Armadillo is a C++ linear algebra library (matrix maths) aiming towards a good balance between speed and ease of use. The API is deliberately similar to Matlab's. Integer, floating point, and complex numbers are supported, as well as a subset of trigonometric and statistics functions. Various matrix decompositions are provided through optional integration with LAPACK and ATLAS numerics libraries. A delayed evaluation approach, based on template meta-programming, is used (during compile time) to combine several operations into one and reduce or eliminate the need for temporaries.
scikits.learn is a Python module that integrates classic machine learning algorithms in the tightly-knit world of scientific Python packages. It aims to provide simple and efficient solutions to learning problems that are accessible to everybody and reusable in various contexts: machine-learning as a versatile tool for science and engineering.
pyuds is a Python library for measuring uncertainty in the Dempster-Shafer theory of evidence. The functionals supported are the Generalized Hartley (GH) uncertainty functional, Generalized Shannon (GS) uncertainty functional, and Aggregate Uncertainty (AU) functional. The library can be utilized either through its API, or through a user-friendly Web interface.
SHOGUN is a machine learning toolbox whose focus is on large scale kernel methods and especially on Support Vector Machines (SVM). It provides a generic SVM object interfacing to several different SVM implementations, all making use of the same underlying, efficient kernel implementations. Apart from SVMs and regression, SHOGUN also features a number of linear methods like Linear Discriminant Analysis (LDA), Linear Programming Machine (LPM), (Kernel) Perceptrons, and algorithms to train hidden Markov models. SHOGUN can be used from within C++, Matlab, R, Octave, and Python.
Treba is a commandline tool for training, decoding, and calculating with weighted (probabilistic) finite state automata (WFSA/PFSA). Training algorithms include Baum-Welch (EM), Viterbi training, and Baum-Welch augmented with deterministic annealing. Treba is optimized for speed and numerical stability, and training algorithms can be run multi-threaded on hardware with multiple cores/CPUs. Forward, backward, and Viterbi decoding are supported. Automata for training/decoding are read from a text file, or can be generated randomly or with uniform transition probabilities with different topologies (ergodic or fully connected, Bakis or left-to-right, or deterministic). Observations used for training or decoding are read from text files compatible with AT&T finite state tools and OpenFST.
Thinknowlogy is grammar-based software, designed to utilize the Natural Laws of Intelligence in grammar, in order to create intelligence through natural language in software. This is demonstrated by programming in natural language, reasoning in natural language and drawing conclusions (more detailed than scientific solutions), making assumptions (with self-adjusting level of uncertainty), asking questions (about gaps in the knowledge), and detecting conflicts in the knowledge. It builds semantics autonomously (with no vocabularies or words lists), detecting some cases of semantic ambiguity. It is multi-grammar, proving that Natural Laws of Intelligence are universal.
Milk is a machine learning toolkit in Python. Its focus is on supervised classification with several classifiers available: SVMs (based on libsvm), k-NN, random forests, and decision trees. It also performs feature selection. These classifiers can be combined in many ways to form different classification systems. For unsupervised learning, milk supports k-means clustering and affinity propagation.
MyMediaLite is a lightweight, multi-purpose library of recommender system algorithms. It addresses the two most common scenarios in collaborative filtering: rating prediction (e.g. on a scale of 1 to 5 stars), and item prediction from implicit feedback (e.g. from clicks or purchase actions). It contains dozens of recommender engines, including state-of-the-art matrix factorization methods. It also supports real-time updates to the recommender engines, storing engines to disk and reloading them again, and several evaluation measures to compare the accuracy of different recommender system methods. Three command-line programs that offer most of the functionality contained in the library are included.