Introduction to Machine Learning.

Basic Concepts and Learning

Paradigms.

Radu Ionescu, Prof. PhD.

raducu.ionescu@gmail.com

Faculty of Mathematics and Computer Science

University of Bucharest
 Machine Learning
Artificial Intelligence

Natural Language
Kernel Methods Processing

Deep Learning Nearest Neighbors Knowledge

Representation
Markov Models Machine K-means
Learning Expert Systems
Bayesian Models Random Forest

Hierarchical Clustering Computer Vision

Image Processing
Signal Processing
Instructors
• Lectures:
Radu Ionescu (raducu.ionescu@gmail.com)

• Labs:
Alin Croitoru (alincroitoru97@gmail.com)
Eduard Poesina (eduardgabriel.poe@gmail.com)
Vlad Hondru (vlad.hondru25@gmail.com)

• Website:
https://practical-ml-fmi.github.io/ML/
 Grading System
• Your final grade is composed of:
 50% for Project 1
 50% for Project 2
• Both projects are individual!
• Each project consists of employing machine learning methods
on a specific data set
• Project 1 is about participating in a Kaggle competition
 The competition will be launched in a couple of weeks
• Project 2 is about comparing two unsupervised approaches
 There are many datasets out there, so no overlap allowed
among students!
 Methods and data sets must be chosen beforehand!
 Grading System
• Project 1 must be presented no later than week 10
• Project 2 must be presented no later than the day of the “exam”
• There will be no paper exam, only oral exam!
• The average grade of projects 1 and 2 must be >= 5
• The project consists of the code implementation in Python (any
library is allowed) and a PDF report including (2 points):
 a description of the data set (for project 2 only)
 a description of the implemented machine learning methods
 figures and / or tables with results / hyperparameter tuning
 comments / interpretation for the results
 conclusion
 Grading System

First project consists of implementing some machine learning
method(s) for the proposed Kaggle challenge (TBA)

The grades will be proportional to your model’s accuracy:
- Top 1-20 => your grade can be up to 10
- Top 21-50 => your grade can be up to 9
- Top 51-80 => your grade can be up to 8
- Top 81-100 => your grade can be up to 7
- Top 101-120 => your grade can be up to 6
- Others => your grade can be up to 5

Ranks can change depending on the final number of
participants
 Grading System

Submit projects to: practical.ml.fmi@gmail.com

Submit only .py files only! (.ipynb not accepted)

We will set deadlines (during every evaluation session) for:
 choosing the project
 submitting the project
 presenting the project
• If you don’t know the dates, please ask! Don’t wait until the
presentation day!
 Grading System
• Extra points during lectures / labs
 awarded only in the first round of evaluation
• Lectures:
 awarded based on the ranking of answers on Kahoot
 top 3 get up to 0.3 points per lecture, next 3 up to 0.2 points and
so on
• Labs:
 first to answer solve an exercise gets 0.2 points
 maximum 0.4 points per lab for each student
• Up to 1 bonus point during lectures (added to final grade)
• Up to 1 bonus point during labs (added to final grade)
• Maybe up to 2 bonus points for some data annotation (TBD)
 (NO) Collaboration Policy

• Collaboration
Each student must write their own code for the project(s)
Borrowing code from web sources with copy & paste is
not permitted under any circumstances

• No tolerance on plagiarism
Neither ethical nor in your best interest
Code will be checked automatically and manually!
Don’t cheat. We will find out!
We are serious about this!
Examples of unacceptable plagiarism
Examples of unacceptable plagiarism
Examples of unacceptable plagiarism
Examples of unacceptable plagiarism
Examples of unacceptable plagiarism
Examples of unacceptable plagiarism
Examples of acceptable code
Examples of acceptable code
Examples of acceptable code
 What is artificial intelligence (AI)?
• The ultimate goal of artificial intelligence is to build systems
able to reach human intelligence levels
• Turing test: a computer is said to possess human-level
intelligence if a remote human interrogator, within a fixed time
frame, cannot distinguish between the computer and a human
subject based on their replies to various questions posed by the
interrogator
Perhaps we are going in the right
direction?
 What is machine learning (ML)?

• Many AI researchers consider the ultimate goal of AI can

be achieved by imitating the way humans learn
• Machine Learning – is the scientific study of algorithms
and statistical models that computer systems use to learn
from observations, without being explicitly programmed
• In this context, learning refers to:
 recognizing complex patterns in data
 making intelligent decisions based on data observations
 Classic Programming

Data
Computer Output
Program

Machine Learning
Data
Computer Program
Output
 A well-posed machine learning
problem
• What problems can be solved* with machine learning?
• Well-posed machine learning problem:
"A computer program is said to learn from experience E with
respect to some class of tasks T and performance measure
P, if its performance at tasks in T, as measured by P,
improves with experience E.” – Tom Mitchell
(*) implies a certain degree of accuracy
 A well-posed machine learning
problem
• Arthur Samuel (1959) wrote a program for playing checkers
(perhaps the first program based on the concept of learning, as
defined by Tom Mitchell)
• The program played 10K games against itself
• The program was designed to find the good and bad positions
on the board from the current state, based on the probability of
winning or losing
• In this example:
 E = 10000 games
 T = play checkers
 P = win or lose
 Strong AI versus Weak AI
• Strong / generic / true AI
(see the Turing test and its extensions)

• Weak / narrow AI
(focuses on a specific well-posed problem)
When do we use machine learning?
• We use ML when it is hard (impossible) to define a set of
rules by hand / to write a program based on explicit rules

• Examples of tasks that be solved through machine

learning:
 face detection
 speech recognition
 stock price prediction
 object recognition
 The essence of machine learning
• A pattern exists
• We cannot express it programmatically
• We have data on it
 What is machine learning?
• [Arthur Samuel, 1959] field of study that:
• gives computers the ability to learn without being
explicitly programmed

• [Kevin Murphy] algorithms that:

• automatically detect patterns in data
• use the uncovered patterns to predict future data or
other outcomes of interest

• [Tom Mitchell] algorithms that:

• improve their performance (P)
• at some task (T)
• with experience (E)
Brief history of AI

(C) Dhruv Batra 30

 Brief history of AI
• “We propose that a 2 month, 10 man study of artificial
intelligence be carried out during the summer of 1956 at
Dartmouth College in Hanover, New Hampshire.”
• The study is to proceed on the basis of the conjecture that
every aspect of learning or any other feature of intelligence
can in principle be so precisely described that a machine
can be made to simulate it.
• An attempt will be made to find how to make machines
use language, form abstractions and concepts, solve kinds
of problems now reserved for humans, and improve
themselves.
• We think that a significant advance can be made in one or
more of these problems if a carefully selected group of
scientists work on it together for a summer.”
 Brief history of AI

• 1960-1980s: ”AI Winter”

• 1990s: Neural networks dominate, essentially because of

the discovery of the backpropagation for training neural
networks with two or more layers

• 2000s: Kernel methods dominate, essentially because of

the instability of training neural networks

• 2010s: The comeback of neural networks, essentially

because of the discovery of deep learning
 Why are things working today?

• More compute
power

• More data Accuracy

• Better
algorithms /
models
Amount of Training Data
 ML in a nutshell

• Tens of thousands of machine learning algorithms

 Researchers publish hundreds new every year

• Decades of ML research oversimplified:

 Learn a mapping f from the input X to the output Y,
i.e.:

 Example: X are emails, Y: {spam, not-spam}

 ML in a nutshell
• Input: X (images, texts, emails…)

• Output: Y (spam or not-spam…)

• (Unknown) Target Function:

• (the “true” mapping / reality)

• Data
• ,…

• Model / Hypothesis Class

•
•
 ML in a nutshell

• Every machine learning algorithm has three

components:
 Representation / Model Class
 Evaluation / Objective Function
 Optimization
 Where does ML fit in?

Biology Applied
Neuroscience Maths

• Biology of learning • Optimization

• Inspiring paradigms • Linear algebra
• E.g.: neural networks • Derivatives
Machine • E.g.: local minimum
Learning

Computer
Statistics
Science

• Algorithms • Estimation techniques

• Data structures • Theoretical frameworks
• Complexity analysis • Optimality, efficiency
• E.g.: k-d trees • E.g.: Bayes rule
 Learning paradigms
• Standard learning paradigms:
 Supervised learning
 Unsupervised learning
 Semi-supervised learning
 Reinforcement learning

• Non-standard paradigms:
 Active learning
 Transfer learning
 Transductive learning
 Supervised learning
• We have a set of labeled training samples
• Example 1: object recognition in images annotated
with corresponding class labels

Car Person

Person Dog
Car
 Supervised learning
• Example 2: handwritten digit recognition (on the MNIST data
set)

• Images of 28 x 28 pixels
• We can represent each image as a vector x of 784 components
• We train a classifier such that:
 Supervised learning
• Example 2 (continued): handwritten digit recognition (on the
MNIST data set)

• Starting with a training set of about 60K images (about 6000 images
per class)
• … the error rate can go down to 0.23% (using convolutional neural
networks)
• Among the first (learning-based) systems used in a large-scale
commercial setting for postal code and bank cheque processing
 Supervised learning
• Example 3: face detection

• One approach consists of sliding a window over the image

• The goal is to classify each window into one of the two
possible classes: face or not-face
• The original problem is transformed into a classification
problem
 Supervised learning
• Example 3: face detection

• We start with a set of face images with different variations

such as age, gender, illumination, pose, but no translations
• … and a larger set of images that do not contain full faces
 Supervised learning
• Example 4: spam detection

• The task is to classify an email into spam or not-spam

• The occurrence of the word “Dollars” is a good indicator of spam
• A possible representation is a vector of word frequencies
We count the words…
obtaining X
 The spam detection algorithm

Confidence /
performance
guarantee?

Why linear
combination?
Why these words?
Where do the weights
come from?
 Supervised learning
• Example 5: predicting stock prices on the market

• The goal is to predict the price at a future date, for example in a

few days
• This is a regression task, since the output is continuous
 Supervised learning
• Example 6: image difficulty prediction [Ionescu et al. CVPR2016]

• The goal is to predict the time necessary for a human to solve a

visual search task (data set available for project 2!)
• This is a regression task, since the output is continuous
 Canonical forms of supervised
learning problems
• Classification

• Regression
 Age estimation in images
• Classification?

• Regression?
What age?
The supervised learning paradigm
 Supervised learning models
• Naive Bayes (lecture 2)
• k-Nearest Neighbors (lecture 3)
• Decision trees and random forests (lecture 4)
• Support Vector Machines (lecture 5, 6)
• Kernel methods (lecture 5)
• Kernel Ridge Regression (lecture 5)
• Neural networks (lectures 7, 8, 9)
• Many others…
 Unsupervised learning
• We have an unlabeled training set of samples
• Example 1: clustering images based on similarity
 Unsupervised learning
• Example 1: clustering MNIST images based on
similarity [Georgescu et al. ICIP2019]
 Unsupervised learning
• Example 2: unsupervised features learning
 Unsupervised learning
• Example 2: unsupervised features learning for
abnormal event detection [Ionescu et al. CVPR2019]
 Unsupervised learning
• Example 3: clustering mammals by family, species, etc.

• The task is to generate the phylogenetic tree based on DNA

 Canonical forms of unsupervised
learning problems
• Clustering

• Dimensionality Reduction
 Unsupervised learning models
• K-means clustering (lecture 10, 11)
• DBScan (lecture 12)
• Hierarchical clustering (lecture 12)
• Principal Component Analysis (lecture 13)
• t-Distributed Stochastic Neighbor Embedding
(lecture 13)
• Hidden Markov Models
• Many others…
 Semi-supervised learning
• We have a training set of samples that are partially
annotated with class labels
• Example 1: object recognition in images, some of
which are annotated with corresponding class labels

Car Dog
Person
 Reinforcement learning
• How does it work?
• The system learns intelligent behavior using a
reinforcement signal (reward)
• The reward is given after several actions are taken (it
does not come after every action)
• Time matters (data is sequential, not i.i.d.)
• The actions of the system can influence the data
 Reinforcement learning
• Example 1: learning to play Go
• +/- reward for winning / losing the game
 Reinforcement learning
• Example 2: teaching a robot to ride a bike
• +/- reward for moving forward / falling
 Reinforcement learning
• Example 3: learning to play Pong from image pixels
• +/- reward for increasing
• personal / adversary score
Reinforcement learning paradigm
Formalizing as Markov Decision
Process
Formalizing as Markov Decision
Process
 Formalizing as Markov Decision
Process
• Solution based on dynamic programming (small
graphs) or approximation (large graphs)
• Goal: select the actions that maximize the total final
reward
• The actions can have long-term consequences
• Sacrificing the immediate reward can lead to higher
rewards on the long term
 Formalizing as Markov Decision
Process

• AlphaGo example:
 Narrator 1: “That’s a very strange move”
 Narrator 2: “I thought it was a mistake”
 But actually, “the move turned the course of the
game. AlphaGo went on to win Game Two, and at
the post-game press conference, Lee Sedol was in
shock.”
 https://www.wired.com/2016/03/two-moves-alphago-
lee-sedol-redefined-future/
 Active learning
• Given a large set of unlabeled samples, we have to
choose a small subset for annotation in order to
obtain a good classification model
 Transfer learning
• Starting with a model trained for a certain task /
domain, use the model for a different task / domain

More specific object classes,

face recognition,
texture classification, etc.
 Transfer learning
• Adapt the model to specific test samples
• Example 1: facial expression recognition [Georgescu et al. Access2019]
 Transfer learning
• Example 2: zero-shot learning
Many interesting applications, but…
• What is ethical and what is not?
• Trolley paradox
Many interesting applications, but…
• What is ethical and what is not?
• Trolley paradox
Many interesting applications, but…
• What is ethical and what is not?
• Trolley paradox
• http://moralmachine.mit.edu
Bibliography