mloss.org new software

scikitlearn 0.1

Fri, 19 Mar 2010 08:14:29 -0000

The focus of the scikit-learn is on having reliable and portable machine learning algorithm accessible in Python. Effort is put in reducing the dependencies and having excellent Python bindings, for instance by keeping a close look on the memory.

OpenOpt 0.28

Dmitrey Kroshko — Mon, 15 Mar 2010 19:12:17 -0000

Universal Python-written numerical optimization toolbox. Problems: NLP, LP, QP, SDP, SOCP, DFP(Non-linear Data Fit), NSP(nonsmooth), MILP, LSP, LLSP, MMP, GLP, MINLP etc. Connects to dozens of solvers (some are C- or Fortran-written).

Provides graphic output of convergence and some more numerical optimization "MUST HAVE" features.

Our another tool FuncDesigner allows to involve automatic differentiation + more convenient modelling of some optimization problems and SLEs (systems of linear equations, possibly sparse/overdetermined).

Pyriel 1.0

tom fawcett — Sun, 14 Mar 2010 03:51:18 -0000

Pyriel is an experimental rule learning system written in Python. Given a set of data labeled with class names, it will learn a set of data classification rules of the form:

If condition1 AND condition2 AND ... AND conditionN ==> CLASS

Pyriel has a number of desirable properties for data mining:

Because PRIE maximizes ROC performance, it naturally handles skewed datasets.
PRIE is able to handle multiple classes. It will attempt to optimize the combined AUC for any number of classes simultaneously.
PRIE's output is a single rulelist and thus is relatively intelligible and modular. To use this rulelist on a new unseen instance, the rules are evaluated sequentially and the first one matching determines the class and probability. Some data mining practitioners consider rulelists to be more intelligible than rulesets because only a single rule matches a new instance.
Because PRIE uses a rulelist whose rules are ordered decreasing by class likelihood, the rulelist may be used naturally with the ROC convex hull (Provost and Fawcett, 2001). In use, if operating conditions (class skew and relative error costs) are known, the rulelist can be truncated to eliminate rules that will never affect a classification decision.
PRIE handles numerical attributes naturally, using the ROC curve implicitly to identify promising discretizations. Other classification models may discretize variables in a preprocessing pass or may use techniques unrelated to model construction. PRIE considers every discretization of a continuous attribute to comprise a separate point in ROC space, and handles these the same as any other discrete attribute.
PRIE can handle set-valued attributes (Cohen, 1996), in which an attribute of an instance may take on a set of discrete values instead of a single one. Such features are useful, for example, in text classification domains in which the set may represent the "bag of words" of a text document.

PRIE is unusual in that it uses basic principles from rule learning and computational geometry to focus the search for promising rule combinations. The result is a system that can learn rulelists with high AUC scores.

ALGLIB 2.4.0

Sergey Bochkanov — Wed, 10 Mar 2010 14:07:35 -0000

ALGLIB is an open source numerical analysis library distributed under GPL 2+. It implements both general numerical algorithms and machine learning algorithms.

As for machine learning algorithms, ALGLIB implements: MLPs, early stopping ensembles, LDA, PCA, k-means++, several other data analysis algorithms. As for general numerical algorithms, ALGLIB implements linear algebra functions (subset of LAPACK), linear solvers, ODE solvers, nonlinear optimization, interpolation/fitting, integration, fast transforms.

ALGLIB can be used from C#, C, FreePascal, VBA and other languages. It is the only numerical analysis library which uses automatic translation to generate source code written in different programming languages with 100% identical functionality. So, C# version is written entirely in C#, C++ version is entirely C++, but they provide exactly the same interface.

dlib ml 17.26

Davis King — Sun, 07 Mar 2010 21:37:42 -0000

A C++ toolkit containing machine learning algorithms and tools that facilitate creating complex software in C++ to solve real world problems.

The library provides efficient implementations of the following algorithms:

support vector machines for classification
relevance vector machines for regression and classification
reduced set approximation of SV decision surfaces
online kernel RLS regression
online kernelized centroid estimation/one class classifier
online SVM classification
kernel k-means clustering
radial basis function networks
kernelized recursive feature ranking
Bayesian network inference using junction trees or MCMC

The library also comes with extensive documentation and example programs that walk the user through the use of these machine learning techniques.

dlib also comes with a fast matrix library that lets the user use a simple Matlab like syntax. It is also capable of using BLAS libraries such as ATLAS or the Intel MKL when available. Additionally, the use of BLAS is transparent to the user, that is, the dlib matrix object uses BLAS internally to optimize all the various forms of matrix multiplication while still allowing the user to use a simple Matlab like syntax.

Continuous Time Bayesian Network Reasoning and Learning Engine 1.0.2

Christian Shelton, Yu Fan, William Lam, Joon Lee, Jing Xu — Fri, 05 Mar 2010 23:45:20 -0000

CTBN-RLE is a continuous time Bayesian network reasoning and learning engine. A continuous time Bayesian network (CTBN) provides a compact (factored) description of a continuous-time Markov process. This software provides libraries and executables for most of the algorithms developed for CTBNs. For learning, CTBN-RLE implements structure and parameter learning for both complete and partial data. For inference, it implements exact inference and Gibbs and importance sampling approximate inference for any type of evidence pattern. Additionally, the library supplies visualization methods for graphically displaying CTBNs or trajectories of evidence.

Error Correcting Output Codes Library 0.1

Sergio Escalera, Oriol Pujol, Petia Radeva — Fri, 05 Mar 2010 16:49:12 -0000

In this work we present an open source Error-Correcting Output Codes (ECOC) library. The ECOC framework is a powerful tool to deal with multi-class categorization problems. This library contains both state-of-the-art coding (one-versus-one, one-versus-all, dense random, sparse random, DECOC, forest-ECOC, and ECOC-ONE) and decoding designs hamming, euclidean, inverse hamming, laplacian, Beta-density, attenuated, loss-based, probabilistic kernel-based, and loss-weighted) with the parameters defined by the authors, as well as the option to include your own coding, decoding, and base classifier.

PyBrain 0.3

Wed, 03 Mar 2010 15:00:08 -0000

PyBrain is a versatile machine learning library for Python. Its goal is to provide flexible, easy-to-use yet still powerful algorithms for machine learning tasks, including a variety of predefined environments and benchmarks to test and compare algorithms. Implemented algorithms include Long Short-Term Memory (LSTM), policy gradient methods, (multidimensional) recurrent neural networks and evolution strategies like CMA-ES, NES or FEM.

Jstacs 1.3.1

Jens Keilwagen, Jan Grau, Andre Gohr — Tue, 02 Mar 2010 14:15:46 -0000

Sequence analysis is one of the major subjects of bioinformatics. Several existing libraries combine the representation of biological sequences with exact and approximate pattern matching as well as alignment algorithms. We present Jstacs, an open source Java library, which focuses on the statistical analysis of biological sequences instead. Jstacs comprises an efficient representation of sequence data and provides implementations of many statistical models with generative and discriminative approaches for parameter learning. Using Jstacs, classifiers can be assessed and compared on test datasets or by cross-validation experiments evaluating several performance measures. Due to its strictly object-oriented design Jstacs is easy to use and readily extensible.

NaN toolbox 2.1.0

alois schloegl — Sun, 28 Feb 2010 22:00:21 -0000

The NaN-toolbox provides a number of statistics functions and machine learning methods for the use with Octave and Matlab. The functions can handle data with missing values encoded as NaNs, weighting of data samples, and multi-class classification (using a one-versus-rest scheme). There is a common interface to a number of different classification methods (including FDA, LDA, Naive Bayes, QDA, RDA, sparse classifiers, interfaces to some SVMs, regression/PLS, Wiener-Hopf).