About: MSVMpack is a Multi-class Support Vector Machine (M-SVM) package. It is dedicated to SVMs which can handle more than two classes without relying on decomposition methods and implements the four M-SVM models from the literature: Weston and Watkins M-SVM, Crammer and Singer M-SVM, Lee, Lin and Wahba M-SVM, and the M-SVM2 of Guermeur and Monfrini.

• Added Matlab interface.

About: MultiBoost is a multi-purpose boosting package implemented in C++. It is based on the multi-class/multi-task AdaBoost.MH algorithm [Schapire-Singer, 1999]. Basic base learners (stumps, trees, products, Haar filters for image processing) can be easily complemented by new data representations and the corresponding base learners, without interfering with the main boosting engine.

• A new fast (sublinear in the number of instances) stump algorithm is implemented. The gain in time is proportional to the sparsity of the features (it is significant when a lot of instances take the most frequent feature value). See Section B.2 in the documentation.
• A parametrized early stopping option is added in --traintest mode. We stop if the (smoothed) test error does not improve for a certain number of iterations. See Section 4.1.3 in the documentation.

About: A broad collection of script-friendly command-line tools for machine learning and data mining tasks. (The command-line tools wrap functionality from a C++ class library.)

See the change log at http://waffles.sourceforge.net/changelog.html

About: A platform-independent C++ framework for machine learning, graphical models, and computer vision research and development.

Version 1.5.1:

• Bug fixes and performance improvements in drwnPCA and drwnKMeans

Version 1.5:

• Win32 threading implementation (drwnThreadPool)
• Added standard command line option for setting random seed
• Made drwnPersistentStorage thread safe (on Linux and Mac OS X)
• Added drwnAverageRegions function
• Added fast superpixel code (drwnFastSuperpixels)
• Implemented drwnPersistentRecord interface for drwnSuperpixelContainer
• Enhanced drwnSuperpixelContainer with additional member functions
• Added image inpainting routines (drwnInPaint)
• Bug fixes and performance improvements

Version 1.4:

• dense and sparse linear program solver
• upgraded to Eigen 3.1.1
• sparse dot product
• thread safe persistent storage
• improved installation documentation
• bug fixes and performance improvements

About: This project is a C++ toolkit containing machine learning algorithms and tools that facilitate creating complex software in C++ to solve real world problems.

In addition to some bug fixes, this release also brings the following notable improvements to the library:

• The SURF feature extraction tool has higher matching accuracy than in previous dlib releases.
• The cutting plane optimizer is now faster
• A new tool for computing the singular value decomposition of very large matrices
• A new tool for performing canonical correlation analysis on large datasets
• A new tool for easily writing parallel for loops

About: The SHOGUN machine learning toolbox's focus is on large scale learning methods with focus on Support Vector Machines (SVM), providing interfaces to python, octave, matlab, r and the command line.

This release also contains several enhancements, cleanups and bugfixes:

Features

• Linear Time MMD two-sample test now works on streaming-features, which allows to perform tests on infinite amounts of data. A block size may be specified for fast processing. The below features were also added. By Heiko Strathmann.
• It is now possible to ask streaming features to produce an instance of streamed features that are stored in memory and returned as a CFeatures* object of corresponding type. See CStreamingFeatures::get_streamed_features().
• New concept of artificial data generator classes: Based on streaming features. First implemented instances are CMeanShiftDataGenerator and CGaussianBlobsDataGenerator. Use above new concepts to get non-streaming data if desired.
• Accelerated projected gradient multiclass logistic regression classifier by Sergey Lisitsyn.
• New CCSOSVM based structured output solver by Viktor Gal
• A collection of kernel selection methods for MMD-based kernel two- sample tests, including optimal kernel choice for single and combined kernels for the linear time MMD. This finishes the kernel MMD framework and also comes with new, more illustrative examples and tests. By Heiko Strathmann.
• Alpha version of Perl modular interface developed by Christian Montanari.
• New framework for unit-tests based on googletest and googlemock by Viktor Gal. A (growing) number of unit-tests from now on ensures basic funcionality of our framework. Since the examples do not have to take this role anymore, they should become more ilustrative in the future.
• Changed the core of dimension reduction algorithms to the Tapkee library.

Bugfixes

• Fix for shallow copy of gaussian kernel by Matt Aasted.
• Fixed a bug when using StringFeatures along with kernel machines in cross-validation which cause an assertion error. Thanks to Eric (yoo)!
• Fix for 3-class case training of MulticlassLibSVM reported by Arya Iranmehr that was suggested by Oksana Bayda.
• Fix for wrong Spectrum mismatch RBF construction in static interfaces reported by Nona Kermani.
• Fix for wrong include in SGMatrix causing build fail on Mac OS X (thanks to @bianjiang).
• Fixed a bug that caused kernel machines to return non-sense when using custom kernel matrices with subsets attached to them.
• Fix for parameter dictionary creationg causing dereferencing null pointers with gaussian processes parameter selection.
• Fixed a bug in exact GP regression that caused wrong results.
• Fixed a bug in exact GP regression that produced memory errors/crashes.
• Fix for a bug with static interfaces causing all outputs to be -1/+1 instead of real scores (reported by Kamikawa Masahisa).

Cleanup and API Changes

• SGStringList is now based on SGReferencedData.
• "confidences" in context of CLabel and subclasses are now "values".
• CLinearTimeMMD constructor changes, only streaming features allowed.
• CDataGenerator will soon be removed and replaced by new streaming- based classes.
• SGVector, SGMatrix, SGSparseVector, SGSparseVector, SGSparseMatrix refactoring: Now contains load/save routines, relevant functions from CMath, and implementations went to .cpp file.

About: The CTBN-RLE is a C++ package of executables and libraries for inference and learning algorithms for continuous time Bayesian networks (CTBNs).

Markov decision processes added (Kan & Shelton 2008) [ctmdp.h]

Mean field inference added (Cohn, El-Hay, Friedman, & Kupferman 2009) [meanfieldinf.h]

Factored uniformization for filtering added (Celikkaya & Shelton 2011) [uniformizedfactoredinf.h]

Auxilliary Gibbs sampling added (Rao & Teh 2011) [gibbsauxsampler.h]

Multi-threading for EM added

many speed improvements

unit testing improved [tst/]

new demo "main" programs added [demo/]

file format changed to XML-ish format (with old methods still there for conversion)

matrix switched to Eigen package (with option to return to old matrix)

glpk now included

initial cmake functionality

About: The scikit-learn aims to provide state of the art standard machine learning algorithms in Python.

Update for 0.13.1

About: The GPML toolbox is a flexible and generic Octave 3.2.x and Matlab 7.x implementation of inference and prediction in Gaussian Process (GP) models.

We now support inference on large datasets using the FITC approximation for non-Gaussian likelihoods for EP and Laplace's approximation. New likelihood functions: mixture likelihood, Poisson likelihood, label noise. We added two MCMC samplers.

About: A Tool for Embedding Strings in Vector Spaces

Support for positional n-grams with shift (similar to weighted-degree kernel with shift) has been added. Several minor bugs have been fixed.

About: libDAI provides free & open source implementations of various (approximate) inference methods for graphical models with discrete variables, including Bayesian networks and Markov Random Fields.

Release 0.3.1 fixes various bugs. The issues on 64-bit Windows platforms have been fixed and libDAI now offers full 64-bit support on all supported platforms (Linux, Mac OSX, Windows).

About: Mulan is an open-source Java library for learning from multi-label datasets. Multi-label datasets consist of training examples of a target function that has multiple binary target variables. This means that each item of a multi-label dataset can be a member of multiple categories or annotated by many labels (classes). This is actually the nature of many real world problems such as semantic annotation of images and video, web page categorization, direct marketing, functional genomics and music categorization into genres and emotions.

Learners

• BinaryRelevance.java: improved data handling that avoids copying the entire input space, leading to important speedups in case of large datasets and very large number of labels.
• RAkEL.java: updated technical information, added a check for the case where the number of labels is less or equal than the size of the subset.
• MultiLabelKNN.java: now checks whether the number of instances is less than the number of requested nearest neighbors.
• Addition of AdaBoostMH.java, an explicit implementation of AdaBoost.MH as combination of AdaBoostM1 and IncludeLabelsClassifier.
• Addition of MLPTO.java, the Multi Label Probabilistic Threshold Optimizer (MLTPTO) thresholding technique.
• Addition of ApproximateExampleBasedFMeasureOptimizer.java, an approximate method for the maximization of example-based F-measure.

Measures/Evaluation

• Addition of Specificity measure (example-based, micro/macro label-based)
• Addition of Mean Average Interpolated Precision (MAiP), Geometric Mean Average Precision (GMAP), Geometric Mean Average Interpolated Precision (GMAiP).
• New methods for stratified multi-label evaluation.
• Added support for outputting per label results for all measures that implement the MacroAverageMeasure interface.
• Simplifying the "strictness" issue of information retrieval measures, by adopting specific assumptions (outlined in the new class InformationRetrievalMeasures.java) to handle special cases, instead of the less clear and useful solution of outputting NaN and the less realistic solution or ignoring special cases.

Bug fixes

• Bug fix in LabelsBuilder.java.
• Bug fix in Ranker.java.
• Bug-fix in ThresholdPrediction.java.
• Fix for bug occurring when loading the XSD for mulan data outside the command-line environment (e.g. web applications).
• Javadoc comment updates.

API changes

• Upgrade to Java 1.6
• Upgrade to JUnit 4.10
• Upgrade to Weka 3.7.6.

Miscellaneous

• Meaningful messages are now shown when a DataLoadException is thrown.
• PT6(PT6Transformation.java): renamed to IncludeLabelsTransformation.java.
• MultiLabelInstances now support serialization, as needed by the improved binary relevance transformation.
• BinaryRelevanceAttributeEvaluator.java: updated according to latest BR improvements.

About: A Java framework for statistical analysis and classification of biological sequences

February 2, 2012: Jstacs 2.0 released

Jstacs 2.0 changes many names and the structure of several packages. It is not code-compatible with Jstacs 1.5 and earlier

RESTRUCTURING and RENAMING:

former ScoringFunction, NormalizableScoringFunction, Model

• new base-interface SequenceScore
• new sub-interface StatisticalModel of SequenceScore for all statistical models with sub-iterfaces DifferentiableStatisticalModel and TrainableStatisticalModel
• new interface DifferentiableSequenceScore replaces ScoringFunction
• new interface DifferentiableStatisticalModel replaces NormalizableScoringFunction
• new interface TrainableStatisticalModel replaces Model
• new abstract class AbstractDifferentiableSequenceScore
• new abstract class AbstractDifferentiableStatisticalModel replaces AbstractNormalizableScoringFunction
• new abstract class AbstractTrainableStatisticalModel replaces AbstractModel
• former Models renamed to TrainSM
• former ScoringFunction renamed to DiffSS or DiffSM
• getProbFor removed from TrainableStatisticalModel (former Model) and conceptually replaced by getLogProbFor
• getLogScore(Sequence,int,int) with changed meaning of arguments: getLogScore(Sequence,start,end) instead of getLogScore(Sequence,start,length)
• isTrained() replaced by common method isInitialized()

Parameters and Results

• new super-class of Parameters and Results: AnnotatedEntity
• common list-type for Parameters and Results: AnnotatedEntityList
• Renaming: CollectionParameter -> SelectionParameter, MultiSelectionCollectionParameter -> MultiSelectionParameter, new super-class AbstractSelectionParameter
• major refactoring due to common hierarchy and code-cleanup
• lazy evaluation of Parameter/ParameterSet hierarchies moved from ParameterSet (loadParameters()) to ParameterSetContainer (constructor on class)
• SubclassFinder adapted to lazy evaluation

performance measures

• new abstract super-class AbstractPerformanceMeasure of all performance measures
• new interface NumericalPerformanceMeasure for all performance measures that return a single number (as opposed, e.g., to curves)
• new class PerformanceMeasureParameterSet for a collection of general performance measures
• new class NumericalPerformanceMeasureParameterSet for a collection of NumericalPerformanceMeasures
• used in evaluate-method of AbstractClassifier and in ClassifierAssessments

further changes

• Sample renamed to DataSet
• evaluate and evaluateAll in AbstractClassifier joined
• new class IndependentProductDiffSS as super-class of IndepedentProductDiffSM (former IndependentProductScoringFunction)
• new class UniformDiffSS as super-class of UniformDiffSM (former UniformScoringFunction)

NEW FUNCTIONALITY:

• multi-threaded implementation of Baum-Welch and Viterbi training of hidden Markov models
• new Interface Singleton that can be used for singleton instances to save memory, current examples: DNAAlphabet, DNAAlphabetContainer, ProteinAlphabet
• added ProteinAlphabet
• added possibility to use NaN-values with ContinuousAlphabets
• added ArbitraryFloatSequence including static methods for DataSet creation for cases where double-precision is not needed
• new performance measure MaximumFMeasure
• access to Parameters in ParameterSets and Results in ResultSets by name
• emitDataSet in BayesianNetworkDiffSM
• new static method Time.getTimeInstance that returns UserTime or RealTime depending on availability of shared lib
• SubclassFinder allows for adding own base packages
• new method overlaps() in LocatedSequenceAnnotationWithLength
• AbstractTerminationCondition used in ScoreClassifier and sub-classes
• public method propagateESS in HMMFactory
• new method generateLog in DirichletMRG for drawing log-values
• added DifferentiableStatisticalModelFactory

BUGFIXES/IMPROVEMENTS:

• bugfix in propagation of equivalent sample size in HMMFactory
• bugfix in random initialization of BasicHigherOrderTransition
• improved Alignment implementation
• SafeOutputStream with new static factory method getSafeOutputStream, write methods now work on Objects

DOCUMENTATION:

• improved Javadocs in many classes and packages
• new Cookbook with extensive documentation and explanation

MISC:

• output of NonParsableException more verbose
• Exceptions in multi-threaded code now lead to exit of program instead of only stopping the thread
• update of RServe/RClient

About: Link Prediction Made Easy

v1.2.2

• Fixed MAJOR issue related to github migration several months ago. The original github commit neglected to import empty folders. This caused parts of the project compilation procedure to fail. Any users of LPmade who downloaded the most recent version from github over the last several months would have encountered this build error and should download the most recent version. This change updates the network library makefile to create the empty folders and gets around the issue. Very sorry to anybody that this may have inconvenienced, but thanks for hanging in there if you diagnosed and solved it yourself.

• Fixed issue with auroc on 32-bit architectures that caused integer wraparounds that produced incorrect results.

About: Locally Weighted Projection Regression (LWPR) is a recent algorithm that achieves nonlinear function approximation in high dimensional spaces with redundant and irrelevant input dimensions. At its [...]

Version 1.2.4

• Corrected typo in lwpr.c (wrong function name for multi-threaded helper function on Unix systems) Thanks to Jose Luis Rivero

About: Infrastructure for representing, manipulating and analyzing transaction data and frequent patterns.

Initial Announcement on mloss.org.

About: The SSA Toolbox is an efficient, platform-independent, standalone implementation of the Stationary Subspace Analysis algorithm with a friendly graphical user interface and a bridge to Matlab. Stationary Subspace Analysis (SSA) is a general purpose algorithm for the explorative analysis of non-stationary data, i.e. data whose statistical properties change over time. SSA helps to detect, investigate and visualize temporal changes in complex high-dimensional data sets.

• Various bugfixes.

About: CARP: The Clustering Algorithms’ Referee Package

Added generalized overlap, more metrics for comparing partitionings

About: The SUMO Toolbox is a Matlab toolbox that automatically builds accurate surrogate models (also known as metamodels or response surface models) of a given data source (e.g., simulation code, data set, script, ...) within the accuracy and time constraints set by the user. The toolbox minimizes the number of data points (which it selects automatically) since they are usually expensive.

Incremental update, fixing some cosmetic issues, coincides with JMLR publication.

About: The DL-Learner framework contains several algorithms for supervised concept learning in Description Logics (DLs) and OWL.

See http://dl-learner.org/wiki/ChangeLog.