About: A fast and scalable graph-based clustering algorithm based on the eigenvectors of the nonlinear 1-Laplacian.

• improved optimization of ncut and rcut criterion
• optimized eigenvector initialization
• changed default values for number of runs
• several internal optimizations
• made console output more informative

About: The Accord.NET Framework is a .NET machine learning framework combined with audio and image processing libraries completely written in C#. It is a complete framework for building production-grade computer vision, computer audition, signal processing and statistics applications even for commercial use. A comprehensive set of sample applications provide a fast start to get up and running quickly, and an extensive online documentation helps fill in the details.

For a complete list of changes, please see the full release notes at the release details page at:

https://github.com/accord-net/framework/releases/tag/v3.8.0

About: A generalized version of spectral clustering using the graph p-Laplacian.

various internal optimizations

About: A mutual information library for C and Mex bindings for MATLAB. Aimed at feature selection, and provides simple methods to calculate mutual information, conditional mutual information, entropy, conditional entropy, Renyi entropy/mutual information, and weighted variants of Shannon entropies/mutual informations. Works with discrete distributions, and expects column vectors of features.

Fixed a Windows compilation bug. MIToolbox v3 should now compile using Visual Studio.

About: NaN-toolbox is a statistics and machine learning toolbox for handling data with and without missing values.

Changes in v.3.1.2 - improve configuration and build system - support of more platforms (including Octave 4.2.0) improved

Changes in v.3.0.3 - improve compatibility for Octave on Windows

Changes in v.3.0.1 - fix packaging for octave

Changes in v.2.8.5 - bug fix: trimmean - compiler support for gcc-5 and clang - fix typos

For details see the CHANGELOG at http://pub.ist.ac.at/~schloegl/matlab/NaN/CHANGELOG

About: FEAST provides implementations of common mutual information based filter feature selection algorithms (mim, mifs, mrmr, cmim, icap, jmi, disr, fcbf, etc), and an implementation of RELIEF. Written for C/C++ & Matlab.

Major refactoring of FEAST to improve speed and portability.

• FEAST now clones the input data if it's floating point and discretises it to unsigned ints once in a single pass. This improves the speed by about 30%.
• FEAST now has unsigned int entry points which avoid this discretisation and are much faster if the data is already categorical.
• Added weighted feature selection algorithms to FEAST which can be used for cost-sensitive feature selection.
• Added a Java API using JNI.
• FEAST now returns the internal score for each feature according to the criterion. Available in all three APIs.
• Rearranged the repository to make it easier to work with. Header files are now in `include`, source in `src`, the MATLAB API is in `matlab/` and the Java API is in `java/`.
• FEAST now compiles cleanly using `-std=c89 -Wall -Werror`.

About: learn optimized scoring systems using MATLAB and the CPLEX Optimization Studio

Initial Announcement on mloss.org.

About: This is an optimization library based on Social Impact Theory(SITO). The optimizer works in the same way as PSO and GA.

bug removed

About: Nowadays this is very popular to use the deep architectures in machine learning. Deep Belief Networks (DBNs) are deep architectures that use a stack of Restricted Boltzmann Machines (RBM) to create a powerful generative model using training data. DBNs have many abilities such as feature extraction and classification that are used in many applications including image processing, speech processing, text categorization, etc. This paper introduces a new object oriented toolbox with the most important abilities needed for the implementation of DBNs. According to the results of the experiments conducted on the MNIST (image), ISOLET (speech), and the 20 Newsgroups (text) datasets, it was shown that the toolbox can learn automatically a good representation of the input from unlabeled data with better discrimination between different classes. Also on all the aforementioned datasets, the obtained classification errors are comparable to those of the state of the art classifiers. In addition, the toolbox supports different sampling methods (e.g. Gibbs, CD, PCD and our new FEPCD method), different sparsity methods (quadratic, rate distortion and our new normal method), different RBM types (generative and discriminative), GPU based, etc. The toolbox is a user-friendly open source software in MATLAB and Octave and is freely available on the website.

New in toolbox

• Using GPU in Backpropagation
• Revision of some demo scripts
• Function approximation with multiple outputs
• Feature extraction with GRBM in first layer

About: testing mloss.org

Initial Announcement on mloss.org.

About: Local high-order regularization for semi-supervised learning

Initial Announcement on mloss.org.

About: A Matlab implementation of Sparse PCA using the inverse power method for nonlinear eigenproblems.

• Added deflation scheme to compute multiple principal components
• Several internal runtime and memory optimizations
• API change: sparsePCA.m is now used to compute multiple components; use computeTradeOffCurve.m to reproduce the examples in the NIPS paper

About: APRIL-ANN toolkit (A Pattern Recognizer In Lua with Artificial Neural Networks). This toolkit incorporates ANN algorithms (as dropout, stacked denoising auto-encoders, convolutional neural networks), with other pattern recognition methods as hidden makov models (HMMs) among others.

• Updated home repository link to follow april-org github organization.
• Improved serialize/deserialize functions, reimplemented all the serialization procedure.
• Added exceptions support to LuaPkg and APRIL-ANN, allowing to capture C++ errors into Lua code.
• Added set class.
• Added series class.
• Added data_frame class, similar to Python Pandas DataFrame.
• Serialization and deserilization have been updated with more robust and reusable API, implemented in util.serialize() and util.deserialize() functions.
• Added matrix.ext.broadcast utility (similar to broadcast in numpy).
• Added ProbablisitcMatrixANNComponent, which allow to implement probabilistic mixtures of posteriors and/or likelihoods.
• Added batch normalization ANN component.
• Allowing matrix.join to add new axis.
• Added methods prod(), cumsum() and cumprod() at matrix classes.
• Added methods count_eq() and count_neq() at matrix classes.
• Serializable objects API have been augmented with methods ctor_name() and
  ctor_params() in Lua, refered to luaCtorName() and luaCtorParams() in C++.
• Added cast.to to dynamic cast C++ objects pushed into Lua, allowing to convert base class objects into any of its derived classes.
• Added matrix.sparse as valid values for targets in ann.loss.mse and
  ann.loss.cross_entropy.
• Changed matrix metamethods __index and __newindex, allowing to use
  matrix objects with standard Lua operator[].
• Added matrix.masked_fill and matrix.masked_copy matrix.
• Added matrix.indexed_fill and matrix.indexed_copy matrix.
• Added ann.components.probabilistic_matrix, and its corresponding specializations ann.components.left_probabilistic_matrix and
  ann.components.right_probabilistic_matrix.
• Added operator[] in the right side of matrix operations.
• Added ann.components.transpose.
• Added max_gradients_norm in traianble.supervised_trainer, to avoid gradients exploding.
• Added ann.components.actf.sparse_logistic a logistic activation function with sparsity penalty.
• Simplified math.add, math.sub, ... and other math extensions for reductions, their original behavior can be emulated by using bind function.
• Added bind function to freeze any positional argument of any Lua function.
• Function stats.boot uses multiple_unpack to allow a table of sizes and the generation of multiple index matrices.
• Added multiple_unpack Lua function.
• Added __tostring metamethod to numeric memory blocks in Lua.
• Added dataset.token.sparse_matrix, a dataset which allow to traverse by rows a sparse matrix instance.
• Added matrix.sparse.builders.dok, a builder which uses the Dictionary-of-Keys format to construct a sparse matrix from scratch.
• Added method data to numeric matrix classes.
• Added methods values, indices, first_index to sparse matrix class.
• Fixed bugs when reading bad formed CSV files.
• Fixed bugs at statistical distributions.
• FloatRGB bug solved on equal (+=, -=, ...) operators. This bug affected ImageRGB operations such as resize.
• Solved problems when chaining methods in Lua, some objects end to be garbage collected.
• Improved support of strings in auto-completion (rlcompleter package).
• Solved bug at SparseMatrix<T> when reading it from a file.
• Solved bug in Image<T>::rotate90_cw methods.
• Solved bug in SparseMatrix::toDense() method.

C/C++

• Better LuaTable accessors, using [] operator.
• Implementation of matrix __index, __newindex and __call metamethods in C++.
• Implementation of matProd(), matCumSum() and matCumProd() functions.
• Implementation of matCountEq() and matCountNeq() functions for
  Matrix<T>.
• Updated matrix_ext_operations.h to change API of matrix operations. All functions have been overloaded to accept an in-place operation and another version which receives a destination matrix.
• Adding iterators to language models.
• Added MatrixScalarMap2 which receives as input2 a SparaseMatrix instance. This functions needs to be generalized to work with CPU and CUDA.
• The method SparseMatrix<T>::fromDenseMatrix() uses a DOKBuilder object to build the sparse matrix.
• The conversion of a Matrix<T> into a SparseMatrix<T> has been changed from a constructor overload to the static method
  SparseMatrix<T>::fromDenseMatrix().
• Added support for IPyLua.
• Optimized matrix access for confusion matrix.
• Minor changes in class.lua.
• Improved binding to avoid multiple object copies when pushing C++ objects.
• Added Git commit hash and compilation time.

About: LOFS is a software toolbox for online streaming feature selection

Initial Announcement on mloss.org.

About: The Cognitive Foundry is a modular Java software library of machine learning components and algorithms designed for research and applications.

• General:
  • Upgraded MTJ to 1.0.3.
• Common:
  • Added package for hash function computation including Eva, FNV-1a, MD5, Murmur2, Prime, SHA1, SHA2
  • Added callback-based forEach implementations to Vector and InfiniteVector, which can be faster for iterating through some vector types.
  • Optimized DenseVector by removing a layer of indirection.
  • Added method to compute set of percentiles in UnivariateStatisticsUtil and fixed issue with percentile interpolation.
  • Added utility class for enumerating combinations.
  • Adjusted ScalarMap implementation hierarchy.
  • Added method for copying a map to VectorFactory and moved createVectorCapacity up from SparseVectorFactory.
  • Added method for creating square identity matrix to MatrixFactory.
  • Added Random implementation that uses a cached set of values.
• Learning:
  • Implemented feature hashing.
  • Added factory for random forests.
  • Implemented uniform distribution over integer values.
  • Added Chi-squared similarity.
  • Added KL divergence.
  • Added general conditional probability distribution.
  • Added interfaces for Regression, UnivariateRegression, and MultivariateRegression.
  • Fixed null pointer exception that can happen in K-means with an empty cluster.
  • Fixed name of maxClusters property on AgglomerativeClusterer (was called maxMinDistance).
• Text:
  • Improvements to LDA Gibbs sampler.

About: This code is provided by Jun Wan. It is used in the Chalearn one-shot learning gesture challenge (round 2). This code includes: bag of features, 3D MoSIFT-based features (i.e. 3D MoSIFT, 3D EMoSIFT and 3D SMoSIFT), and the MFSK feature.

Initial Announcement on mloss.org.