• Bugfixes:
  - Fix cookbook examples for website

Features

• Drop all <math.h> function calls [Viktor Gal]
• Use c++11 std::isnan, std:isfinite, std::isinf [Viktor Gal]

Bugfixes

• Port ipython notebooks to be python3 compatible [Viktor Gal]
• Use the shogun-static library on Windows when linking the interface library [Viktor Gal]
• Fix python typemap when compiling with MSVC [Viktor Gal]
• Fix ShogunConfig.cmake paths [Viktor Gal]
• Fix meta example parser bug in parallel builds [Esben Sørig]

Bugfixes:

• Fix installation of examples on WIN32 [Viktor Gal]

Bugfixes

• Install headers of GPL models when LICENSE_GPL_SHOGUN is enabled [Viktor Gal]
• Always turn on LIBSHOGUN_BUILD_STATIC when compiling with MSVC [Viktor Gal]
• Fix ipython notebook errors [Viktor Gal]

• This release is dedicated for Heiko's successful PhD defense!

• Add conda-forge packages, to get prebuilt binaries via the cross-platform conda package manager [Dougal Sutherland]

• Change interface cmake variables to INTERFACE_*

• Move GPL code to gpl submodule [Heiko Strathmann]

Features

• Enable using BLAS/LAPACK from Eigen by default [Viktor Gal]
• Add iterators to SGVector and SGMatrix [Viktor Gal]
• Significantly lower the runtime of KernelPCA (GSoC '17) [Michele Mazzoni]
• Refactor FisherLDA and LDA solvers (GSoC '17) [Michele Mazzoni]
• Add automated test for trained model serialization (GSoC '17) [Michele Mazzoni]
• Enable SWIG director classes by default [Viktor Gal]
• Vectorize DotFeatures covariance/mean calculation [Michele Mazzoni]
• Support for premature stopping of model training (GSoC '17) [Giovanni De Toni]
• Add support for observable variables (GSoC '17) [Giovanni De Toni]
• Use TFLogger to serialize observed variables for TensorBoard (GSoC '17) [Giovanni De Toni]
• Drop CMath::dot and SGVector::dot and use linalg::dot [Viktor Gal]
• Added class probabilities for BaggingMachine (GSoC '17) [Olivier Nguyen]

Bugfixes

• Fix transpose bug in Ruby typemap for matrices [Elias Saalmann]
• Fix MKL detection and linking; use mkl_rt when available [Viktor Gal]
• Fix Windows static linking [Viktor Gal]
• Fix SWIG interface compilation on Windows [qcrist]
• Fix CircularBuffer bug that broke parsing of big CSV and LibSVM files #1991 [Viktor Gal]
• Fix R interface when using clang to compile the interface [Viktor Gal]

• Add native MS Windows support [Viktor Gal]
• Shogun requires the compiler to support C++11 features
• Shogun cloud online: Jupyter notebook with Shogun from the browser, https://cloud.shogun.ml

Features

• LDA now supports 32, 64 and 128 bit floating point numbers [Chris Goldsworthy]
• Add SHOGUN_NUM_THREADS enviroment variable to control the number of threads
  used by the models in runtime [Viktor Gal]
• Added Scala Interface to the build [Abhinav Rai]
• Major re-writing and API changes in kernel statistical hypothesis
  testing framework, significant speed up in permutation test for
  quadratic time MMD, new kernel selection algorithms for quadratic time MMD [Soumyajit De]

Bugfixes:

• Fix build error of R interface for R>=3.3.0, #3460 [Heiko Strathmann]
• Make the code compatible with Eigen 3.3.0 [Viktor Gal]
• Fix number of CPUs detected on Linux [Viktor Gal]
• Fix multi-threading in KMeansBase [Viktor Gal]
• Make ExponentialARDKernel thread-safe [Viktor Gal]
• Make PRNG thread-safe [Viktor Gal]
• Fix python interface when using libshogun compiled with OpenMP [Viktor Gal]
• Fix CART to work with cross-validation [Fernando Iglesias]

Cleanup, efficiency updates, and API Changes:

• Port multi-threading to use OpenMP backend in Kernel [Viktor Gal]
• Fix false sharing in EuclideanDistance [Viktor Gal]
• Fix out of source build of the whole project [Viktor Gal]
• Add LIBSHOGUN cmake flag to turn off libshogun compilation [Viktor Gal]
• Export Shogun target with cmake to enable to build modular interfaces to a
  pre-compiled libshogun on the system without requiring to compile
  libshogun itself [Viktor Gal]

Notes

• Contains major rewrite and clean-up of developer documentation in doc/readme [Heiko Strathmann, Lea Götz]
• Known issue: Octave multithreaded crashes, currently bindings are initialized single-threaded, #3772 [Heiko Strathmann]

Features

• GSoC 2016 project of Saurabh Mahindre: Major efficiency improvements for KMeans, LARS, Random Forests, Bagging, KNN.
• Add new Shogun cookbook for documentation and testing across all target languages [Heiko Strathmann, Sergey Lisitsyn, Esben Sorig, Viktor Gal].
• Added option to learn CombinedKernel weights with GP approximate inference [Wu Lin].
• LARS now supports 32, 64, and 128 bit floating point numbers [Chris Goldsworthy].

Bugfixes:

• Fix gTest segfaults with GCC >= 6.0.0 [Björn Esser].
• Make Java and CSharp install-dir configurable [Björn Esser].
• Autogenerate modshogun.rb with correct module-suffix [Björn Esser].
• Fix KMeans++ initialization [Saurabh Mahindre].

Cleanup, efficiency updates, and API Changes:

• Make Eigen3 a hard requirement. Bundle if not found on system. [Heiko Strathmann]
• Drop ALGLIB (GPL) dependency in CStatistics and ship CDFLIB (public domain) instead [Heiko Strathmann]
• Drop p-value estimation in model-selection [Heiko Strathmann]
• Static interfaces have been removed [Viktor Gal]
• New base class ShiftInvariantKernel of which GaussianKernel inherits [Rahul De].

NOTE

This version contains a new CMake option USE_GPL_SHOGUN, which when set to OFF will exclude all GPL codes from Shogun [Heiko Strathmann].

This is a new feature and cleanup release.

Features:

• Added GEMPLP for approximate inference to the structured output framework [Jiaolong Xu].
• Effeciency improvements of the FITC framework for GP inference (FITC_Laplce, FITC, VarDTC) [Wu Lin].
• Added optimisation of inducing variables in sparse GP inference [Wu Lin].
• Added optimisation methods for GP inference (Newton, Cholesky, LBFGS, ...) [Wu Lin].
• Added Automatic Relevance Determination (ARD) kernel functionality for variational GP inference [Wu Lin].
• Updated Notebook for variational GP inference [Wu Lin].
• New framework for stochastic optimisation (L1/2 loss, mirror descent, proximal gradients, adagrad, SVRG, RMSProp, adadelta, ...) [Wu Lin].
• New Shogun meta-language for automatically generating code listings in all target languages [Esben Sörig].
• Added periodic kernel [Esben Sörig].
• Add gradient output functionality in Neural Nets [Sanuj Sharma].

Bugfixes:

• Fixes for java_modular build using OpenJDK [Björn Esser].
• Catch uncaught exceptions in Neural Net code [Khaled Nasr].
• Fix build of modular interfaces with SWIG 3.0.5 on MacOSX [Björn Esser].
• Fix segfaults when calling delete[] twice on SGMatrix-instances [Björn Esser].
• Fix for building with full-hardening-(CXX|LD)FLAGS [Björn Esser].
• Patch SWIG to fix a problem with SWIG and Python >= 3.5 [Björn Esser].
• Add modshogun.rb: make sure narray is loaded before modshogun.so [Björn Esser].
• set working-dir properly when running R (#2654) [Björn Esser].

Cleanup, efficiency updates, and API Changes:

• Added GPU based dot-products to linalg [Rahul De].
• Added scale methods to linalg [Rahul De].
• Added element wise products to linalg [Rahul De].
• Added element-wise unary operators in linalg [Rahul De].
• Dropped parameter migration framework [Heiko Strathmann].
• Disabled Python integration tests by default [Sergey Lisitsyn, Heiko Strathmann].

• This release features the work of our 8 GSoC 2014 students [student; mentors]:
  • OpenCV Integration and Computer Vision Applications [Abhijeet Kislay; Kevin Hughes]
  • Large-Scale Multi-Label Classification [Abinash Panda; Thoralf Klein]
  • Large-scale structured prediction with approximate inference [Jiaolong Xu; Shell Hu]
  • Essential Deep Learning Modules [Khaled Nasr; Sergey Lisitsyn, Theofanis Karaletsos]
  • Fundamental Machine Learning: decision trees, kernel density estimation [Parijat Mazumdar ; Fernando Iglesias]
  • Shogun Missionary & Shogun in Education [Saurabh Mahindre; Heiko Strathmann]
  • Testing and Measuring Variable Interactions With Kernels [Soumyajit De; Dino Sejdinovic, Heiko Strathmann]
  • Variational Learning for Gaussian Processes [Wu Lin; Heiko Strathmann, Emtiyaz Khan]
• This release also contains several cleanups and bugfixes:
  • Features:
    • New Shogun project description [Heiko Strathmann]
    • ID3 algorithm for decision tree learning [Parijat Mazumdar]
    • New modes for PCA matrix factorizations: SVD & EVD, in-place or reallocating [Parijat Mazumdar]
    • Add Neural Networks with linear, logistic and softmax neurons [Khaled Nasr]
    • Add kernel multiclass strategy examples in multiclass notebook [Saurabh Mahindre]
    • Add decision trees notebook containing examples for ID3 algorithm [Parijat Mazumdar]
    • Add sudoku recognizer ipython notebook [Alejandro Hernandez]
    • Add in-place subsets on features, labels, and custom kernels [Heiko Strathmann]
    • Add Principal Component Analysis notebook [Abhijeet Kislay]
    • Add Multiple Kernel Learning notebook [Saurabh Mahindre]
    • Add Multi-Label classes to enable Multi-Label classification [Thoralf Klein]
    • Add rectified linear neurons, dropout and max-norm regularization to neural networks [Khaled Nasr]
    • Add C4.5 algorithm for multiclass classification using decision trees [Parijat Mazumdar]
    • Add support for arbitrary acyclic graph-structured neural networks [Khaled Nasr]
    • Add CART algorithm for classification and regression using decision trees [Parijat Mazumdar]
    • Add CHAID algorithm for multiclass classification and regression using decision trees [Parijat Mazumdar]
    • Add Convolutional Neural Networks [Khaled Nasr]
    • Add Random Forests algorithm for ensemble learning using CART [Parijat Mazumdar]
    • Add Restricted Botlzmann Machines [Khaled Nasr]
    • Add Stochastic Gradient Boosting algorithm for ensemble learning [Parijat Mazumdar]
    • Add Deep contractive and denoising autoencoders [Khaled Nasr]
    • Add Deep belief networks [Khaled Nasr]
  • Bugfixes:
    • Fix reference counting bugs in CList when reference counting is on [Heiko Strathmann, Thoralf Klein, lambday]
    • Fix memory problem in PCA::apply_to_feature_matrix [Parijat Mazumdar]
    • Fix crash in LeastAngleRegression for the case D greater than N [Parijat Mazumdar]
    • Fix memory violations in bundle method solvers [Thoralf Klein]
    • Fix fail in library_mldatahdf5.cpp example when http://mldata.org is not working properly [Parijat Mazumdar]
    • Fix memory leaks in Vowpal Wabbit, LibSVMFile and KernelPCA [Thoralf Klein]
    • Fix memory and control flow issues discovered by Coverity [Thoralf Klein]
    • Fix R modular interface SWIG typemap (Requires SWIG >= 2.0.5) [Matt Huska]
  • Cleanup and API Changes:
    • PCA now depends on Eigen3 instead of LAPACK [Parijat Mazumdar]
    • Removing redundant and fixing implicit imports [Thoralf Klein]
    • Hide many methods from SWIG, reducing compile memory by 500MiB [Heiko Strathmann, Fernando Iglesias, Thoralf Klein]

we are pleased to announce Shogun 3.2.0 !

This release also contains several cleanups and bugfixes:

• Features:
  • Fully support python3 now
  • Add mini-batch k-means [Parijat Mazumdar]
  • Add k-means++ for more details see the notebook [Parijat Mazumdar]
  • Add sub-sequence string kernel [lambday]
• Bugfixes:
  • Compile fixes for upcoming swig3.0
  • Speedup for gaussian process' apply()
  • Improve unit / integration test checks
  • libbmrm uninitialized memory reads
  • libocas uninitialized memory reads
  • Octave 3.8 compile fixes [Orion Poplawski]
  • Fix java modular compile error [Bjoern Esser]