknown’.
sparse_input_ [boolean,] True if the input data to transform is given as a sparse matrix, False
otherwise.

See also:
label_binarize function to perform the transform operation of LabelBinarizer with fixed classes.
sklearn.preprocessing.OneHotEncoder encode categorical features using a one-hot aka one-of-K
scheme.
Examples

>>> lb.fit([1, 2, 6, 4, 2])
>>> lb.classes_
array([1, 2, 4, 6])
>>> lb.transform([1, 6])
array([[1, 0, 0, 0],
[0, 0, 0, 1]])
Binary targets transform to a column vector
>>> lb.fit_transform(['yes', 'no', 'no', 'yes'])
array([[1],
[0],
[0],
[1]])
Passing a 2D matrix for multilabel classification

>>> lb.fit(np.array([[0, 1, 1], [1, 0, 0]]))
>>> lb.classes_
array([0, 1, 2])
>>> lb.transform([0, 1, 2, 1])
array([[1, 0, 0],
[0, 1, 0],
[0, 0, 1],
[0, 1, 0]])
Methods
fit(y) Fit label binarizer

fit_transform(y) Fit label binarizer and transform multi-class labels to
binary labels.
inverse_transform(Y[, threshold]) Transform binary labels back to multi-class labels
transform(y) Transform multi-class labels to binary labels
__init__(neg_label=0, pos_label=1, sparse_output=False)

fit(y)
Fit label binarizer
Parameters
y [array of shape [n_samples,] or [n_samples, n_classes]] Target values. The 2-d matrix
should only contain 0 and 1, represents multilabel classification.
Returns
fit_transform(y)
Fit label binarizer and transform multi-class labels to binary labels.
The output of transform is sometimes referred to as the 1-of-K coding scheme.
Parameters
y [array or sparse matrix of shape [n_samples,] or [n_samples, n_classes]] Target values.
The 2-d matrix should only contain 0 and 1, represents multilabel classification. Sparse
matrix can be CSR, CSC, COO, DOK, or LIL.
Returns
Y [array or CSR matrix of shape [n_samples, n_classes]] Shape will be [n_samples, 1] for
binary problems.
Parameters
Returns
inverse_transform(Y, threshold=None)
Transform binary labels back to multi-class labels
Parameters
Y [numpy array or sparse matrix with shape [n_samples, n_classes]] Target values. All
sparse matrices are converted to CSR before inverse transformation.
threshold [float or None] Threshold used in the binary and multi-label cases.
Use 0 when Y contains the output of decision_function (classifier). Use 0.5 when Y con-
tains the output of predict_proba.
If None, the threshold is assumed to be half way between neg_label and pos_label.
Returns
y [numpy array or CSR matrix of shape [n_samples] Target values.]
Notes
In the case when the binary labels are fractional (probabilistic), inverse_transform chooses the class with
the greatest value. Typically, this allows to use the output of a linear model’s decision_function method
directly as the input of inverse_transform.

of a nested object.
Returns
self
transform(y)
Transform multi-class labels to binary labels
The output of transform is sometimes referred to by some authors as the 1-of-K coding scheme.
Parameters
y [array or sparse matrix of shape [n_samples,] or [n_samples, n_classes]] Target values.
The 2-d matrix should only contain 0 and 1, represents multilabel classification. Sparse
matrix can be CSR, CSC, COO, DOK, or LIL.
Returns
Y [numpy array or CSR matrix of shape [n_samples, n_classes]] Shape will be [n_samples,
1] for binary problems.
6.32.6 sklearn.preprocessing.LabelEncoder
class sklearn.preprocessing.LabelEncoder
Encode labels with value between 0 and n_classes-1.
Attributes
classes_ [array of shape (n_class,)] Holds the label for each class.
See also:
sklearn.preprocessing.OrdinalEncoder encode categorical features using a one-hot or ordinal

encoding scheme.
Examples
LabelEncoder can be used to normalize labels.

>>> le.fit([1, 2, 2, 6])
LabelEncoder()
>>> le.classes_
array([1, 2, 6])
>>> le.transform([1, 1, 2, 6])
array([0, 0, 1, 2]...)
>>> le.inverse_transform([0, 0, 1, 2])
array([1, 1, 2, 6])

It can also be used to transform non-numerical labels (as long as they are hashable and comparable) to numerical
labels.
>>> le.fit(["paris", "paris", "tokyo", "amsterdam"])
LabelEncoder()
>>> list(le.classes_)
['amsterdam', 'paris', 'tokyo']
>>> le.transform(["tokyo", "tokyo", "paris"])
array([2, 2, 1]...)
>>> list(le.inverse_transform([2, 2, 1]))
['tokyo', 'tokyo', 'paris']
Methods
fit(y) Fit label encoder

fit_transform(y) Fit label encoder and return encoded labels
inverse_transform(y) Transform labels back to original encoding.
transform(y) Transform labels to normalized encoding.

fit(y)
Fit label encoder
Parameters
y [array-like of shape (n_samples,)] Target values.
Returns
fit_transform(y)
Fit label encoder and return encoded labels
Parameters
y [array-like of shape [n_samples]] Target values.
Returns
y [array-like of shape [n_samples]]
Parameters
Returns

inverse_transform(y)
Transform labels back to original encoding.
Parameters
Returns
y [numpy array of shape [n_samples]]
of a nested object.
Returns
self
transform(y)
Transform labels to normalized encoding.
Parameters
y [array-like of shape [n_samples]] Target values.
Returns
y [array-like of shape [n_samples]]
6.32.7 sklearn.preprocessing.MultiLabelBinarizer
class sklearn.preprocessing.MultiLabelBinarizer(classes=None, sparse_output=False)

Transform between iterable of iterables and a multilabel format
Although a list of sets or tuples is a very intuitive format for multilabel data, it is unwieldy to process. This
transformer converts between this intuitive format and the supported multilabel format: a (samples x classes)
binary matrix indicating the presence of a class label.
Parameters
classes [array-like of shape [n_classes] (optional)] Indicates an ordering for the class labels. All
entries should be unique (cannot contain duplicate classes).
sparse_output [boolean (default: False),] Set to true if output binary array is desired in CSR
sparse format
Attributes
classes_ [array of labels] A copy of the classes parameter where provided, or otherwise, the
sorted set of classes found when fitting.
See also:
sklearn.preprocessing.OneHotEncoder encode categorical features using a one-hot aka one-of-K

scheme.

Examples

>>> mlb = MultiLabelBinarizer()
>>> mlb.fit_transform([(1, 2), (3,)])
array([[1, 1, 0],
[0, 0, 1]])
>>> mlb.classes_
array([1, 2, 3])
>>> mlb.fit_transform([set(['sci-fi', 'thriller']), set(['comedy'])])

array([[0, 1, 1],
[1, 0, 0]])
>>> list(mlb.classes_)
['comedy', 'sci-fi', 'thriller']
Methods
fit(y) Fit the label sets binarizer, storing classes_

fit_transform(y) Fit the label sets binarizer and transform the given label
sets
inverse_transform(yt) Transform the given indicator matrix into label sets
transform(y) Transform the given label sets
__init__(classes=None, sparse_output=False)
fit(y)
Fit the label sets binarizer, storing classes_
Parameters
y [iterable of iterables] A set of labels (any orderable and hashable object) for each sample.
If the classes parameter is set, y will not be iterated.
Returns
self [returns this MultiLabelBinarizer instance]
fit_transform(y)
Fit the label sets binarizer and transform the given label sets
Parameters
Returns
y_indicator [array or CSR matrix, shape (n_samples, n_classes)] A matrix such that
y_indicator[i, j] = 1 iff classes_[j] is in y[i], and 0 otherwise.
Parameters

Returns
inverse_transform(yt)
Transform the given indicator matrix into label sets
Parameters
yt [array or sparse matrix of shape (n_samples, n_classes)] A matrix containing only 1s ands
0s.
Returns
y [list of tuples] The set of labels for each sample such that y[i] consists of classes_[j] for
each yt[i, j] == 1.
of a nested object.
Returns
self
transform(y)
Transform the given label sets
Parameters
Returns
y_indicator [array or CSR matrix, shape (n_samples, n_classes)] A matrix such that
y_indicator[i, j] = 1 iff classes_[j] is in y[i], and 0 otherwise.
6.32.8 sklearn.preprocessing.MaxAbsScaler
class sklearn.preprocessing.MaxAbsScaler(copy=True)
Scale each feature by its maximum absolute value.
This estimator scales and translates each feature individually such that the maximal absolute value of each
feature in the training set will be 1.0. It does not shift/center the data, and thus does not destroy any sparsity.
This scaler can also be applied to sparse CSR or CSC matrices.
Parameters
copy [boolean, optional, default is True] Set to False to perform inplace scaling and avoid a
copy (if the input is already a numpy array).
Attributes

scale_ [ndarray, shape (n_features,)] Per feature relative scaling of the data.
New in version 0.17: scale_ attribute.
max_abs_ [ndarray, shape (n_features,)] Per feature maximum absolute value.
n_samples_seen_ [int] The number of samples processed by the estimator. Will be reset on
new calls to fit, but increments across partial_fit calls.
See also:
maxabs_scale Equivalent function without the estimator API.
Notes
NaNs are treated as missing values: disregarded in fit, and maintained in transform.
For a comparison of the different scalers, transformers, and normalizers, see exam-
ples/preprocessing/plot_all_scaling.py.
Examples
>>> from sklearn.preprocessing import MaxAbsScaler

>>> X = [[ 1., -1., 2.],
... [ 2., 0., 0.],
... [ 0., 1., -1.]]
>>> transformer = MaxAbsScaler().fit(X)
>>> transformer
MaxAbsScaler(copy=True)
array([[ 0.5, -1. , 1. ],
[ 1. , 0. , 0. ],
[ 0. , 1. , -0.5]])
Methods
fit(X[, y]) Compute the maximum absolute value to be used for

later scaling.
inverse_transform(X) Scale back the data to the original representation
partial_fit(X[, y]) Online computation of max absolute value of X for later
scaling.
transform(X) Scale the data
__init__(copy=True)
fit(X, y=None)
Compute the maximum absolute value to be used for later scaling.
Parameters
X [{array-like, sparse matrix}, shape [n_samples, n_features]] The data used to compute the

per-feature minimum and maximum used for later scaling along the features axis.
Parameters
Returns
Parameters
Returns
Scale back the data to the original representation
Parameters
X [{array-like, sparse matrix}] The data that should be transformed back.
Online computation of max absolute value of X for later scaling. All of X is processed as a single batch.
This is intended for cases when fit is not feasible due to very large number of n_samples or because X is
read from a continuous stream.
Parameters
mean and standard deviation used for later scaling along the features axis.
y Ignored
of a nested object.
Returns
self
transform(X)
Scale the data
Parameters
X [{array-like, sparse matrix}] The data that should be scaled.

Examples using sklearn.preprocessing.MaxAbsScaler
6.32.9 sklearn.preprocessing.MinMaxScaler
class sklearn.preprocessing.MinMaxScaler(feature_range=(0, 1), copy=True)

Transforms features by scaling each feature to a given range.
This estimator scales and translates each feature individually such that it is in the given range on the training set,
e.g. between zero and one.
The transformation is given by:

where min, max = feature_range.

The transformation is calculated as:
X_scaled = scale * X + min - X.min(axis=0) * scale

where scale = (max - min) / (X.max(axis=0) - X.min(axis=0))
This transformation is often used as an alternative to zero mean, unit variance scaling.
Parameters
feature_range [tuple (min, max), default=(0, 1)] Desired range of transformed data.
copy [boolean, optional, default True] Set to False to perform inplace row normalization and
avoid a copy (if the input is already a numpy array).
Attributes
min_ [ndarray, shape (n_features,)] Per feature adjustment for minimum. Equivalent to min -
X.min(axis=0) * self.scale_
scale_ [ndarray, shape (n_features,)] Per feature relative scaling of the data. Equivalent to (max
- min) / (X.max(axis=0) - X.min(axis=0))
data_min_ [ndarray, shape (n_features,)] Per feature minimum seen in the data
New in version 0.17: data_min_
data_max_ [ndarray, shape (n_features,)] Per feature maximum seen in the data
New in version 0.17: data_max_
data_range_ [ndarray, shape (n_features,)] Per feature range (data_max_ -
data_min_) seen in the data
New in version 0.17: data_range_
See also:
minmax_scale Equivalent function without the estimator API.

Notes
Examples
>>> from sklearn.preprocessing import MinMaxScaler
>>> data = [[-1, 2], [-0.5, 6], [0, 10], [1, 18]]
>>> scaler = MinMaxScaler()
>>> print(scaler.fit(data))
MinMaxScaler(copy=True, feature_range=(0, 1))
>>> print(scaler.data_max_)
[ 1. 18.]
>>> print(scaler.transform(data))
[[0. 0. ]
[0.25 0.25]
[0.5 0.5 ]
[1. 1. ]]
>>> print(scaler.transform([[2, 2]]))
[[1.5 0. ]]
Methods
fit(X[, y]) Compute the minimum and maximum to be used for

later scaling.
inverse_transform(X) Undo the scaling of X according to feature_range.
partial_fit(X[, y]) Online computation of min and max on X for later scal-
ing.
transform(X) Scaling features of X according to feature_range.
__init__(feature_range=(0, 1), copy=True)

fit(X, y=None)
Compute the minimum and maximum to be used for later scaling.
Parameters
X [array-like, shape [n_samples, n_features]] The data used to compute the per-feature min-
imum and maximum used for later scaling along the features axis.
Parameters


Returns
Parameters
Returns
Undo the scaling of X according to feature_range.
Parameters
X [array-like, shape [n_samples, n_features]] Input data that will be transformed. It cannot
be sparse.
Online computation of min and max on X for later scaling. All of X is processed as a single batch. This
is intended for cases when fit is not feasible due to very large number of n_samples or because X is read
from a continuous stream.
Parameters
X [array-like, shape [n_samples, n_features]] The data used to compute the mean and stan-
dard deviation used for later scaling along the features axis.
y Ignored
of a nested object.
Returns
self
transform(X)
Scaling features of X according to feature_range.
Parameters
X [array-like, shape [n_samples, n_features]] Input data that will be transformed.
Examples using sklearn.preprocessing.MinMaxScaler


6.32.10 sklearn.preprocessing.Normalizer
class sklearn.preprocessing.Normalizer(norm=’l2’, copy=True)

Normalize samples individually to unit norm.
Each sample (i.e. each row of the data matrix) with at least one non zero component is rescaled independently
of other samples so that its norm (l1 or l2) equals one.
This transformer is able to work both with dense numpy arrays and scipy.sparse matrix (use CSR format if you
want to avoid the burden of a copy / conversion).
Scaling inputs to unit norms is a common operation for text classification or clustering for instance. For instance
the dot product of two l2-normalized TF-IDF vectors is the cosine similarity of the vectors and is the base
similarity metric for the Vector Space Model commonly used by the Information Retrieval community.
Parameters
norm [‘l1’, ‘l2’, or ‘max’, optional (‘l2’ by default)] The norm to use to normalize each non
zero sample.
copy [boolean, optional, default True] set to False to perform inplace row normalization and
avoid a copy (if the input is already a numpy array or a scipy.sparse CSR matrix).
See also:
normalize Equivalent function without the estimator API.
Notes
This estimator is stateless (besides constructor parameters), the fit method does nothing but is useful when used
in a pipeline.
Examples
>>> from sklearn.preprocessing import Normalizer

>>> X = [[4, 1, 2, 2],
... [1, 3, 9, 3],
... [5, 7, 5, 1]]
>>> transformer = Normalizer().fit(X) # fit does nothing.
>>> transformer
Normalizer(copy=True, norm='l2')
array([[0.8, 0.2, 0.4, 0.4],
[0.1, 0.3, 0.9, 0.3],
[0.5, 0.7, 0.5, 0.1]])
Methods


transform(X[, y, copy]) Scale each non zero row of X to unit norm
__init__(norm=’l2’, copy=True)
fit(X, y=None)
Parameters
X [array-like]
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
Scale each non zero row of X to unit norm
Parameters
X [{array-like, sparse matrix}, shape [n_samples, n_features]] The data to normalize, row
by row. scipy.sparse matrices should be in CSR format to avoid an un-necessary copy.

copy [bool, optional (default: None)] Copy the input X or not.
Examples using sklearn.preprocessing.Normalizer

6.32.11 sklearn.preprocessing.OneHotEncoder
class sklearn.preprocessing.OneHotEncoder(n_values=None, categorical_features=None,

categories=None, sparse=True, dtype=<class
‘numpy.float64’>, handle_unknown=’error’)
Encode categorical integer features as a one-hot numeric array.
The input to this transformer should be an array-like of integers or strings, denoting the values taken on by
categorical (discrete) features. The features are encoded using a one-hot (aka ‘one-of-K’ or ‘dummy’) encoding
scheme. This creates a binary column for each category and returns a sparse matrix or dense array.
By default, the encoder derives the categories based on the unique values in each feature. Alternatively, you can
also specify the categories manually. The OneHotEncoder previously assumed that the input features take on
values in the range [0, max(values)). This behaviour is deprecated.
This encoding is needed for feeding categorical data to many scikit-learn estimators, notably linear models and
SVMs with the standard kernels.
Note: a one-hot encoding of y labels should use a LabelBinarizer instead.
Parameters
categories [‘auto’ or a list of lists/arrays of values, default=’auto’.] Categories (unique values)
per feature:
• ‘auto’ : Determine categories automatically from the training data.
• list : categories[i] holds the categories expected in the ith column. The passed
categories should not mix strings and numeric values within a single feature, and should
be sorted in case of numeric values.
The used categories can be found in the categories_ attribute.
sparse [boolean, default=True] Will return sparse matrix if set True else will return an array.
dtype [number type, default=np.float] Desired dtype of output.
handle_unknown [‘error’ or ‘ignore’, default=’error’.] Whether to raise an error or ignore if
an unknown categorical feature is present during transform (default is to raise). When this
parameter is set to ‘ignore’ and an unknown category is encountered during transform, the
resulting one-hot encoded columns for this feature will be all zeros. In the inverse transform,
an unknown category will be denoted as None.
n_values [‘auto’, int or array of ints, default=’auto’] Number of values per feature.
• ‘auto’ : determine value range from training data.
• int [number of categorical values per feature.] Each feature value should be in
range(n_values)
• array [n_values[i] is the number of categorical values in] X[:, i]. Each feature
value should be in range(n_values[i])

Deprecated since version 0.20: The n_values keyword was deprecated in version 0.20 and
will be removed in 0.22. Use categories instead.
categorical_features [‘all’ or array of indices or mask, default=’all’] Specify what features are
treated as categorical.
• ‘all’: All features are treated as categorical.
• array of indices: Array of categorical feature indices.
• mask: Array of length n_features and with dtype=bool.
Non-categorical features are always stacked to the right of the matrix.
Deprecated since version 0.20: The categorical_features keyword was deprecated in version
0.20 and will be removed in 0.22. You can use the ColumnTransformer instead.
Attributes
categories_ [list of arrays] The categories of each feature determined during fitting (in order of
the features in X and corresponding with the output of transform).
active_features_ [array] DEPRECATED: The active_features_ attribute was
deprecated in version 0.20 and will be removed 0.22.
feature_indices_ [array of shape (n_features,)] DEPRECATED: The
feature_indices_ attribute was deprecated in version 0.20 and will be removed
0.22.
n_values_ [array of shape (n_features,)] DEPRECATED: The n_values_ attribute was
deprecated in version 0.20 and will be removed 0.22.
See also:
sklearn.preprocessing.OrdinalEncoder performs an ordinal (integer) encoding of the categorical

features.
sklearn.feature_extraction.DictVectorizer performs a one-hot encoding of dictionary items
(also handles string-valued features).
sklearn.feature_extraction.FeatureHasher performs an approximate one-hot encoding of dic-
tionary items or strings.
sklearn.preprocessing.LabelBinarizer binarizes labels in a one-vs-all fashion.
sklearn.preprocessing.MultiLabelBinarizer transforms between iterable of iterables and a
multilabel format, e.g. a (samples x classes) binary matrix indicating the presence of a class label.
Examples
Given a dataset with two features, we let the encoder find the unique values per feature and transform the data
to a binary one-hot encoding.
>>> from sklearn.preprocessing import OneHotEncoder

>>> enc = OneHotEncoder(handle_unknown='ignore')
>>> X = [['Male', 1], ['Female', 3], ['Female', 2]]
>>> enc.fit(X)
...
dtype=<... 'numpy.float64'>, handle_unknown='ignore',

>>> enc.categories_
[array(['Female', 'Male'], dtype=object), array([1, 2, 3], dtype=object)]
>>> enc.transform([['Female', 1], ['Male', 4]]).toarray()
array([[1., 0., 1., 0., 0.],
[0., 1., 0., 0., 0.]])
>>> enc.inverse_transform([[0, 1, 1, 0, 0], [0, 0, 0, 1, 0]])
array([['Male', 1],
[None, 2]], dtype=object)
>>> enc.get_feature_names()
array(['x0_Female', 'x0_Male', 'x1_1', 'x1_2', 'x1_3'], dtype=object)
Methods
fit(X[, y]) Fit OneHotEncoder to X.

fit_transform(X[, y]) Fit OneHotEncoder to X, then transform X.
get_feature_names([input_features]) Return feature names for output features.
inverse_transform(X) Convert the back data to the original representation.
transform(X) Transform X using one-hot encoding.
__init__(n_values=None, categorical_features=None, categories=None, sparse=True, dtype=<class

‘numpy.float64’>, handle_unknown=’error’)
active_features_
DEPRECATED: The active_features_ attribute was deprecated in version 0.20 and will be removed
0.22.
feature_indices_
DEPRECATED: The feature_indices_ attribute was deprecated in version 0.20 and will be removed
0.22.
fit(X, y=None)
Fit OneHotEncoder to X.
Parameters
X [array-like, shape [n_samples, n_features]] The data to determine the categories of each
feature.
Returns
self
Fit OneHotEncoder to X, then transform X.
Equivalent to fit(X).transform(X) but more convenient.
Parameters
X [array-like, shape [n_samples, n_features]] The data to encode.
Returns
X_out [sparse matrix if sparse=True else a 2-d array] Transformed input.
get_feature_names(input_features=None)
Return feature names for output features.

Parameters
input_features [list of string, length n_features, optional] String names for input features if
available. By default, “x0”, “x1”, . . . “xn_features” is used.
Returns
output_feature_names [array of string, length n_output_features]
Parameters
Returns
Convert the back data to the original representation.
In case unknown categories are encountered (all zero’s in the one-hot encoding), None is used to represent
this category.
Parameters
X [array-like or sparse matrix, shape [n_samples, n_encoded_features]] The transformed
data.
Returns
X_tr [array-like, shape [n_samples, n_features]] Inverse transformed array.
n_values_
DEPRECATED: The n_values_ attribute was deprecated in version 0.20 and will be removed 0.22.
of a nested object.
Returns
self
transform(X)
Transform X using one-hot encoding.
Parameters
Returns
X_out [sparse matrix if sparse=True else a 2-d array] Transformed input.
Examples using sklearn.preprocessing.OneHotEncoder


6.32.12 sklearn.preprocessing.OrdinalEncoder
class sklearn.preprocessing.OrdinalEncoder(categories=’auto’, dtype=<class

Encode categorical features as an integer array.
The input to this transformer should be an array-like of integers or strings, denoting the values taken on by
categorical (discrete) features. The features are converted to ordinal integers. This results in a single column of
integers (0 to n_categories - 1) per feature.
Parameters
categories [‘auto’ or a list of lists/arrays of values.] Categories (unique values) per feature:
• ‘auto’ : Determine categories automatically from the training data.
• list : categories[i] holds the categories expected in the ith column. The passed
categories should not mix strings and numeric values, and should be sorted in case of
numeric values.
The used categories can be found in the categories_ attribute.
dtype [number type, default np.float64] Desired dtype of output.
Attributes
categories_ [list of arrays] The categories of each feature determined during fitting (in order of
the features in X and corresponding with the output of transform).
See also:
sklearn.preprocessing.OneHotEncoder performs a one-hot encoding of categorical features.

sklearn.preprocessing.LabelEncoder encodes target labels with values between 0 and n_classes-
1.
Examples
Given a dataset with two features, we let the encoder find the unique values per feature and transform the data
to an ordinal encoding.
>>> from sklearn.preprocessing import OrdinalEncoder

>>> enc = OrdinalEncoder()
>>> X = [['Male', 1], ['Female', 3], ['Female', 2]]
>>> enc.fit(X)
...
OrdinalEncoder(categories='auto', dtype=<... 'numpy.float64'>)
>>> enc.categories_
[array(['Female', 'Male'], dtype=object), array([1, 2, 3], dtype=object)]
>>> enc.transform([['Female', 3], ['Male', 1]])
array([[0., 2.],
[1., 0.]])
>>> enc.inverse_transform([[1, 0], [0, 1]])

array([['Male', 1],
['Female', 2]], dtype=object)

Methods
fit(X[, y]) Fit the OrdinalEncoder to X.

inverse_transform(X) Convert the data back to the original representation.
transform(X) Transform X to ordinal codes.
__init__(categories=’auto’, dtype=<class ‘numpy.float64’>)

fit(X, y=None)
Fit the OrdinalEncoder to X.
Parameters
X [array-like, shape [n_samples, n_features]] The data to determine the categories of each
feature.
Returns
self
Parameters
Returns
Parameters
Returns
Convert the data back to the original representation.
Parameters
X [array-like or sparse matrix, shape [n_samples, n_encoded_features]] The transformed
data.
Returns
X_tr [array-like, shape [n_samples, n_features]] Inverse transformed array.

of a nested object.
Returns
self
transform(X)
Transform X to ordinal codes.
Parameters
Returns
X_out [sparse matrix or a 2-d array] Transformed input.
6.32.13 sklearn.preprocessing.PolynomialFeatures
class sklearn.preprocessing.PolynomialFeatures(degree=2, interaction_only=False, in-

clude_bias=True)
Generate polynomial and interaction features.
Generate a new feature matrix consisting of all polynomial combinations of the features with degree less than
or equal to the specified degree. For example, if an input sample is two dimensional and of the form [a, b], the
degree-2 polynomial features are [1, a, b, a^2, ab, b^2].
Parameters
degree [integer] The degree of the polynomial features. Default = 2.
interaction_only [boolean, default = False] If true, only interaction features are produced: fea-
tures that are products of at most degree distinct input features (so not x[1] ** 2,
x[0] * x[2] ** 3, etc.).
include_bias [boolean] If True (default), then include a bias column, the feature in which all
polynomial powers are zero (i.e. a column of ones - acts as an intercept term in a linear
model).
Attributes
powers_ [array, shape (n_output_features, n_input_features)] powers_[i, j] is the exponent of
the jth input in the ith output.
n_input_features_ [int] The total number of input features.
n_output_features_ [int] The total number of polynomial output features. The number of out-
put features is computed by iterating over all suitably sized combinations of input features.
Notes
Be aware that the number of features in the output array scales polynomially in the number of features of the
input array, and exponentially in the degree. High degrees can cause overfitting.
See examples/linear_model/plot_polynomial_interpolation.py

Examples
>>> X
array([[0, 1],
[2, 3],
[4, 5]])
>>> poly = PolynomialFeatures(2)
array([[ 1., 0., 1., 0., 0., 1.],
[ 1., 2., 3., 4., 6., 9.],
[ 1., 4., 5., 16., 20., 25.]])
>>> poly = PolynomialFeatures(interaction_only=True)
array([[ 1., 0., 1., 0.],
[ 1., 2., 3., 6.],
[ 1., 4., 5., 20.]])
Methods
fit(X[, y]) Compute number of output features.

get_feature_names([input_features]) Return feature names for output features
transform(X) Transform data to polynomial features
__init__(degree=2, interaction_only=False, include_bias=True)

fit(X, y=None)
Compute number of output features.
Parameters
X [array-like, shape (n_samples, n_features)] The data.
Returns
self [instance]
Parameters
Returns
get_feature_names(input_features=None)
Return feature names for output features
Parameters

input_features [list of string, length n_features, optional] String names for input features if
available. By default, “x0”, “x1”, . . . “xn_features” is used.
Returns
output_feature_names [list of string, length n_output_features]
Parameters
Returns
of a nested object.
Returns
self
transform(X)
Transform data to polynomial features
Parameters
X [array-like or sparse matrix, shape [n_samples, n_features]] The data to transform, row
by row. Sparse input should preferably be in CSC format.
Returns
XP [np.ndarray or CSC sparse matrix, shape [n_samples, NP]] The matrix of features, where
NP is the number of polynomial features generated from the combination of inputs.
Examples using sklearn.preprocessing.PolynomialFeatures
6.32.14 sklearn.preprocessing.PowerTransformer
class sklearn.preprocessing.PowerTransformer(method=’yeo-johnson’, standardize=True,

copy=True)
Apply a power transform featurewise to make data more Gaussian-like.
Power transforms are a family of parametric, monotonic transformations that are applied to make data more
Gaussian-like. This is useful for modeling issues related to heteroscedasticity (non-constant variance), or other
situations where normality is desired.
Currently, PowerTransformer supports the Box-Cox transform and the Yeo-Johnson transform. The optimal
parameter for stabilizing variance and minimizing skewness is estimated through maximum likelihood.

Box-Cox requires input data to be strictly positive, while Yeo-Johnson supports both positive or negative data.
By default, zero-mean, unit-variance normalization is applied to the transformed data.
Parameters
method [str, (default=’yeo-johnson’)] The power transform method. Available methods are:
• ‘yeo-johnson’ [1], works with positive and negative values
• ‘box-cox’ [2], only works with strictly positive values
standardize [boolean, default=True] Set to True to apply zero-mean, unit-variance normaliza-
tion to the transformed output.
copy [boolean, optional, default=True] Set to False to perform inplace computation during
transformation.
Attributes
lambdas_ [array of float, shape (n_features,)] The parameters of the power transformation for
the selected features.
See also:
power_transform Equivalent function without the estimator API.

QuantileTransformer Maps data to a standard normal distribution with the parameter out-
put_distribution=’normal’.
Notes
References
[1], [2]
Examples

>>> from sklearn.preprocessing import PowerTransformer
>>> pt = PowerTransformer()
>>> data = [[1, 2], [3, 2], [4, 5]]
>>> print(pt.fit(data))
PowerTransformer(copy=True, method='yeo-johnson', standardize=True)
>>> print(pt.lambdas_)
[ 1.386... -3.100...]
>>> print(pt.transform(data))
[[-1.316... -0.707...]
[ 0.209... -0.707...]
[ 1.106... 1.414...]]

Methods
fit(X[, y]) Estimate the optimal parameter lambda for each feature.
fit_transform(X[, y])
inverse_transform(X) Apply the inverse power transformation using the fitted
lambdas.
transform(X) Apply the power transform to each feature using the fit-
ted lambdas.
__init__(method=’yeo-johnson’, standardize=True, copy=True)

fit(X, y=None)
Estimate the optimal parameter lambda for each feature.
The optimal lambda parameter for minimizing skewness is estimated on each feature independently using
maximum likelihood.
Parameters
X [array-like, shape (n_samples, n_features)] The data used to estimate the optimal trans-
formation parameters.
y [Ignored]
Returns
self [object]
Parameters
Returns
Apply the inverse power transformation using the fitted lambdas.
The inverse of the Box-Cox transformation is given by:
if lambda == 0:
X = exp(X_trans)
else:
X = (X_trans * lambda + 1) ** (1 / lambda)
The inverse of the Yeo-Johnson transformation is given by:
if X >= 0 and lambda == 0:

X = exp(X_trans) - 1
elif X >= 0 and lambda != 0:
X = (X_trans * lambda + 1) ** (1 / lambda) - 1
elif X < 0 and lambda != 2:
X = 1 - (-(2 - lambda) * X_trans + 1) ** (1 / (2 - lambda))

elif X < 0 and lambda == 2:

X = 1 - exp(-X_trans)
Parameters
X [array-like, shape (n_samples, n_features)] The transformed data.
Returns
X [array-like, shape (n_samples, n_features)] The original data
of a nested object.
Returns
self
transform(X)
Apply the power transform to each feature using the fitted lambdas.
Parameters
X [array-like, shape (n_samples, n_features)] The data to be transformed using a power
transformation.
Returns
X_trans [array-like, shape (n_samples, n_features)] The transformed data.
Examples using sklearn.preprocessing.PowerTransformer

6.32.15 sklearn.preprocessing.QuantileTransformer
class sklearn.preprocessing.QuantileTransformer(n_quantiles=1000, out-

put_distribution=’uniform’, ig-
nore_implicit_zeros=False, subsam-
ple=100000, random_state=None,
copy=True)
Transform features using quantiles information.
This method transforms the features to follow a uniform or a normal distribution. Therefore, for a given feature,
this transformation tends to spread out the most frequent values. It also reduces the impact of (marginal) outliers:
this is therefore a robust preprocessing scheme.
The transformation is applied on each feature independently. The cumulative distribution function of a feature
is used to project the original values. Features values of new/unseen data that fall below or above the fitted range
will be mapped to the bounds of the output distribution. Note that this transform is non-linear. It may distort
linear correlations between variables measured at the same scale but renders variables measured at different
scales more directly comparable.


Parameters
n_quantiles [int, optional (default=1000)] Number of quantiles to be computed. It corresponds
to the number of landmarks used to discretize the cumulative distribution function.
output_distribution [str, optional (default=’uniform’)] Marginal distribution for the trans-
formed data. The choices are ‘uniform’ (default) or ‘normal’.
ignore_implicit_zeros [bool, optional (default=False)] Only applies to sparse matrices. If True,
the sparse entries of the matrix are discarded to compute the quantile statistics. If False,
these entries are treated as zeros.
subsample [int, optional (default=1e5)] Maximum number of samples used to estimate the
quantiles for computational efficiency. Note that the subsampling procedure may differ
for value-identical sparse and dense matrices.
the RandomState instance used by np.random. Note that this is used by subsampling and
smoothing noise.
copy [boolean, optional, (default=True)] Set to False to perform inplace transformation and
Attributes
quantiles_ [ndarray, shape (n_quantiles, n_features)] The values corresponding the quantiles of
reference.
references_ [ndarray, shape(n_quantiles, )] Quantiles of references.
See also:
quantile_transform Equivalent function without the estimator API.

PowerTransformer Perform mapping to a normal distribution using a power transform.
StandardScaler Perform standardization that is faster, but less robust to outliers.
RobustScaler Perform robust standardization that removes the influence of outliers but does not put outliers
and inliers on the same scale.
Notes
Examples

>>> from sklearn.preprocessing import QuantileTransformer
>>> X = np.sort(rng.normal(loc=0.5, scale=0.25, size=(25, 1)), axis=0)
>>> qt = QuantileTransformer(n_quantiles=10, random_state=0)

>>> qt.fit_transform(X)
array([...])
Methods
fit(X[, y]) Compute the quantiles used for transforming.

inverse_transform(X) Back-projection to the original space.
transform(X) Feature-wise transformation of the data.
__init__(n_quantiles=1000, output_distribution=’uniform’, ignore_implicit_zeros=False, subsam-

ple=100000, random_state=None, copy=True)
fit(X, y=None)
Compute the quantiles used for transforming.
Parameters
X [ndarray or sparse matrix, shape (n_samples, n_features)] The data used to scale
along the features axis. If a sparse matrix is provided, it will be converted into a
sparse csc_matrix. Additionally, the sparse matrix needs to be nonnegative if ig-
nore_implicit_zeros is False.
Returns
self [object]
Parameters
Returns
Parameters
Returns
Back-projection to the original space.
Parameters

Returns
Xt [ndarray or sparse matrix, shape (n_samples, n_features)] The projected data.
of a nested object.
Returns
self
transform(X)
Feature-wise transformation of the data.
Parameters
Returns
Xt [ndarray or sparse matrix, shape (n_samples, n_features)] The projected data.
Examples using sklearn.preprocessing.QuantileTransformer

6.32.16 sklearn.preprocessing.RobustScaler
class sklearn.preprocessing.RobustScaler(with_centering=True, with_scaling=True, quan-

tile_range=(25.0, 75.0), copy=True)
Scale features using statistics that are robust to outliers.
This Scaler removes the median and scales the data according to the quantile range (defaults to IQR: Interquartile
Range). The IQR is the range between the 1st quartile (25th quantile) and the 3rd quartile (75th quantile).
Centering and scaling happen independently on each feature by computing the relevant statistics on the samples
in the training set. Median and interquartile range are then stored to be used on later data using the transform
method.
Standardization of a dataset is a common requirement for many machine learning estimators. Typically this is
done by removing the mean and scaling to unit variance. However, outliers can often influence the sample mean
/ variance in a negative way. In such cases, the median and the interquartile range often give better results.

Parameters
with_centering [boolean, True by default] If True, center the data before scaling. This will
cause transform to raise an exception when attempted on sparse matrices, because cen-
tering them entails building a dense matrix which in common use cases is likely to be too
large to fit in memory.
with_scaling [boolean, True by default] If True, scale the data to interquartile range.
quantile_range [tuple (q_min, q_max), 0.0 < q_min < q_max < 100.0] Default: (25.0, 75.0) =
(1st quantile, 3rd quantile) = IQR Quantile range used to calculate scale_.
copy [boolean, optional, default is True] If False, try to avoid a copy and do inplace scaling
instead. This is not guaranteed to always work inplace; e.g. if the data is not a NumPy array
or scipy.sparse CSR matrix, a copy may still be returned.
Attributes
center_ [array of floats] The median value for each feature in the training set.
scale_ [array of floats] The (scaled) interquartile range for each feature in the training set.
See also:
robust_scale Equivalent function without the estimator API.

sklearn.decomposition.PCA Further removes the linear correlation across features with
‘whiten=True’.
Notes
https://en.wikipedia.org/wiki/Median https://en.wikipedia.org/wiki/Interquartile_range
Examples
>>> from sklearn.preprocessing import RobustScaler

>>> X = [[ 1., -2., 2.],
... [ -2., 1., 3.],
... [ 4., 1., -2.]]
>>> transformer = RobustScaler().fit(X)
>>> transformer
RobustScaler(copy=True, quantile_range=(25.0, 75.0), with_centering=True,
with_scaling=True)
array([[ 0. , -2. , 0. ],
[-1. , 0. , 0.4],
[ 1. , 0. , -1.6]])

Methods
fit(X[, y]) Compute the median and quantiles to be used for scal-
ing.
inverse_transform(X) Scale back the data to the original representation
transform(X) Center and scale the data.
__init__(with_centering=True, with_scaling=True, quantile_range=(25.0, 75.0), copy=True)

fit(X, y=None)
Compute the median and quantiles to be used for scaling.
Parameters
X [array-like, shape [n_samples, n_features]] The data used to compute the median and
quantiles used for later scaling along the features axis.
Parameters
Returns
Parameters
Returns
Parameters
X [array-like] The data used to scale along the specified axis.
of a nested object.
Returns

self
transform(X)
Center and scale the data.
Parameters
X [{array-like, sparse matrix}] The data used to scale along the specified axis.
Examples using sklearn.preprocessing.RobustScaler
6.32.17 sklearn.preprocessing.StandardScaler
class sklearn.preprocessing.StandardScaler(copy=True, with_mean=True, with_std=True)

Standardize features by removing the mean and scaling to unit variance
The standard score of a sample x is calculated as:
z = (x - u) / s
where u is the mean of the training samples or zero if with_mean=False, and s is the standard deviation of the
training samples or one if with_std=False.
Centering and scaling happen independently on each feature by computing the relevant statistics on the samples
in the training set. Mean and standard deviation are then stored to be used on later data using the transform
method.
Standardization of a dataset is a common requirement for many machine learning estimators: they might behave
badly if the individual features do not more or less look like standard normally distributed data (e.g. Gaussian
with 0 mean and unit variance).
For instance many elements used in the objective function of a learning algorithm (such as the RBF kernel of
Support Vector Machines or the L1 and L2 regularizers of linear models) assume that all features are centered
around 0 and have variance in the same order. If a feature has a variance that is orders of magnitude larger
that others, it might dominate the objective function and make the estimator unable to learn from other features
correctly as expected.
This scaler can also be applied to sparse CSR or CSC matrices by passing with_mean=False to avoid breaking
the sparsity structure of the data.
Parameters
copy [boolean, optional, default True] If False, try to avoid a copy and do inplace scaling in-
stead. This is not guaranteed to always work inplace; e.g. if the data is not a NumPy array
or scipy.sparse CSR matrix, a copy may still be returned.
with_mean [boolean, True by default] If True, center the data before scaling. This does not
work (and will raise an exception) when attempted on sparse matrices, because centering
them entails building a dense matrix which in common use cases is likely to be too large to
fit in memory.
with_std [boolean, True by default] If True, scale the data to unit variance (or equivalently, unit
standard deviation).
Attributes

scale_ [ndarray or None, shape (n_features,)] Per feature relative scaling of the data. This is
calculated using np.sqrt(var_). Equal to None when with_std=False.
New in version 0.17: scale_
mean_ [ndarray or None, shape (n_features,)] The mean value for each feature in the training
set. Equal to None when with_mean=False.
var_ [ndarray or None, shape (n_features,)] The variance for each feature in the training set.
Used to compute scale_. Equal to None when with_std=False.
n_samples_seen_ [int or array, shape (n_features,)] The number of samples processed by the
estimator for each feature. If there are not missing samples, the n_samples_seen will
be an integer, otherwise it will be an array. Will be reset on new calls to fit, but increments
across partial_fit calls.
See also:
scale Equivalent function without the estimator API.

sklearn.decomposition.PCA Further removes the linear correlation across features with
‘whiten=True’.
Notes
We use a biased estimator for the standard deviation, equivalent to numpy.std(x, ddof=0). Note that the choice
of ddof is unlikely to affect model performance.
Examples

>>> data = [[0, 0], [0, 0], [1, 1], [1, 1]]
>>> scaler = StandardScaler()
>>> print(scaler.fit(data))
StandardScaler(copy=True, with_mean=True, with_std=True)
>>> print(scaler.mean_)
[0.5 0.5]
>>> print(scaler.transform(data))
[[-1. -1.]
[-1. -1.]
[ 1. 1.]
[ 1. 1.]]
>>> print(scaler.transform([[2, 2]]))
[[3. 3.]]
Methods
fit(X[, y]) Compute the mean and std to be used for later scaling.


inverse_transform(X[, copy]) Scale back the data to the original representation
partial_fit(X[, y]) Online computation of mean and std on X for later scal-
ing.
transform(X[, y, copy]) Perform standardization by centering and scaling
__init__(copy=True, with_mean=True, with_std=True)

fit(X, y=None)
Compute the mean and std to be used for later scaling.
Parameters
y Ignored
Parameters
Returns
Parameters
Returns
inverse_transform(X, copy=None)
Parameters
X [array-like, shape [n_samples, n_features]] The data used to scale along the features axis.
Returns
X_tr [array-like, shape [n_samples, n_features]] Transformed array.
Online computation of mean and std on X for later scaling. All of X is processed as a single batch. This
is intended for cases when fit is not feasible due to very large number of n_samples or because X is read
from a continuous stream.

The algorithm for incremental mean and std is given in Equation 1.5a,b in Chan, Tony F., Gene H. Golub,
and Randall J. LeVeque. “Algorithms for computing the sample variance: Analysis and recommendations.”
The American Statistician 37.3 (1983): 242-247:
Parameters
y Ignored
of a nested object.
Returns
self
Perform standardization by centering and scaling
Parameters
X [array-like, shape [n_samples, n_features]] The data used to scale along the features axis.
Examples using sklearn.preprocessing.StandardScaler
preprocessing.add_dummy_feature(X[, value]) Augment dataset with an additional dummy feature.



preprocessing.binarize(X[, threshold, copy]) Boolean thresholding of array-like or scipy.sparse matrix
preprocessing.label_binarize(y, classes[, . . . ]) Binarize labels in a one-vs-all fashion
preprocessing.maxabs_scale(X[, axis, copy]) Scale each feature to the [-1, 1] range without breaking the
sparsity.
preprocessing.minmax_scale(X[, . . . ]) Transforms features by scaling each feature to a given
range.
preprocessing.normalize(X[, norm, axis, . . . ]) Scale input vectors individually to unit norm (vector
length).
preprocessing.quantile_transform(X[, axis, Transform features using quantiles information.
. . . ])
preprocessing.robust_scale(X[, axis, . . . ]) Standardize a dataset along any axis
preprocessing.scale(X[, axis, with_mean, . . . ]) Standardize a dataset along any axis
preprocessing.power_transform(X[, method, Power transforms are a family of parametric, mono-
. . . ]) tonic transformations that are applied to make data more
Gaussian-like.
6.32.18 sklearn.preprocessing.add_dummy_feature
sklearn.preprocessing.add_dummy_feature(X, value=1.0)
Augment dataset with an additional dummy feature.
This is useful for fitting an intercept term with implementations which cannot otherwise fit it directly.
Parameters
X [{array-like, sparse matrix}, shape [n_samples, n_features]] Data.
value [float] Value to use for the dummy feature.
Returns
X [{array, sparse matrix}, shape [n_samples, n_features + 1]] Same data with dummy feature
added as first column.
Examples
>>> from sklearn.preprocessing import add_dummy_feature

>>> add_dummy_feature([[0, 1], [1, 0]])
array([[1., 0., 1.],
[1., 1., 0.]])
6.32.19 sklearn.preprocessing.binarize
sklearn.preprocessing.binarize(X, threshold=0.0, copy=True)

Boolean thresholding of array-like or scipy.sparse matrix
Parameters
X [{array-like, sparse matrix}, shape [n_samples, n_features]] The data to binarize, element by
element. scipy.sparse matrices should be in CSR or CSC format to avoid an un-necessary
copy.

threshold [float, optional (0.0 by default)] Feature values below or equal to this are replaced by
0, above it by 1. Threshold may not be less than 0 for operations on sparse matrices.
copy [boolean, optional, default True] set to False to perform inplace binarization and avoid a
copy (if the input is already a numpy array or a scipy.sparse CSR / CSC matrix and if axis
is 1).
See also:
Binarizer Performs binarization using the Transformer API (e.g. as part of a preprocessing sklearn.
pipeline.Pipeline).
6.32.20 sklearn.preprocessing.label_binarize
sklearn.preprocessing.label_binarize(y, classes, neg_label=0, pos_label=1,

sparse_output=False)
Binarize labels in a one-vs-all fashion
Several regression and binary classification algorithms are available in scikit-learn. A simple way to extend
these algorithms to the multi-class classification case is to use the so-called one-vs-all scheme.
This function makes it possible to compute this transformation for a fixed set of class labels known ahead of
time.
Parameters
y [array-like] Sequence of integer labels or multilabel data to encode.
classes [array-like of shape [n_classes]] Uniquely holds the label for each class.
neg_label [int (default: 0)] Value with which negative labels must be encoded.
pos_label [int (default: 1)] Value with which positive labels must be encoded.
sparse_output [boolean (default: False),] Set to true if output binary array is desired in CSR
sparse format
Returns
Y [numpy array or CSR matrix of shape [n_samples, n_classes]] Shape will be [n_samples, 1]
for binary problems.
See also:
LabelBinarizer class used to wrap the functionality of label_binarize and allow for fitting to classes inde-
pendently of the transform operation
Examples
>>> from sklearn.preprocessing import label_binarize

>>> label_binarize([1, 6], classes=[1, 2, 4, 6])
array([[1, 0, 0, 0],
[0, 0, 0, 1]])
The class ordering is preserved:
>>> label_binarize([1, 6], classes=[1, 6, 4, 2])

array([[1, 0, 0, 0],
[0, 1, 0, 0]])

Binary targets transform to a column vector
>>> label_binarize(['yes', 'no', 'no', 'yes'], classes=['no', 'yes'])

array([[1],
[0],
[0],
[1]])
Examples using sklearn.preprocessing.label_binarize

6.32.21 sklearn.preprocessing.maxabs_scale
sklearn.preprocessing.maxabs_scale(X, axis=0, copy=True)

Scale each feature to the [-1, 1] range without breaking the sparsity.
This estimator scales each feature individually such that the maximal absolute value of each feature in the
training set will be 1.0.
This scaler can also be applied to sparse CSR or CSC matrices.
Parameters
axis [int (0 by default)] axis used to scale along. If 0, independently scale each feature, other-
wise (if 1) scale each sample.
See also:
MaxAbsScaler Performs scaling to the [-1, 1] range using the‘‘Transformer‘‘ API (e.g. as part of a prepro-
cessing sklearn.pipeline.Pipeline).
Notes
NaNs are treated as missing values: disregarded to compute the statistics, and maintained during the data trans-
formation.
6.32.22 sklearn.preprocessing.minmax_scale
sklearn.preprocessing.minmax_scale(X, feature_range=(0, 1), axis=0, copy=True)

Transforms features by scaling each feature to a given range.
This estimator scales and translates each feature individually such that it is in the given range on the training set,
i.e. between zero and one.
The transformation is given by (when axis=0):


where min, max = feature_range.

The transformation is calculated as (when axis=0):
X_scaled = scale * X + min - X.min(axis=0) * scale

where scale = (max - min) / (X.max(axis=0) - X.min(axis=0))
This transformation is often used as an alternative to zero mean, unit variance scaling.
New in version 0.17: minmax_scale function interface to sklearn.preprocessing.MinMaxScaler.
Parameters
feature_range [tuple (min, max), default=(0, 1)] Desired range of transformed data.
axis [int (0 by default)] axis used to scale along. If 0, independently scale each feature, other-
wise (if 1) scale each sample.
See also:
MinMaxScaler Performs scaling to a given range using the‘‘Transformer‘‘ API (e.g. as part of a preprocess-
ing sklearn.pipeline.Pipeline).
Notes
Examples using sklearn.preprocessing.minmax_scale
6.32.23 sklearn.preprocessing.normalize
sklearn.preprocessing.normalize(X, norm=’l2’, axis=1, copy=True, return_norm=False)

Scale input vectors individually to unit norm (vector length).
Parameters
X [{array-like, sparse matrix}, shape [n_samples, n_features]] The data to normalize, element
by element. scipy.sparse matrices should be in CSR format to avoid an un-necessary copy.
norm [‘l1’, ‘l2’, or ‘max’, optional (‘l2’ by default)] The norm to use to normalize each non
zero sample (or each non-zero feature if axis is 0).

axis [0 or 1, optional (1 by default)] axis used to normalize the data along. If 1, independently
normalize each sample, otherwise (if 0) normalize each feature.
avoid a copy (if the input is already a numpy array or a scipy.sparse CSR matrix and if axis
is 1).
return_norm [boolean, default False] whether to return the computed norms
Returns
X [{array-like, sparse matrix}, shape [n_samples, n_features]] Normalized input X.
norms [array, shape [n_samples] if axis=1 else [n_features]] An array of norms along given
axis for X. When X is sparse, a NotImplementedError will be raised for norm ‘l1’ or ‘l2’.
See also:
Normalizer Performs normalization using the Transformer API (e.g. as part of a preprocessing
sklearn.pipeline.Pipeline).
Notes
6.32.24 sklearn.preprocessing.quantile_transform
sklearn.preprocessing.quantile_transform(X, axis=0, n_quantiles=1000, out-

put_distribution=’uniform’, ig-
nore_implicit_zeros=False, subsample=100000,
random_state=None, copy=False)
Transform features using quantiles information.
This method transforms the features to follow a uniform or a normal distribution. Therefore, for a given feature,
this transformation tends to spread out the most frequent values. It also reduces the impact of (marginal) outliers:
this is therefore a robust preprocessing scheme.
The transformation is applied on each feature independently. The cumulative distribution function of a feature
is used to project the original values. Features values of new/unseen data that fall below or above the fitted range
will be mapped to the bounds of the output distribution. Note that this transform is non-linear. It may distort
linear correlations between variables measured at the same scale but renders variables measured at different
scales more directly comparable.
Parameters
X [array-like, sparse matrix] The data to transform.
axis [int, (default=0)] Axis used to compute the means and standard deviations along. If 0,
transform each feature, otherwise (if 1) transform each sample.
n_quantiles [int, optional (default=1000)] Number of quantiles to be computed. It corresponds
to the number of landmarks used to discretize the cumulative distribution function.
output_distribution [str, optional (default=’uniform’)] Marginal distribution for the trans-
formed data. The choices are ‘uniform’ (default) or ‘normal’.

ignore_implicit_zeros [bool, optional (default=False)] Only applies to sparse matrices. If True,

the sparse entries of the matrix are discarded to compute the quantile statistics. If False,
these entries are treated as zeros.
subsample [int, optional (default=1e5)] Maximum number of samples used to estimate the
quantiles for computational efficiency. Note that the subsampling procedure may differ
for value-identical sparse and dense matrices.
the RandomState instance used by np.random. Note that this is used by subsampling and
smoothing noise.
copy [boolean, optional, (default=True)] Set to False to perform inplace transformation and
Attributes
quantiles_ [ndarray, shape (n_quantiles, n_features)] The values corresponding the quantiles of
reference.
references_ [ndarray, shape(n_quantiles, )] Quantiles of references.
See also:
QuantileTransformer Performs quantile-based scaling using the Transformer API (e.g. as part of a
preprocessing sklearn.pipeline.Pipeline).
power_transform Maps data to a normal distribution using a power transformation.
scale Performs standardization that is faster, but less robust to outliers.
robust_scale Performs robust standardization that removes the influence of outliers but does not put out-
liers and inliers on the same scale.
Notes
Examples

>>> from sklearn.preprocessing import quantile_transform
>>> X = np.sort(rng.normal(loc=0.5, scale=0.25, size=(25, 1)), axis=0)
>>> quantile_transform(X, n_quantiles=10, random_state=0)
...
array([...])
Examples using sklearn.preprocessing.quantile_transform

6.32.25 sklearn.preprocessing.robust_scale
sklearn.preprocessing.robust_scale(X, axis=0, with_centering=True, with_scaling=True,

quantile_range=(25.0, 75.0), copy=True)
Standardize a dataset along any axis
Center to the median and component wise scale according to the interquartile range.
Parameters
X [array-like] The data to center and scale.
axis [int (0 by default)] axis used to compute the medians and IQR along. If 0, independently
scale each feature, otherwise (if 1) scale each sample.
with_centering [boolean, True by default] If True, center the data before scaling.
with_scaling [boolean, True by default] If True, scale the data to unit variance (or equivalently,
unit standard deviation).
quantile_range [tuple (q_min, q_max), 0.0 < q_min < q_max < 100.0] Default: (25.0, 75.0) =
(1st quantile, 3rd quantile) = IQR Quantile range used to calculate scale_.
copy [boolean, optional, default is True] set to False to perform inplace row normalization and
avoid a copy (if the input is already a numpy array or a scipy.sparse CSR matrix and if axis
is 1).
See also:
RobustScaler Performs centering and scaling using the Transformer API (e.g. as part of a preprocess-
Notes
This implementation will refuse to center scipy.sparse matrices since it would make them non-sparse and would
potentially crash the program with memory exhaustion problems.
Instead the caller is expected to either set explicitly with_centering=False (in that case, only variance scaling
will be performed on the features of the CSR matrix) or to call X.toarray() if he/she expects the materialized
dense array to fit in memory.
To avoid memory copy the caller should pass a CSR matrix.
6.32.26 sklearn.preprocessing.scale
sklearn.preprocessing.scale(X, axis=0, with_mean=True, with_std=True, copy=True)

Standardize a dataset along any axis
Center to the mean and component wise scale to unit variance.
Parameters

X [{array-like, sparse matrix}] The data to center and scale.

axis [int (0 by default)] axis used to compute the means and standard deviations along. If 0,
independently standardize each feature, otherwise (if 1) standardize each sample.
with_mean [boolean, True by default] If True, center the data before scaling.
with_std [boolean, True by default] If True, scale the data to unit variance (or equivalently, unit
standard deviation).
avoid a copy (if the input is already a numpy array or a scipy.sparse CSC matrix and if axis
is 1).
See also:
StandardScaler Performs scaling to unit variance using the‘‘Transformer‘‘ API (e.g. as part of a prepro-
cessing sklearn.pipeline.Pipeline).
Notes
This implementation will refuse to center scipy.sparse matrices since it would make them non-sparse and would
potentially crash the program with memory exhaustion problems.
Instead the caller is expected to either set explicitly with_mean=False (in that case, only variance scaling will
be performed on the features of the CSC matrix) or to call X.toarray() if he/she expects the materialized dense
array to fit in memory.
To avoid memory copy the caller should pass a CSC matrix.
NaNs are treated as missing values: disregarded to compute the statistics, and maintained during the data trans-
formation.
We use a biased estimator for the standard deviation, equivalent to numpy.std(x, ddof=0). Note that the choice
of ddof is unlikely to affect model performance.
Examples using sklearn.preprocessing.scale
6.32.27 sklearn.preprocessing.power_transform
sklearn.preprocessing.power_transform(X, method=’warn’, standardize=True, copy=True)

Power transforms are a family of parametric, monotonic transformations that are applied to make data more
Gaussian-like. This is useful for modeling issues related to heteroscedasticity (non-constant variance), or other
situations where normality is desired.
Currently, power_transform supports the Box-Cox transform and the Yeo-Johnson transform. The optimal pa-
rameter for stabilizing variance and minimizing skewness is estimated through maximum likelihood.
Box-Cox requires input data to be strictly positive, while Yeo-Johnson supports both positive or negative data.
By default, zero-mean, unit-variance normalization is applied to the transformed data.

Parameters
X [array-like, shape (n_samples, n_features)] The data to be transformed using a power trans-
formation.
method [str] The power transform method. Available methods are:
• ‘yeo-johnson’ [1], works with positive and negative values
• ‘box-cox’ [2], only works with strictly positive values
The default method will be changed from ‘box-cox’ to ‘yeo-johnson’ in version 0.23. To
suppress the FutureWarning, explicitly set the parameter.
standardize [boolean, default=True] Set to True to apply zero-mean, unit-variance normaliza-
tion to the transformed output.
copy [boolean, optional, default=True] Set to False to perform inplace computation during
transformation.
Returns
X_trans [array-like, shape (n_samples, n_features)] The transformed data.
See also:
PowerTransformer Equivalent transformation with the Transformer API (e.g. as part of a preprocess-
quantile_transform Maps data to a standard normal distribution with the parameter out-
put_distribution=’normal’.
Notes
References
[1], [2]
Examples

>>> from sklearn.preprocessing import power_transform
>>> data = [[1, 2], [3, 2], [4, 5]]
>>> print(power_transform(data, method='box-cox'))
[[-1.332... -0.707...]
[ 0.256... -0.707...]
[ 1.076... 1.414...]]

6.33 sklearn.random_projection: Random projection
Random Projection transformers

Random Projections are a simple and computationally efficient way to reduce the dimensionality of the data by trading
a controlled amount of accuracy (as additional variance) for faster processing times and smaller model sizes.
The dimensions and distribution of Random Projections matrices are controlled so as to preserve the pairwise distances
between any two samples of the dataset.
The main theoretical result behind the efficiency of random projection is the Johnson-Lindenstrauss lemma (quoting
Wikipedia):
In mathematics, the Johnson-Lindenstrauss lemma is a result concerning low-distortion embeddings of
points from high-dimensional into low-dimensional Euclidean space. The lemma states that a small set
of points in a high-dimensional space can be embedded into a space of much lower dimension in such a
way that distances between the points are nearly preserved. The map used for the embedding is at least
Lipschitz, and can even be taken to be an orthogonal projection.
User guide: See the Random Projection section for further details.
random_projection. Reduce dimensionality through Gaussian random projec-

GaussianRandomProjection([. . . ]) tion
random_projection. Reduce dimensionality through sparse random projection
SparseRandomProjection([. . . ])
6.33.1 sklearn.random_projection.GaussianRandomProjection
class sklearn.random_projection.GaussianRandomProjection(n_components=’auto’,
eps=0.1, ran-
dom_state=None)
Reduce dimensionality through Gaussian random projection
The components of the random matrix are drawn from N(0, 1 / n_components).
Parameters
n_components [int or ‘auto’, optional (default = ‘auto’)] Dimensionality of the target projection
space.
n_components can be automatically adjusted according to the number of samples in the
dataset and the bound given by the Johnson-Lindenstrauss lemma. In that case the quality
of the embedding is controlled by the eps parameter.
It should be noted that Johnson-Lindenstrauss lemma can yield very conservative estimated
of the required number of components as it makes no assumption on the structure of the
dataset.
eps [strictly positive float, optional (default=0.1)] Parameter to control the quality of the embed-
ding according to the Johnson-Lindenstrauss lemma when n_components is set to ‘auto’.
Smaller values lead to better embedding and higher number of dimensions (n_components)
in the target projection space.
random_state [int, RandomState instance or None, optional (default=None)] Control the
pseudo random number generator used to generate the matrix at fit time. If int, random_state
6.33. sklearn.random_projection: Random projection 2171

Attributes
n_component_ [int] Concrete number of components computed when n_components=”auto”.
components_ [numpy array of shape [n_components, n_features]] Random matrix used for the
projection.
See also:
SparseRandomProjection
Examples

>>> from sklearn.random_projection import GaussianRandomProjection
>>> X = np.random.rand(100, 10000)
>>> transformer = GaussianRandomProjection()
>>> X_new.shape
(100, 3947)
Methods
fit(X[, y]) Generate a sparse random projection matrix

transform(X) Project the data by using matrix product with the ran-
dom matrix
__init__(n_components=’auto’, eps=0.1, random_state=None)

fit(X, y=None)
Generate a sparse random projection matrix
Parameters
X [numpy array or scipy.sparse of shape [n_samples, n_features]] Training set: only the
shape is used to find optimal random matrix dimensions based on the theory referenced in
the afore mentioned papers.
y Ignored
Returns
self
Parameters


Returns
Parameters
Returns
of a nested object.
Returns
self
transform(X)
Project the data by using matrix product with the random matrix
Parameters
X [numpy array or scipy.sparse of shape [n_samples, n_features]] The input data to project
into a smaller dimensional space.
Returns
X_new [numpy array or scipy sparse of shape [n_samples, n_components]] Projected array.
6.33.2 sklearn.random_projection.SparseRandomProjection
class sklearn.random_projection.SparseRandomProjection(n_components=’auto’,
density=’auto’, eps=0.1,
dense_output=False, ran-
dom_state=None)
Reduce dimensionality through sparse random projection
Sparse random matrix is an alternative to dense random projection matrix that guarantees similar embedding
quality while being much more memory efficient and allowing faster computation of the projected data.
If we note s = 1 / density the components of the random matrix are drawn from:
• -sqrt(s) / sqrt(n_components) with probability 1 / 2s
• 0 with probability 1 - 1 / s
• +sqrt(s) / sqrt(n_components) with probability 1 / 2s
Parameters

n_components [int or ‘auto’, optional (default = ‘auto’)] Dimensionality of the target projection
space.
n_components can be automatically adjusted according to the number of samples in the
dataset and the bound given by the Johnson-Lindenstrauss lemma. In that case the quality
of the embedding is controlled by the eps parameter.
It should be noted that Johnson-Lindenstrauss lemma can yield very conservative estimated
of the required number of components as it makes no assumption on the structure of the
dataset.
density [float in range ]0, 1], optional (default=’auto’)] Ratio of non-zero component in the
random projection matrix.
If density = ‘auto’, the value is set to the minimum density as recommended by Ping Li et
al.: 1 / sqrt(n_features).
Use density = 1 / 3.0 if you want to reproduce the results from Achlioptas, 2001.
eps [strictly positive float, optional, (default=0.1)] Parameter to control the quality of the em-
bedding according to the Johnson-Lindenstrauss lemma when n_components is set to ‘auto’.
Smaller values lead to better embedding and higher number of dimensions (n_components)
in the target projection space.
dense_output [boolean, optional (default=False)] If True, ensure that the output of the random
projection is a dense numpy array even if the input and random projection matrix are both
sparse. In practice, if the number of components is small the number of zero components in
the projected data will be very small and it will be more CPU and memory efficient to use a
dense representation.
If False, the projected data uses a sparse representation if the input is sparse.
random_state [int, RandomState instance or None, optional (default=None)] Control the
pseudo random number generator used to generate the matrix at fit time. If int, random_state
Attributes
n_component_ [int] Concrete number of components computed when n_components=”auto”.
components_ [CSR matrix with shape [n_components, n_features]] Random matrix used for
the projection.
density_ [float in range 0.0 - 1.0] Concrete density computed from when density = “auto”.
See also:
GaussianRandomProjection
References
[1], [2]

Examples

>>> from sklearn.random_projection import SparseRandomProjection
>>> X = np.random.rand(100, 10000)
>>> transformer = SparseRandomProjection()
>>> X_new.shape
(100, 3947)
>>> # very few components are non-zero
>>> np.mean(transformer.components_ != 0)
0.0100...
Methods
fit(X[, y]) Generate a sparse random projection matrix

transform(X) Project the data by using matrix product with the ran-
dom matrix
__init__(n_components=’auto’, density=’auto’, eps=0.1, dense_output=False, ran-

dom_state=None)
fit(X, y=None)
Generate a sparse random projection matrix
Parameters
X [numpy array or scipy.sparse of shape [n_samples, n_features]] Training set: only the
shape is used to find optimal random matrix dimensions based on the theory referenced in
the afore mentioned papers.
y Ignored
Returns
self
Parameters
Returns

Parameters
Returns
of a nested object.
Returns
self
transform(X)
Project the data by using matrix product with the random matrix
Parameters
X [numpy array or scipy.sparse of shape [n_samples, n_features]] The input data to project
into a smaller dimensional space.
Returns
X_new [numpy array or scipy sparse of shape [n_samples, n_components]] Projected array.
Examples using sklearn.random_projection.SparseRandomProjection

random_projection. Find a ‘safe’ number of components to randomly project to

johnson_lindenstrauss_min_dim(. . . )
6.33.3 sklearn.random_projection.johnson_lindenstrauss_min_dim
sklearn.random_projection.johnson_lindenstrauss_min_dim(n_samples, eps=0.1)
Find a ‘safe’ number of components to randomly project to
The distortion introduced by a random projection p only changes the distance between two points by a factor (1
+- eps) in an euclidean space with good probability. The projection p is an eps-embedding as defined by:
(1 - eps) ||u - v||^2 < ||p(u) - p(v)||^2 < (1 + eps) ||u - v||^2
Where u and v are any rows taken from a dataset of shape [n_samples, n_features], eps is in ]0, 1[ and p is a
projection by a random Gaussian N(0, 1) matrix with shape [n_components, n_features] (or a sparse Achlioptas
matrix).
The minimum number of components to guarantee the eps-embedding is given by:
n_components >= 4 log(n_samples) / (eps^2 / 2 - eps^3 / 3)
Note that the number of dimensions is independent of the original number of features but instead depends on
the size of the dataset: the larger the dataset, the higher is the minimal dimensionality of an eps-embedding.


Parameters
n_samples [int or numpy array of int greater than 0,] Number of samples. If an array is given,
it will compute a safe number of components array-wise.
eps [float or numpy array of float in ]0,1[, optional (default=0.1)] Maximum distortion rate as
defined by the Johnson-Lindenstrauss lemma. If an array is given, it will compute a safe
number of components array-wise.
Returns
n_components [int or numpy array of int,] The minimal number of components to guarantee
with good probability an eps-embedding with n_samples.
References
[1], [2]
Examples
>>> johnson_lindenstrauss_min_dim(1e6, eps=0.5)

663
>>> johnson_lindenstrauss_min_dim(1e6, eps=[0.5, 0.1, 0.01])

array([ 663, 11841, 1112658])
>>> johnson_lindenstrauss_min_dim([1e4, 1e5, 1e6], eps=0.1)

array([ 7894, 9868, 11841])
Examples using sklearn.random_projection.johnson_lindenstrauss_min_dim
6.34 sklearn.semi_supervised Semi-Supervised Learning
The sklearn.semi_supervised module implements semi-supervised learning algorithms. These algorithms

utilized small amounts of labeled data and large amounts of unlabeled data for classification tasks. This module
includes Label Propagation.
User guide: See the Semi-Supervised section for further details.
semi_supervised.LabelPropagation([kernel, Label Propagation classifier

. . . ])
semi_supervised.LabelSpreading([kernel, LabelSpreading model for semi-supervised learning
. . . ])
6.34. sklearn.semi_supervised Semi-Supervised Learning 2177

6.34.1 sklearn.semi_supervised.LabelPropagation
class sklearn.semi_supervised.LabelPropagation(kernel=’rbf’, gamma=20, n_neighbors=7,

alpha=None, max_iter=1000, tol=0.001,
n_jobs=None)
Label Propagation classifier
Parameters
kernel [{‘knn’, ‘rbf’, callable}] String identifier for kernel function to use or the kernel function
itself. Only ‘rbf’ and ‘knn’ strings are valid inputs. The function passed should take two
inputs, each of shape [n_samples, n_features], and return a [n_samples, n_samples] shaped
weight matrix.
gamma [float] Parameter for rbf kernel
n_neighbors [integer > 0] Parameter for knn kernel
alpha [float] Clamping factor.
Deprecated since version 0.19: This parameter will be removed in 0.21. ‘alpha’ is fixed to
zero in ‘LabelPropagation’.
max_iter [integer] Change maximum number of iterations allowed
tol [float] Convergence tolerance: threshold to consider the system at steady state
Attributes
X_ [array, shape = [n_samples, n_features]] Input array.
classes_ [array, shape = [n_classes]] The distinct labels used in classifying instances.
label_distributions_ [array, shape = [n_samples, n_classes]] Categorical distribution for each
item.
transduction_ [array, shape = [n_samples]] Label assigned to each item via the transduction.
See also:
LabelSpreading Alternate label propagation strategy more robust to noise
References
Xiaojin Zhu and Zoubin Ghahramani. Learning from labeled and unlabeled data with label propagation. Tech-
nical Report CMU-CALD-02-107, Carnegie Mellon University, 2002 http://pages.cs.wisc.edu/~jerryzhu/pub/
CMU-CALD-02-107.pdf
Examples


>>> from sklearn.semi_supervised import LabelPropagation
>>> label_prop_model = LabelPropagation()
>>> random_unlabeled_points = rng.rand(len(iris.target)) < 0.3
>>> labels = np.copy(iris.target)
>>> labels[random_unlabeled_points] = -1
>>> label_prop_model.fit(iris.data, labels)
...
LabelPropagation(...)
Methods
fit(X, y)
predict(X) Performs inductive inference across the model.
predict_proba(X) Predict probability for each possible outcome.
labels.
__init__(kernel=’rbf’, gamma=20, n_neighbors=7, alpha=None, max_iter=1000, tol=0.001,

n_jobs=None)
Parameters
Returns
predict(X)
Performs inductive inference across the model.
Parameters
X [array_like, shape = [n_samples, n_features]]
Returns
y [array_like, shape = [n_samples]] Predictions for input data
predict_proba(X)
Predict probability for each possible outcome.
Compute the probability estimates for each single sample in X and each possible outcome seen during
training (categorical distribution).
Parameters

Returns
probabilities [array, shape = [n_samples, n_classes]] Normalized probability distributions
across class labels
Parameters
Returns
of a nested object.
Returns
self
6.34.2 sklearn.semi_supervised.LabelSpreading
class sklearn.semi_supervised.LabelSpreading(kernel=’rbf’, gamma=20, n_neighbors=7,

alpha=0.2, max_iter=30, tol=0.001,
n_jobs=None)
LabelSpreading model for semi-supervised learning
This model is similar to the basic Label Propagation algorithm, but uses affinity matrix based on the normalized
graph Laplacian and soft clamping across the labels.
Parameters
kernel [{‘knn’, ‘rbf’, callable}] String identifier for kernel function to use or the kernel function
itself. Only ‘rbf’ and ‘knn’ strings are valid inputs. The function passed should take two
inputs, each of shape [n_samples, n_features], and return a [n_samples, n_samples] shaped
weight matrix
gamma [float] parameter for rbf kernel
n_neighbors [integer > 0] parameter for knn kernel
alpha [float] Clamping factor. A value in (0, 1) that specifies the relative amount that an instance
should adopt the information from its neighbors as opposed to its initial label. alpha=0
means keeping the initial label information; alpha=1 means replacing all initial information.
max_iter [integer] maximum number of iterations allowed
tol [float] Convergence tolerance: threshold to consider the system at steady state

Attributes
X_ [array, shape = [n_samples, n_features]] Input array.
classes_ [array, shape = [n_classes]] The distinct labels used in classifying instances.
label_distributions_ [array, shape = [n_samples, n_classes]] Categorical distribution for each
item.
transduction_ [array, shape = [n_samples]] Label assigned to each item via the transduction.
See also:
LabelPropagation Unregularized graph based semi-supervised learning
References
Dengyong Zhou, Olivier Bousquet, Thomas Navin Lal, Jason Weston, Bernhard Schoelkopf. Learning with
local and global consistency (2004) http://citeseer.ist.psu.edu/viewdoc/summary?doi=10.1.1.115.3219
Examples

>>> from sklearn.semi_supervised import LabelSpreading
>>> label_prop_model = LabelSpreading()
>>> random_unlabeled_points = rng.rand(len(iris.target)) < 0.3
>>> labels = np.copy(iris.target)
>>> labels[random_unlabeled_points] = -1
>>> label_prop_model.fit(iris.data, labels)
...
LabelSpreading(...)
Methods
fit(X, y) Fit a semi-supervised label propagation model based

predict(X) Performs inductive inference across the model.
predict_proba(X) Predict probability for each possible outcome.
labels.
__init__(kernel=’rbf’, gamma=20, n_neighbors=7, alpha=0.2, max_iter=30, tol=0.001,

n_jobs=None)

fit(X, y)
Fit a semi-supervised label propagation model based
All the input data is provided matrix X (labeled and unlabeled) and corresponding label matrix y with a
dedicated marker value for unlabeled samples.
Parameters
X [array-like, shape = [n_samples, n_features]] A {n_samples by n_samples} size matrix
will be created from this
y [array_like, shape = [n_samples]] n_labeled_samples (unlabeled points are marked as -1)
All unlabeled samples will be transductively assigned labels
Returns
Parameters
Returns
predict(X)
Performs inductive inference across the model.
Parameters
Returns
y [array_like, shape = [n_samples]] Predictions for input data
predict_proba(X)
Predict probability for each possible outcome.
Compute the probability estimates for each single sample in X and each possible outcome seen during
training (categorical distribution).
Parameters
Returns
probabilities [array, shape = [n_samples, n_classes]] Normalized probability distributions
across class labels
Parameters


Returns
of a nested object.
Returns
self
Examples using sklearn.semi_supervised.LabelSpreading

6.35 sklearn.svm: Support Vector Machines
The sklearn.svm module includes Support Vector Machine algorithms.

User guide: See the Support Vector Machines section for further details.
6.35.1 Estimators
svm.LinearSVC([penalty, loss, dual, tol, C, . . . ]) Linear Support Vector Classification.

svm.LinearSVR([epsilon, tol, C, loss, . . . ]) Linear Support Vector Regression.
svm.NuSVC([nu, kernel, degree, gamma, . . . ]) Nu-Support Vector Classification.
svm.NuSVR([nu, C, kernel, degree, gamma, . . . ]) Nu Support Vector Regression.
svm.OneClassSVM ([kernel, degree, gamma, . . . ]) Unsupervised Outlier Detection.
svm.SVC([C, kernel, degree, gamma, coef0, . . . ]) C-Support Vector Classification.
svm.SVR([kernel, degree, gamma, coef0, tol, . . . ]) Epsilon-Support Vector Regression.
sklearn.svm.LinearSVC
class sklearn.svm.LinearSVC(penalty=’l2’, loss=’squared_hinge’, dual=True, tol=0.0001,

C=1.0, multi_class=’ovr’, fit_intercept=True, intercept_scaling=1,
class_weight=None, verbose=0, random_state=None, max_iter=1000)
Linear Support Vector Classification.
Similar to SVC with parameter kernel=’linear’, but implemented in terms of liblinear rather than libsvm, so it
has more flexibility in the choice of penalties and loss functions and should scale better to large numbers of
samples.
This class supports both dense and sparse input and the multiclass support is handled according to a one-vs-the-
rest scheme.
6.35. sklearn.svm: Support Vector Machines 2183


Parameters
penalty [string, ‘l1’ or ‘l2’ (default=’l2’)] Specifies the norm used in the penalization. The ‘l2’
penalty is the standard used in SVC. The ‘l1’ leads to coef_ vectors that are sparse.
loss [string, ‘hinge’ or ‘squared_hinge’ (default=’squared_hinge’)] Specifies the loss function.
‘hinge’ is the standard SVM loss (used e.g. by the SVC class) while ‘squared_hinge’ is the
square of the hinge loss.
dual [bool, (default=True)] Select the algorithm to either solve the dual or primal optimization
problem. Prefer dual=False when n_samples > n_features.
tol [float, optional (default=1e-4)] Tolerance for stopping criteria.
C [float, optional (default=1.0)] Penalty parameter C of the error term.
multi_class [string, ‘ovr’ or ‘crammer_singer’ (default=’ovr’)] Determines the multi-class
strategy if y contains more than two classes. "ovr" trains n_classes one-vs-rest clas-
sifiers, while "crammer_singer" optimizes a joint objective over all classes. While
crammer_singer is interesting from a theoretical perspective as it is consistent, it is seldom
used in practice as it rarely leads to better accuracy and is more expensive to compute. If
"crammer_singer" is chosen, the options loss, penalty and dual will be ignored.
fit_intercept [boolean, optional (default=True)] Whether to calculate the intercept for this
model. If set to false, no intercept will be used in calculations (i.e. data is expected to
be already centered).
intercept_scaling [float, optional (default=1)] When self.fit_intercept is True, instance vector
x becomes [x, self.intercept_scaling], i.e. a “synthetic” feature with constant
value equals to intercept_scaling is appended to the instance vector. The intercept becomes
intercept_scaling * synthetic feature weight Note! the synthetic feature weight is subject to
l1/l2 regularization as all other features. To lessen the effect of regularization on synthetic
feature weight (and therefore on the intercept) intercept_scaling has to be increased.
class_weight [{dict, ‘balanced’}, optional] Set the parameter C of class i to
class_weight[i]*C for SVC. If not given, all classes are supposed to have weight
one. The “balanced” mode uses the values of y to automatically adjust weights inversely
proportional to class frequencies in the input data as n_samples / (n_classes *
np.bincount(y))
verbose [int, (default=0)] Enable verbose output. Note that this setting takes advantage of a per-
process runtime setting in liblinear that, if enabled, may not work properly in a multithreaded
context.
pseudo random number generator to use when shuffling the data for the dual coordi-
nate descent (if dual=True). When dual=False the underlying implementation of
LinearSVC is not random and random_state has no effect on the results. If int, ran-
max_iter [int, (default=1000)] The maximum number of iterations to be run.
Attributes
coef_ [array, shape = [n_features] if n_classes == 2 else [n_classes, n_features]] Weights as-
signed to the features (coefficients in the primal problem). This is only available in the case
of a linear kernel.

coef_ is a readonly property derived from raw_coef_ that follows the internal memory
layout of liblinear.
See also:
SVC Implementation of Support Vector Machine classifier using libsvm: the kernel can be non-linear but its
SMO algorithm does not scale to large number of samples as LinearSVC does. Furthermore SVC multi-
class mode is implemented using one vs one scheme while LinearSVC uses one vs the rest. It is possible to
implement one vs the rest with SVC by using the sklearn.multiclass.OneVsRestClassifier
wrapper. Finally SVC can fit dense data without memory copy if the input is C-contiguous. Sparse data
will still incur memory copy though.
sklearn.linear_model.SGDClassifier SGDClassifier can optimize the same cost function as Lin-
earSVC by adjusting the penalty and loss parameters. In addition it requires less memory, allows incre-
mental (online) learning, and implements various loss functions and regularization regimes.
Notes
The underlying C implementation uses a random number generator to select features when fitting the model.
It is thus not uncommon to have slightly different results for the same input data. If that happens, try with a
smaller tol parameter.
The underlying implementation, liblinear, uses a sparse internal representation for the data that will incur a
memory copy.
Predict output may not match that of standalone liblinear in certain cases. See differences from liblinear in the
narrative documentation.
References
LIBLINEAR: A Library for Large Linear Classification
Examples

>>> X, y = make_classification(n_features=4, random_state=0)
>>> clf = LinearSVC(random_state=0, tol=1e-5)
>>> clf.fit(X, y)
multi_class='ovr', penalty='l2', random_state=0, tol=1e-05, verbose=0)
[[0.085... 0.394... 0.498... 0.375...]]
[0.284...]
[1]

Methods

labels.
__init__(penalty=’l2’, loss=’squared_hinge’, dual=True, tol=0.0001, C=1.0, multi_class=’ovr’,

fit_intercept=True, intercept_scaling=1, class_weight=None, verbose=0, ran-
dom_state=None, max_iter=1000)
Parameters
Returns
densify()
Returns
self [estimator]
Parameters
y [array-like, shape = [n_samples]] Target vector relative to X
sample_weight [array-like, shape = [n_samples], optional] Array of weights that are as-
signed to individual samples. If not provided, then each sample is given unit weight.
Returns
self [object]

Parameters
Returns
predict(X)
Parameters
Returns
Parameters
Returns
of a nested object.
Returns
self
sparsify()
Returns
self [estimator]

Notes
densify.
Examples using sklearn.svm.LinearSVC

sklearn.svm.LinearSVR
class sklearn.svm.LinearSVR(epsilon=0.0, tol=0.0001, C=1.0, loss=’epsilon_insensitive’,

fit_intercept=True, intercept_scaling=1.0, dual=True, verbose=0,
random_state=None, max_iter=1000)
Linear Support Vector Regression.
Similar to SVR with parameter kernel=’linear’, but implemented in terms of liblinear rather than libsvm, so it
has more flexibility in the choice of penalties and loss functions and should scale better to large numbers of
samples.
This class supports both dense and sparse input.
Parameters
epsilon [float, optional (default=0.1)] Epsilon parameter in the epsilon-insensitive loss function.
Note that the value of this parameter depends on the scale of the target variable y. If unsure,
set epsilon=0.
tol [float, optional (default=1e-4)] Tolerance for stopping criteria.
C [float, optional (default=1.0)] Penalty parameter C of the error term. The penalty is a squared
l2 penalty. The bigger this parameter, the less regularization is used.
loss [string, optional (default=’epsilon_insensitive’)] Specifies the loss function. The epsilon-
insensitive loss (standard SVR) is the L1 loss, while the squared epsilon-insensitive loss
(‘squared_epsilon_insensitive’) is the L2 loss.

fit_intercept [boolean, optional (default=True)] Whether to calculate the intercept for this
model. If set to false, no intercept will be used in calculations (i.e. data is expected to
be already centered).
intercept_scaling [float, optional (default=1)] When self.fit_intercept is True, instance vec-
tor x becomes [x, self.intercept_scaling], i.e. a “synthetic” feature with constant value
equals to intercept_scaling is appended to the instance vector. The intercept becomes in-
tercept_scaling * synthetic feature weight Note! the synthetic feature weight is subject to
l1/l2 regularization as all other features. To lessen the effect of regularization on synthetic
feature weight (and therefore on the intercept) intercept_scaling has to be increased.
dual [bool, (default=True)] Select the algorithm to either solve the dual or primal optimization
problem. Prefer dual=False when n_samples > n_features.
verbose [int, (default=0)] Enable verbose output. Note that this setting takes advantage of a per-
process runtime setting in liblinear that, if enabled, may not work properly in a multithreaded
context.
max_iter [int, (default=1000)] The maximum number of iterations to be run.
Attributes
coef_ [array, shape = [n_features] if n_classes == 2 else [n_classes, n_features]] Weights as-
signed to the features (coefficients in the primal problem). This is only available in the case
of a linear kernel.
coef_ is a readonly property derived from raw_coef_ that follows the internal memory layout
of liblinear.
See also:
LinearSVC Implementation of Support Vector Machine classifier using the same library as this class (liblin-
ear).
SVR Implementation of Support Vector Machine regression using libsvm: the kernel can be non-linear but its
SMO algorithm does not scale to large number of samples as LinearSVC does.
sklearn.linear_model.SGDRegressor SGDRegressor can optimize the same cost function as Lin-
earSVR by adjusting the penalty and loss parameters. In addition it requires less memory, allows incre-
mental (online) learning, and implements various loss functions and regularization regimes.
Examples
>>> from sklearn.svm import LinearSVR

>>> regr = LinearSVR(random_state=0, tol=1e-5)
>>> regr.fit(X, y)
LinearSVR(C=1.0, dual=True, epsilon=0.0, fit_intercept=True,
intercept_scaling=1.0, loss='epsilon_insensitive', max_iter=1000,
random_state=0, tol=1e-05, verbose=0)

[16.35... 26.91... 42.30... 60.47...]
[-4.29...]
[-4.29...]
Methods
diction.
__init__(epsilon=0.0, tol=0.0001, C=1.0, loss=’epsilon_insensitive’, fit_intercept=True, inter-

cept_scaling=1.0, dual=True, verbose=0, random_state=None, max_iter=1000)
Parameters
y [array-like, shape = [n_samples]] Target vector relative to X
sample_weight [array-like, shape = [n_samples], optional] Array of weights that are as-
signed to individual samples. If not provided, then each sample is given unit weight.
Returns
self [object]
Parameters
Returns
predict(X)
Parameters
Returns

Parameters
Returns
of a nested object.
Returns
self
sklearn.svm.NuSVC
class sklearn.svm.NuSVC(nu=0.5, kernel=’rbf’, degree=3, gamma=’auto_deprecated’,

coef0=0.0, shrinking=True, probability=False, tol=0.001,
cache_size=200, class_weight=None, verbose=False, max_iter=-1, deci-
sion_function_shape=’ovr’, random_state=None)
Nu-Support Vector Classification.
Similar to SVC but uses a parameter to control the number of support vectors.
The implementation is based on libsvm.
Parameters
nu [float, optional (default=0.5)] An upper bound on the fraction of training errors and a lower
bound of the fraction of support vectors. Should be in the interval (0, 1].
kernel [string, optional (default=’rbf’)] Specifies the kernel type to be used in the algorithm.
It must be one of ‘linear’, ‘poly’, ‘rbf’, ‘sigmoid’, ‘precomputed’ or a callable. If none is
given, ‘rbf’ will be used. If a callable is given it is used to precompute the kernel matrix.
degree [int, optional (default=3)] Degree of the polynomial kernel function (‘poly’). Ignored
by all other kernels.
gamma [float, optional (default=’auto’)] Kernel coefficient for ‘rbf’, ‘poly’ and ‘sigmoid’.
Current default is ‘auto’ which uses 1 / n_features, if gamma='scale' is passed then it
uses 1 / (n_features * X.var()) as value of gamma. The current default of gamma, ‘auto’,

will change to ‘scale’ in version 0.22. ‘auto_deprecated’, a deprecated version of ‘auto’ is

used as a default indicating that no explicit value of gamma was passed.
coef0 [float, optional (default=0.0)] Independent term in kernel function. It is only significant
in ‘poly’ and ‘sigmoid’.
shrinking [boolean, optional (default=True)] Whether to use the shrinking heuristic.
probability [boolean, optional (default=False)] Whether to enable probability estimates. This
must be enabled prior to calling fit, and will slow down that method.
tol [float, optional (default=1e-3)] Tolerance for stopping criterion.
cache_size [float, optional] Specify the size of the kernel cache (in MB).
class_weight [{dict, ‘balanced’}, optional] Set the parameter C of class i to class_weight[i]*C
for SVC. If not given, all classes are supposed to have weight one. The “balanced” mode
uses the values of y to automatically adjust weights inversely proportional to class frequen-
cies as n_samples / (n_classes * np.bincount(y))
verbose [bool, default: False] Enable verbose output. Note that this setting takes advantage
of a per-process runtime setting in libsvm that, if enabled, may not work properly in a
multithreaded context.
max_iter [int, optional (default=-1)] Hard limit on iterations within solver, or -1 for no limit.
decision_function_shape [‘ovo’, ‘ovr’, default=’ovr’] Whether to return a one-vs-rest (‘ovr’)
decision function of shape (n_samples, n_classes) as all other classifiers, or the original
one-vs-one (‘ovo’) decision function of libsvm which has shape (n_samples, n_classes *
(n_classes - 1) / 2).
Changed in version 0.19: decision_function_shape is ‘ovr’ by default.
New in version 0.17: decision_function_shape=’ovr’ is recommended.
Changed in version 0.17: Deprecated decision_function_shape=’ovo’ and None.
pseudo random number generator used when shuffling the data for probability estimates.
If int, random_state is the seed used by the random number generator; If RandomState
instance, random_state is the random number generator; If None, the random number gen-
erator is the RandomState instance used by np.random.
Attributes
support_ [array-like, shape = [n_SV]] Indices of support vectors.
support_vectors_ [array-like, shape = [n_SV, n_features]] Support vectors.
n_support_ [array-like, dtype=int32, shape = [n_class]] Number of support vectors for each
class.
dual_coef_ [array, shape = [n_class-1, n_SV]] Coefficients of the support vector in the decision
function. For multiclass, coefficient for all 1-vs-1 classifiers. The layout of the coefficients
in the multiclass case is somewhat non-trivial. See the section about multi-class classifica-
tion in the SVM section of the User Guide for details.
coef_ [array, shape = [n_class * (n_class-1) / 2, n_features]] Weights assigned to the features
(coefficients in the primal problem). This is only available in the case of a linear kernel.
coef_ is readonly property derived from dual_coef_ and support_vectors_.
intercept_ [array, shape = [n_class * (n_class-1) / 2]] Constants in decision function.
See also:

SVC Support Vector Machine for classification using libsvm.

LinearSVC Scalable linear Support Vector Machine for classification using liblinear.
Notes
References: LIBSVM: A Library for Support Vector Machines
Examples

>>> X = np.array([[-1, -1], [-2, -1], [1, 1], [2, 1]])
>>> y = np.array([1, 1, 2, 2])
>>> from sklearn.svm import NuSVC
>>> clf = NuSVC(gamma='scale')
>>> clf.fit(X, y)
NuSVC(cache_size=200, class_weight=None, coef0=0.0,
max_iter=-1, nu=0.5, probability=False, random_state=None,
[1]
Methods
decision_function(X) Evaluates the decision function for the samples in X.

fit(X, y[, sample_weight]) Fit the SVM model according to the given training data.
predict(X) Perform classification on samples in X.
labels.
__init__(nu=0.5, kernel=’rbf’, degree=3, gamma=’auto_deprecated’, coef0=0.0, shrinking=True,

probability=False, tol=0.001, cache_size=200, class_weight=None, verbose=False,
max_iter=-1, decision_function_shape=’ovr’, random_state=None)
Evaluates the decision function for the samples in X.
Parameters
Returns
X [array-like, shape (n_samples, n_classes * (n_classes-1) / 2)] Returns the decision func-
tion of the sample for each class in the model. If decision_function_shape=’ovr’, the shape
is (n_samples, n_classes).

Notes
If decision_function_shape=’ovo’, the function values are proportional to the distance of the samples X to
the separating hyperplane. If the exact distances are required, divide the function values by the norm of
the weight vector (coef_). See also this question for further details.
Fit the SVM model according to the given training data.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] Training vectors, where
n_samples is the number of samples and n_features is the number of features. For ker-
nel=”precomputed”, the expected shape of X is (n_samples, n_samples).
y [array-like, shape (n_samples,)] Target values (class labels in classification, real numbers
in regression)
sample_weight [array-like, shape (n_samples,)] Per-sample weights. Rescale C per sample.
Higher weights force the classifier to put more emphasis on these points.
Returns
self [object]
Notes
If X and y are not C-ordered and contiguous arrays of np.float64 and X is not a scipy.sparse.csr_matrix, X
and/or y may be copied.
If X is a dense array, then the other methods will not support sparse matrices as input.
Parameters
Returns
predict(X)
Perform classification on samples in X.
For an one-class model, +1 or -1 is returned.
Parameters
X [{array-like, sparse matrix}, shape (n_samples, n_features)] For kernel=”precomputed”,
the expected shape of X is [n_samples_test, n_samples_train]
Returns
y_pred [array, shape (n_samples,)] Class labels for samples in X.
predict_log_proba
Compute log probabilities of possible outcomes for samples in X.
The model need to have probability information computed at training time: fit with attribute probability
set to True.

Parameters
X [array-like, shape (n_samples, n_features)] For kernel=”precomputed”, the expected
shape of X is [n_samples_test, n_samples_train]
Returns
T [array-like, shape (n_samples, n_classes)] Returns the log-probabilities of the sample for
Notes
The probability model is created using cross validation, so the results can be slightly different than those
obtained by predict. Also, it will produce meaningless results on very small datasets.
predict_proba
set to True.
Parameters
Returns
T [array-like, shape (n_samples, n_classes)] Returns the probability of the sample for each
class in the model. The columns correspond to the classes in sorted order, as they appear
in the attribute classes_.
Notes
Parameters
Returns
of a nested object.

Returns
self
Examples using sklearn.svm.NuSVC
• Non-linear SVM
sklearn.svm.NuSVR
class sklearn.svm.NuSVR(nu=0.5, C=1.0, kernel=’rbf’, degree=3, gamma=’auto_deprecated’,

coef0=0.0, shrinking=True, tol=0.001, cache_size=200, verbose=False,
max_iter=-1)
Nu Support Vector Regression.
Similar to NuSVC, for regression, uses a parameter nu to control the number of support vectors. However,
unlike NuSVC, where nu replaces C, here nu replaces the parameter epsilon of epsilon-SVR.
Parameters
nu [float, optional] An upper bound on the fraction of training errors and a lower bound of the
fraction of support vectors. Should be in the interval (0, 1]. By default 0.5 will be taken.
Attributes

support_vectors_ [array-like, shape = [nSV, n_features]] Support vectors.

dual_coef_ [array, shape = [1, n_SV]] Coefficients of the support vector in the decision func-
tion.
coef_ [array, shape = [1, n_features]] Weights assigned to the features (coefficients in the primal
problem). This is only available in the case of a linear kernel.
intercept_ [array, shape = [1]] Constants in decision function.
See also:
NuSVC Support Vector Machine for classification implemented with libsvm with a parameter to control the
number of support vectors.
SVR epsilon Support Vector Machine for regression implemented with libsvm.
Notes
Examples
>>> from sklearn.svm import NuSVR

>>> clf = NuSVR(gamma='scale', C=1.0, nu=0.1)
>>> clf.fit(X, y)
NuSVR(C=1.0, cache_size=200, coef0=0.0, degree=3, gamma='scale',
verbose=False)
Methods
predict(X) Perform regression on samples in X.
diction.
__init__(nu=0.5, C=1.0, kernel=’rbf’, degree=3, gamma=’auto_deprecated’, coef0=0.0, shrink-

ing=True, tol=0.001, cache_size=200, verbose=False, max_iter=-1)
Parameters


in regression)
Returns
self [object]
Notes
Parameters
Returns
predict(X)
Perform regression on samples in X.
For an one-class model, +1 (inlier) or -1 (outlier) is returned.
Parameters
the expected shape of X is (n_samples_test, n_samples_train).
Returns
y_pred [array, shape (n_samples,)]
Parameters

Returns
of a nested object.
Returns
self
Examples using sklearn.svm.NuSVR
sklearn.svm.OneClassSVM
class sklearn.svm.OneClassSVM(kernel=’rbf’, degree=3, gamma=’auto_deprecated’, coef0=0.0,

tol=0.001, nu=0.5, shrinking=True, cache_size=200, verbose=False,
max_iter=-1, random_state=None)
Unsupervised Outlier Detection.
Estimate the support of a high-dimensional distribution.
Parameters
tol [float, optional] Tolerance for stopping criterion.
nu [float, optional] An upper bound on the fraction of training errors and a lower bound of the
fraction of support vectors. Should be in the interval (0, 1]. By default 0.5 will be taken.
shrinking [boolean, optional] Whether to use the shrinking heuristic.

random_state [int, RandomState instance or None, optional (default=None)] Ignored.
Deprecated since version 0.20: random_state has been deprecated in 0.20 and will be
removed in 0.22.
Attributes
dual_coef_ [array, shape = [1, n_SV]] Coefficients of the support vectors in the decision func-
tion.
coef_ is readonly property derived from dual_coef_ and support_vectors_
intercept_ [array, shape = [1,]] Constant in the decision function.
offset_ [float] Offset used to define the decision function from the raw scores. We have the re-
lation: decision_function = score_samples - offset_. The offset is the opposite of intercept_
and is provided for consistency with other outlier detection algorithms.
Methods
decision_function(X) Signed distance to the separating hyperplane.

fit(X[, y, sample_weight]) Detects the soft boundary of the set of samples X.
score_samples(X) Raw scoring function of the samples.
__init__(kernel=’rbf’, degree=3, gamma=’auto_deprecated’, coef0=0.0, tol=0.001, nu=0.5, shrink-

ing=True, cache_size=200, verbose=False, max_iter=-1, random_state=None)
Signed distance to the separating hyperplane.
Signed distance is positive for an inlier and negative for an outlier.
Parameters
Returns
dec [array-like, shape (n_samples,)] Returns the decision function of the samples.
fit(X, y=None, sample_weight=None, **params)
Detects the soft boundary of the set of samples X.
Parameters

X [{array-like, sparse matrix}, shape (n_samples, n_features)] Set of samples, where

Returns
self [object]
Notes
If X is not a C-ordered contiguous array it is copied.

Parameters
Returns
Parameters
Returns
predict(X)
Parameters
Returns
score_samples(X)
Raw scoring function of the samples.
Parameters
Returns

score_samples [array-like, shape (n_samples,)] Returns the (unshifted) scoring function of

the samples.
of a nested object.
Returns
self
Examples using sklearn.svm.OneClassSVM

• Libsvm GUI
• One-class SVM with non-linear kernel (RBF)
sklearn.svm.SVC
class sklearn.svm.SVC(C=1.0, kernel=’rbf’, degree=3, gamma=’auto_deprecated’, coef0=0.0, shrink-

ing=True, probability=False, tol=0.001, cache_size=200, class_weight=None,
verbose=False, max_iter=-1, decision_function_shape=’ovr’, ran-
dom_state=None)
C-Support Vector Classification.
The implementation is based on libsvm. The fit time complexity is more than quadratic with the number of
samples which makes it hard to scale to dataset with more than a couple of 10000 samples.
The multiclass support is handled according to a one-vs-one scheme.
For details on the precise mathematical formulation of the provided kernel functions and how gamma, coef0 and
degree affect each other, see the corresponding section in the narrative documentation: Kernel functions.
Parameters
It must be one of ‘linear’, ‘poly’, ‘rbf’, ‘sigmoid’, ‘precomputed’ or a callable. If none
is given, ‘rbf’ will be used. If a callable is given it is used to pre-compute the ker-
nel matrix from data matrices; that matrix should be an array of shape (n_samples,
n_samples).


probability [boolean, optional (default=False)] Whether to enable probability estimates. This
must be enabled prior to calling fit, and will slow down that method.
class_weight [{dict, ‘balanced’}, optional] Set the parameter C of class i to class_weight[i]*C
for SVC. If not given, all classes are supposed to have weight one. The “balanced” mode
uses the values of y to automatically adjust weights inversely proportional to class frequen-
cies in the input data as n_samples / (n_classes * np.bincount(y))
decision_function_shape [‘ovo’, ‘ovr’, default=’ovr’] Whether to return a one-vs-rest (‘ovr’)
decision function of shape (n_samples, n_classes) as all other classifiers, or the original
one-vs-one (‘ovo’) decision function of libsvm which has shape (n_samples, n_classes *
(n_classes - 1) / 2). However, one-vs-one (‘ovo’) is always used as multi-class strategy.
Changed in version 0.19: decision_function_shape is ‘ovr’ by default.
New in version 0.17: decision_function_shape=’ovr’ is recommended.
Changed in version 0.17: Deprecated decision_function_shape=’ovo’ and None.
pseudo random number generator used when shuffling the data for probability estimates.
If int, random_state is the seed used by the random number generator; If RandomState
instance, random_state is the random number generator; If None, the random number gen-
erator is the RandomState instance used by np.random.
Attributes
support_vectors_ [array-like, shape = [n_SV, n_features]] Support vectors.
n_support_ [array-like, dtype=int32, shape = [n_class]] Number of support vectors for each
class.
dual_coef_ [array, shape = [n_class-1, n_SV]] Coefficients of the support vector in the decision
function. For multiclass, coefficient for all 1-vs-1 classifiers. The layout of the coefficients
in the multiclass case is somewhat non-trivial. See the section about multi-class classifica-
tion in the SVM section of the User Guide for details.
coef_ [array, shape = [n_class * (n_class-1) / 2, n_features]] Weights assigned to the features
(coefficients in the primal problem). This is only available in the case of a linear kernel.
coef_ is a readonly property derived from dual_coef_ and support_vectors_.
intercept_ [array, shape = [n_class * (n_class-1) / 2]] Constants in decision function.
fit_status_ [int] 0 if correctly fitted, 1 otherwise (will raise warning)

probA_ [array, shape = [n_class * (n_class-1) / 2]]

probB_ [array, shape = [n_class * (n_class-1) / 2]] If probability=True, the parameters
learned in Platt scaling to produce probability estimates from decision values. If prob-
ability=False, an empty array. Platt scaling uses the logistic function 1 / (1 +
exp(decision_value * probA_ + probB_)) where probA_ and probB_ are
learned from the dataset [2]. For more information on the multiclass case and training pro-
cedure see section 8 of [1].
See also:
SVR Support Vector Machine for Regression implemented using libsvm.

LinearSVC Scalable Linear Support Vector Machine for classification implemented using liblinear. Check
the See also section of LinearSVC for more comparison element.
References
[1], [2]
Examples

>>> X = np.array([[-1, -1], [-2, -1], [1, 1], [2, 1]])
>>> y = np.array([1, 1, 2, 2])
>>> clf = SVC(gamma='auto')
>>> clf.fit(X, y)
decision_function_shape='ovr', degree=3, gamma='auto', kernel='rbf',
[1]
Methods
decision_function(X) Evaluates the decision function for the samples in X.

labels.
__init__(C=1.0, kernel=’rbf’, degree=3, gamma=’auto_deprecated’, coef0=0.0, shrinking=True,

probability=False, tol=0.001, cache_size=200, class_weight=None, verbose=False,
max_iter=-1, decision_function_shape=’ovr’, random_state=None)
Evaluates the decision function for the samples in X.
Parameters


Returns
X [array-like, shape (n_samples, n_classes * (n_classes-1) / 2)] Returns the decision func-
tion of the sample for each class in the model. If decision_function_shape=’ovr’, the shape
is (n_samples, n_classes).
Notes
If decision_function_shape=’ovo’, the function values are proportional to the distance of the samples X to
the separating hyperplane. If the exact distances are required, divide the function values by the norm of
the weight vector (coef_). See also this question for further details.
Parameters
in regression)
Returns
self [object]
Notes
Parameters
Returns
predict(X)
Parameters

Returns
predict_log_proba
Compute log probabilities of possible outcomes for samples in X.
set to True.
Parameters
Returns
T [array-like, shape (n_samples, n_classes)] Returns the log-probabilities of the sample for
Notes
predict_proba
set to True.
Parameters
Returns
T [array-like, shape (n_samples, n_classes)] Returns the probability of the sample for each
class in the model. The columns correspond to the classes in sorted order, as they appear
in the attribute classes_.
Notes
Parameters

Returns
of a nested object.
Returns
self
Examples using sklearn.svm.SVC
• Libsvm GUI
• SVM Exercise
• SVM: Maximum margin separating hyperplane

• SVM with custom kernel

• SVM: Weighted samples
• SVM-Kernels
• SVM Margins Example
sklearn.svm.SVR
class sklearn.svm.SVR(kernel=’rbf’, degree=3, gamma=’auto_deprecated’, coef0=0.0, tol=0.001,

C=1.0, epsilon=0.1, shrinking=True, cache_size=200, verbose=False,
max_iter=-1)
Epsilon-Support Vector Regression.
The free parameters in the model are C and epsilon.
Parameters
epsilon [float, optional (default=0.1)] Epsilon in the epsilon-SVR model. It specifies the
epsilon-tube within which no penalty is associated in the training loss function with points
predicted within a distance epsilon from the actual value.

Attributes
dual_coef_ [array, shape = [1, n_SV]] Coefficients of the support vector in the decision func-
tion.
intercept_ [array, shape = [1]] Constants in decision function.
See also:
NuSVR Support Vector Machine for regression implemented using libsvm using a parameter to control the
number of support vectors.
LinearSVR Scalable Linear Support Vector Machine for regression implemented using liblinear.
Notes
Examples

>>> clf = SVR(gamma='scale', C=1.0, epsilon=0.2)
>>> clf.fit(X, y)
SVR(C=1.0, cache_size=200, coef0=0.0, degree=3, epsilon=0.2, gamma='scale',
kernel='rbf', max_iter=-1, shrinking=True, tol=0.001, verbose=False)
Methods
predict(X) Perform regression on samples in X.
diction.
__init__(kernel=’rbf’, degree=3, gamma=’auto_deprecated’, coef0=0.0, tol=0.001, C=1.0, ep-

silon=0.1, shrinking=True, cache_size=200, verbose=False, max_iter=-1)

Parameters
in regression)
Returns
self [object]
Notes
Parameters
Returns
predict(X)
Perform regression on samples in X.
For an one-class model, +1 (inlier) or -1 (outlier) is returned.
Parameters
the expected shape of X is (n_samples_test, n_samples_train).
Returns
y_pred [array, shape (n_samples,)]
Parameters


Returns
of a nested object.
Returns
self
Examples using sklearn.svm.SVR

• Support Vector Regression (SVR) using linear and non-linear kernels
svm.l1_min_c(X, y[, loss, fit_intercept, . . . ]) Return the lowest bound for C such that for C in (l1_min_C,
infinity) the model is guaranteed not to be empty.
sklearn.svm.l1_min_c
sklearn.svm.l1_min_c(X, y, loss=’squared_hinge’, fit_intercept=True, intercept_scaling=1.0)

Return the lowest bound for C such that for C in (l1_min_C, infinity) the model is guaranteed not
to be empty. This applies to l1 penalized classifiers, such as LinearSVC with penalty=’l1’ and lin-
ear_model.LogisticRegression with penalty=’l1’.
This value is valid if class_weight parameter in fit() is not set.
Parameters
X [array-like or sparse matrix, shape = [n_samples, n_features]] Training vector, where
y [array, shape = [n_samples]] Target vector relative to X
loss [{‘squared_hinge’, ‘log’}, default ‘squared_hinge’] Specifies the loss function. With
‘squared_hinge’ it is the squared hinge loss (a.k.a. L2 loss). With ‘log’ it is the loss of
logistic regression models.
fit_intercept [bool, default: True] Specifies if the intercept should be fitted by the model. It
must match the fit() method parameter.
intercept_scaling [float, default: 1] when fit_intercept is True, instance vector x becomes [x,
intercept_scaling], i.e. a “synthetic” feature with constant value equals to intercept_scaling
is appended to the instance vector. It must match the fit() method parameter.
Returns
l1_min_c [float] minimum value for C

Examples using sklearn.svm.l1_min_c
6.35.2 Low-level methods
svm.libsvm.cross_validation Binding of the cross-validation routine (low-level routine)

svm.libsvm.decision_function Predict margin (libsvm name for this is predict_values)
svm.libsvm.fit Train the model using libsvm (low-level method)
svm.libsvm.predict Predict target values of X given a model (low-level method)
svm.libsvm.predict_proba Predict probabilities
sklearn.svm.libsvm.cross_validation
sklearn.svm.libsvm.cross_validation()
Binding of the cross-validation routine (low-level routine)
Parameters
X [array-like, dtype=float, size=[n_samples, n_features]]
Y [array, dtype=float, size=[n_samples]] target vector
svm_type [{0, 1, 2, 3, 4}] Type of SVM: C SVC, nu SVC, one class, epsilon SVR, nu SVR
kernel [{‘linear’, ‘rbf’, ‘poly’, ‘sigmoid’, ‘precomputed’}] Kernel to use in the model: linear,
polynomial, RBF, sigmoid or precomputed.
degree [int] Degree of the polynomial kernel (only relevant if kernel is set to polynomial)
gamma [float] Gamma parameter in rbf, poly and sigmoid kernels. Ignored by other kernels.
0.1 by default.
coef0 [float] Independent parameter in poly/sigmoid kernel.
tol [float] Stopping criteria.
C [float] C parameter in C-Support Vector Classification
nu [float]
cache_size [float]
random_seed [int, optional] Seed for the random number generator used for probability esti-
mates. 0 by default.
Returns
target [array, float]
sklearn.svm.libsvm.decision_function
sklearn.svm.libsvm.decision_function()
Predict margin (libsvm name for this is predict_values)
We have to reconstruct model and parameters to make sure we stay in sync with the python object.

sklearn.svm.libsvm.fit
sklearn.svm.libsvm.fit()
Train the model using libsvm (low-level method)
Parameters
X [array-like, dtype=float64, size=[n_samples, n_features]]
Y [array, dtype=float64, size=[n_samples]] target vector
svm_type [{0, 1, 2, 3, 4}, optional] Type of SVM: C_SVC, NuSVC, OneClassSVM, Ep-
silonSVR or NuSVR respectively. 0 by default.
kernel [{‘linear’, ‘rbf’, ‘poly’, ‘sigmoid’, ‘precomputed’}, optional] Kernel to use in the model:
linear, polynomial, RBF, sigmoid or precomputed. ‘rbf’ by default.
degree [int32, optional] Degree of the polynomial kernel (only relevant if kernel is set to poly-
nomial), 3 by default.
gamma [float64, optional] Gamma parameter in rbf, poly and sigmoid kernels. Ignored by
other kernels. 0.1 by default.
coef0 [float64, optional] Independent parameter in poly/sigmoid kernel. 0 by default.
tol [float64, optional] Numeric stopping criterion (WRITEME). 1e-3 by default.
C [float64, optional] C parameter in C-Support Vector Classification. 1 by default.
nu [float64, optional] 0.5 by default.
epsilon [double, optional] 0.1 by default.
class_weight [array, dtype float64, shape (n_classes,), optional] np.empty(0) by default.
sample_weight [array, dtype float64, shape (n_samples,), optional] np.empty(0) by default.
shrinking [int, optional] 1 by default.
probability [int, optional] 0 by default.
cache_size [float64, optional] Cache size for gram matrix columns (in megabytes). 100 by
default.
max_iter [int (-1 for no limit), optional.] Stop solver after this many iterations regardless of
accuracy (XXX Currently there is no API to know whether this kicked in.) -1 by default.
random_seed [int, optional] Seed for the random number generator used for probability esti-
mates. 0 by default.
Returns
support [array, shape=[n_support]] index of support vectors
support_vectors [array, shape=[n_support, n_features]] support vectors (equivalent to
X[support]). Will return an empty array in the case of precomputed kernel.
n_class_SV [array] number of support vectors in each class.
sv_coef [array] coefficients of support vectors in decision function.
intercept [array] intercept in decision function
probA, probB [array] probability estimates, empty array for probability=False

sklearn.svm.libsvm.predict
sklearn.svm.libsvm.predict()
Predict target values of X given a model (low-level method)
Parameters
X [array-like, dtype=float, size=[n_samples, n_features]]
svm_type [{0, 1, 2, 3, 4}] Type of SVM: C SVC, nu SVC, one class, epsilon SVR, nu SVR
kernel [{‘linear’, ‘rbf’, ‘poly’, ‘sigmoid’, ‘precomputed’}] Type of kernel.
degree [int] Degree of the polynomial kernel.
gamma [float] Gamma parameter in rbf, poly and sigmoid kernels. Ignored by other kernels.
0.1 by default.
coef0 [float] Independent parameter in poly/sigmoid kernel.
Returns
dec_values [array] predicted values.
sklearn.svm.libsvm.predict_proba
sklearn.svm.libsvm.predict_proba()
Predict probabilities
svm_model stores all parameters needed to predict a given value.
For speed, all real work is done at the C level in function copy_predict (libsvm_helper.c).
We have to reconstruct model and parameters to make sure we stay in sync with the python object.
See sklearn.svm.predict for a complete list of parameters.
Parameters
X [array-like, dtype=float]
kernel [{‘linear’, ‘rbf’, ‘poly’, ‘sigmoid’, ‘precomputed’}]
Returns
dec_values [array] predicted values.
6.36 sklearn.tree: Decision Trees
The sklearn.tree module includes decision tree-based models for classification and regression.
User guide: See the Decision Trees section for further details.
tree.DecisionTreeClassifier([criterion, . . . ]) A decision tree classifier.

tree.DecisionTreeRegressor([criterion, . . . ]) A decision tree regressor.
tree.ExtraTreeClassifier([criterion, . . . ]) An extremely randomized tree classifier.
tree.ExtraTreeRegressor([criterion, . . . ]) An extremely randomized tree regressor.

6.36.1 sklearn.tree.DecisionTreeClassifier
class sklearn.tree.DecisionTreeClassifier(criterion=’gini’, splitter=’best’, max_depth=None,

max_features=None, ran-
dom_state=None, max_leaf_nodes=None,
min_impurity_split=None, class_weight=None,
presort=False)
A decision tree classifier.
Parameters
splitter [string, optional (default=”best”)] The strategy used to choose the split at each node.
Supported strategies are “best” to choose the best split and “random” to choose the best
random split.
max_depth [int or None, optional (default=None)] The maximum depth of the tree. If None,
6.36. sklearn.tree: Decision Trees 2215


max_leaf_nodes [int or None, optional (default=None)] Grow a tree with max_leaf_nodes

the right child.
class_weight [dict, list of dicts, “balanced” or None, default=None] Weights associated with
classes in the form {class_label: weight}. If not given, all classes are supposed
to have weight one. For multi-output problems, a list of dicts can be provided in the same
order as the columns of y.
{3:1}, {4:1}].
bincount(y))
presort [bool, optional (default=False)] Whether to presort the data to speed up the finding of
best splits in fitting. For the default settings of a decision tree on large datasets, setting

this to true may slow down the training process. When using either a smaller dataset or a
restricted depth, this may speed up the training.
Attributes
feature_importances_ [array of shape = [n_features]] Return the feature importances.
max_features_ [int,] The inferred value of max_features.
n_classes_ [int or list] The number of classes (for single output problems), or a list containing
the number of classes for each output (for multi-output problems).
tree_ [Tree object] The underlying Tree object. Please refer to help(sklearn.tree.
_tree.Tree) for attributes of Tree object and Understanding the decision tree structure
for basic usage of these attributes.
See also:
DecisionTreeRegressor
Notes
References
[1], [2], [3], [4]
Examples

>>> clf = DecisionTreeClassifier(random_state=0)
>>> cross_val_score(clf, iris.data, iris.target, cv=10)
...
...
array([ 1. , 0.93..., 0.86..., 0.93..., 0.93...,
0.93..., 0.93..., 1. , 0.93..., 1. ])

Methods
apply(X[, check_input]) Returns the index of the leaf that each sample is pre-
dicted as.
decision_path(X[, check_input]) Return the decision path in the tree
fit(X, y[, sample_weight, check_input, . . . ]) Build a decision tree classifier from the training set (X,
y).
predict(X[, check_input]) Predict class or regression value for X.
predict_log_proba(X) Predict class log-probabilities of the input samples X.
predict_proba(X[, check_input]) Predict class probabilities of the input samples X.
labels.
__init__(criterion=’gini’, splitter=’best’, max_depth=None, min_samples_split=2,

min_samples_leaf=1, min_weight_fraction_leaf=0.0, max_features=None,
random_state=None, max_leaf_nodes=None, min_impurity_decrease=0.0,
min_impurity_split=None, class_weight=None, presort=False)
apply(X, check_input=True)
Returns the index of the leaf that each sample is predicted as.
Parameters
X [array_like or sparse matrix, shape = [n_samples, n_features]] The input samples. Inter-
Returns
X_leaves [array_like, shape = [n_samples,]] For each datapoint x in X, return the index of
the leaf x ends up in. Leaves are numbered within [0; self.tree_.node_count),
possibly with gaps in the numbering.
decision_path(X, check_input=True)
Return the decision path in the tree
Parameters
Returns

Return the feature importances.
The importance of a feature is computed as the (normalized) total reduction of the criterion brought by that
feature. It is also known as the Gini importance.
Returns
fit(X, y, sample_weight=None, check_input=True, X_idx_sorted=None)
Build a decision tree classifier from the training set (X, y).
Parameters
X [array-like or sparse matrix, shape = [n_samples, n_features]] The training input samples.
Internally, it will be converted to dtype=np.float32 and if a sparse matrix is provided
to a sparse csc_matrix.
y [array-like, shape = [n_samples] or [n_samples, n_outputs]] The target values (class la-
bels) as integers or strings.
samples are equally weighted. Splits that would create child nodes with net zero or nega-
tive weight are ignored while searching for a split in each node. Splits are also ignored if
they would result in any single class carrying a negative weight in either child node.
X_idx_sorted [array-like, shape = [n_samples, n_features], optional] The indexes of the
sorted training input samples. If many tree are grown on the same dataset, this allows the
ordering to be cached between trees. If None, the data will be sorted here. Don’t use this
parameter unless you know what to do.
Returns
self [object]
Parameters
Returns
predict(X, check_input=True)
Predict class or regression value for X.
For a classification model, the predicted class for each sample in X is returned. For a regression model,
the predicted value based on X is returned.
Parameters

Returns
y [array of shape = [n_samples] or [n_samples, n_outputs]] The predicted classes, or the
predict values.
Predict class log-probabilities of the input samples X.
Parameters
Returns
> 1. The class log-probabilities of the input samples. The order of the classes corresponds
to that in the attribute classes_.
predict_proba(X, check_input=True)
Predict class probabilities of the input samples X.
The predicted class probability is the fraction of samples of the same class in a leaf.
check_input [boolean, (default=True)] Allow to bypass several input checking. Don’t use this parameter
unless you know what you do.
Parameters
Returns
Parameters
Returns

of a nested object.
Returns
self
Examples using sklearn.tree.DecisionTreeClassifier
6.36.2 sklearn.tree.DecisionTreeRegressor
class sklearn.tree.DecisionTreeRegressor(criterion=’mse’, splitter=’best’, max_depth=None,

min_weight_fraction_leaf=0.0, max_features=None,
random_state=None, max_leaf_nodes=None,
min_impurity_split=None, presort=False)
A decision tree regressor.
Parameters
feature selection criterion and minimizes the L2 loss using the mean of each terminal node,
“friedman_mse”, which uses mean squared error with Friedman’s improvement score for
potential splits, and “mae” for the mean absolute error, which minimizes the L1 loss using
the median of each terminal node.
splitter [string, optional (default=”best”)] The strategy used to choose the split at each node.
random split.



the right child.


presort [bool, optional (default=False)] Whether to presort the data to speed up the finding of
best splits in fitting. For the default settings of a decision tree on large datasets, setting
this to true may slow down the training process. When using either a smaller dataset or a
restricted depth, this may speed up the training.
Attributes
feature_importances_ [array of shape = [n_features]] Return the feature importances.
max_features_ [int,] The inferred value of max_features.
tree_ [Tree object] The underlying Tree object. Please refer to help(sklearn.tree.
_tree.Tree) for attributes of Tree object and Understanding the decision tree structure
for basic usage of these attributes.
See also:
DecisionTreeClassifier
Notes
References
[1], [2], [3], [4]
Examples

>>> from sklearn.tree import DecisionTreeRegressor
>>> regressor = DecisionTreeRegressor(random_state=0)
>>> cross_val_score(regressor, boston.data, boston.target, cv=10)

...
...
array([ 0.61..., 0.57..., -0.34..., 0.41..., 0.75...,
0.07..., 0.29..., 0.33..., -1.42..., -1.77...])
Methods
dicted as.
fit(X, y[, sample_weight, check_input, . . . ]) Build a decision tree regressor from the training set (X,
y).
diction.
__init__(criterion=’mse’, splitter=’best’, max_depth=None, min_samples_split=2,

min_samples_leaf=1, min_weight_fraction_leaf=0.0, max_features=None,
min_impurity_split=None, presort=False)
Parameters
Returns
Parameters
Returns

Returns
Build a decision tree regressor from the training set (X, y).
Parameters
y [array-like, shape = [n_samples] or [n_samples, n_outputs]] The target values (real num-
bers). Use dtype=np.float64 and order='C' for maximum efficiency.
tive weight are ignored while searching for a split in each node.
Returns
self [object]
Parameters
Returns
Parameters

Returns
predict values.
Parameters
Returns
of a nested object.
Returns
self
Examples using sklearn.tree.DecisionTreeRegressor
• Decision Tree Regression with AdaBoost

• Decision Tree Regression
• Multi-output Decision Tree Regression
6.36.3 sklearn.tree.ExtraTreeClassifier
class sklearn.tree.ExtraTreeClassifier(criterion=’gini’, splitter=’random’,

max_depth=None, min_samples_split=2,
min_samples_leaf=1, min_weight_fraction_leaf=0.0,
max_features=’auto’, random_state=None,
min_impurity_split=None, class_weight=None)
An extremely randomized tree classifier.

Extra-trees differ from classic decision trees in the way they are built. When looking for the best split to separate
the samples of a node into two groups, random splits are drawn for each of the max_features randomly selected
features and the best split among those is chosen. When max_features is set 1, this amounts to building a totally
random decision tree.
Warning: Extra-trees should only be used within ensemble methods.
Parameters
splitter [string, optional (default=”random”)] The strategy used to choose the split at each node.
random split.


the right child.
class_weight [dict, list of dicts, “balanced” or None, default=None] Weights associated with
classes in the form {class_label: weight}. If not given, all classes are supposed
to have weight one. For multi-output problems, a list of dicts can be provided in the same
order as the columns of y.
{3:1}, {4:1}].
bincount(y))
Attributes
feature_importances_ Return the feature importances.
See also:
ExtraTreeRegressor, sklearn.ensemble.ExtraTreesClassifier, sklearn.ensemble.
ExtraTreesRegressor

Notes
References
[1]
Methods
dicted as.
fit(X, y[, sample_weight, check_input, . . . ]) Build a decision tree classifier from the training set (X,
y).
predict_log_proba(X) Predict class log-probabilities of the input samples X.
predict_proba(X[, check_input]) Predict class probabilities of the input samples X.
labels.
__init__(criterion=’gini’, splitter=’random’, max_depth=None, min_samples_split=2,

min_impurity_split=None, class_weight=None)
Parameters
Returns

Parameters
Returns
Returns
Build a decision tree classifier from the training set (X, y).
Parameters
y [array-like, shape = [n_samples] or [n_samples, n_outputs]] The target values (class la-
bels) as integers or strings.
tive weight are ignored while searching for a split in each node. Splits are also ignored if
they would result in any single class carrying a negative weight in either child node.
Returns
self [object]
Parameters
Returns

Parameters
Returns
predict values.
Predict class log-probabilities of the input samples X.
Parameters
Returns
> 1. The class log-probabilities of the input samples. The order of the classes corresponds
to that in the attribute classes_.
predict_proba(X, check_input=True)
Predict class probabilities of the input samples X.
The predicted class probability is the fraction of samples of the same class in a leaf.
check_input [boolean, (default=True)] Allow to bypass several input checking. Don’t use this parameter
unless you know what you do.
Parameters
Returns
Parameters


Returns
of a nested object.
Returns
self
6.36.4 sklearn.tree.ExtraTreeRegressor
class sklearn.tree.ExtraTreeRegressor(criterion=’mse’, splitter=’random’, max_depth=None,

min_weight_fraction_leaf=0.0, max_features=’auto’,
random_state=None, min_impurity_decrease=0.0,
min_impurity_split=None, max_leaf_nodes=None)
An extremely randomized tree regressor.
Extra-trees differ from classic decision trees in the way they are built. When looking for the best split to separate
the samples of a node into two groups, random splits are drawn for each of the max_features randomly selected
features and the best split among those is chosen. When max_features is set 1, this amounts to building a totally
random decision tree.
Warning: Extra-trees should only be used within ensemble methods.
Parameters
criterion [string, optional (default=”mse”)] The function to meas

Scikit Learn Docs PDF

Uploaded by

Scikit Learn Docs PDF

Uploaded by

scikit-learn user guide

Mar 01, 2019

2 scikit-learn Tutorials 129

3 User Guide 173

4 Glossary of Common Terms and API Elements 623

6 API Reference 1393

7 Developer’s Guide 2293

1.1 Installing scikit-learn

1.1.1 Installing the latest release

pip install -U scikit-learn

conda install scikit-learn

1.1.2 Third-party Distributions

Canopy and Anaconda for all supported platforms

Upgrading with pip install -U scikit-learn or uninstalling pip uninstall scikit-learn is

WinPython for Windows

The WinPython project distributes scikit-learn as an additional plugin.

1.2 Frequently Asked Questions

1.2.1 What is the project name (a lot of people get it wrong)?

2 Chapter 1. Welcome to scikit-learn

1.2.2 How do you pronounce the project name?

sy-kit learn. sci stands for science!

1.2.3 Why scikit?

1.2.4 How can I contribute to scikit-learn?

1.2.5 What’s the best way to get help on scikit-learn usage?

1.2.6 How should I save, export or deploy estimators for production?

See Model persistence.

1.2.7 How can I create a bunch object?

1.2. Frequently Asked Questions 3

1.2.9 What are the inclusion criteria for new algorithms ?

1.2.11 Why did you remove HMMs from scikit-learn?

See Will you add graphical models or sequence prediction to scikit-learn?.

1.2.12 Will you add graphical models or sequence prediction to scikit-learn?

4 Chapter 1. Welcome to scikit-learn

1.2.13 Will you add GPU support?

1.2.14 Do you support PyPy?

1.2.15 How do I deal with string data (or trees, graphs. . . )?

>>> from leven import levenshtein

1.2. Frequently Asked Questions 5

(This uses the third-party edit distance package leven.)

# other imports, custom code, load data, define model...

# call scikit-learn utils with n_jobs > 1 here

6 Chapter 1. Welcome to scikit-learn

1.2.19 Why is my pull request not getting any attention?

1.2.20 How do I set a random_state for an entire execution?

1.2.21 Why do categorical variables need preprocessing in scikit-learn, compared

1.2. Frequently Asked Questions 7

There are several ways to get in touch with the developers.

1.3.1 Mailing List

• The main mailing list is scikit-learn.

1.3.2 User questions

1.3.3 Bug tracker

Some developers like to hang out on channel #scikit-learn on irc.freenode.net.

8 Chapter 1. Welcome to scikit-learn

1.3.5 Documentation resources

1.4 Related Projects

1.4.1 Interoperability and framework enhancements

1.4. Related Projects 9

1.4.2 Other estimators and tasks

10 Chapter 1. Welcome to scikit-learn

• multiisotonic Isotonic regression on multidimensional features.

1.4.3 Statistical learning with Python

Other packages useful for data analysis and machine learning.

Domain specific packages

• scikit-image Image processing and computer vision in python.

1.4. Related Projects 11

1.4.4 Snippets and tidbits

The wiki has more!

1.5.4 Citing scikit-learn

12 Chapter 1. Welcome to scikit-learn

2017). NYU Moore-Sloan Data Science Environment funded An-

eral students to work on the project part-time. Télécom Paris-

from the Alfred P. Sloan Foundation in 2017. The University of

The following students were sponsored by Google to work on

14 Chapter 1. Welcome to scikit-learn

• 2012 - ‘Vlad Niculae‘_, Immanuel Bayer.

Donating to the project