nimfa A Python Library for Nonnegative Matrix Factorizationhttp://mloss.orgUpdates and additions to nimfa A Python Library for Nonnegative Matrix FactorizationenThu, 22 Mar 2012 02:38:18 -0000nimfa A Python Library for Nonnegative Matrix Factorization 1.0<html><p>Nimfa is an open-source Python library that provides a unified interface to nonnegative matrix factorization algorithms. It includes implementations of state-of-the-art factorization methods, initialization approaches, and quality scoring. Both dense and sparse matrix representation are supported. </p> <p>Matrix Factorization Methods </p> <ul> <li> BD - Bayesian nonnegative matrix factorization Gibbs sampler </li> <li> BMF - Binary matrix factorization </li> <li> ICM - Iterated conditional modes nonnegative matrix factorization </li> <li> LFNMF - Fisher nonnegative matrix factorization for learning local features </li> <li> LSNMF - Alternating nonnegative least squares matrix factorization using projected gradient method for subproblems </li> <li> NMF - Standard nonnegative matrix factorization with Euclidean / Kullback-Leibler update equations and Frobenius / divergence / connectivity cost functions </li> <li> NSNMF - Nonsmooth nonnegative matrix factorization </li> <li> PMF - Probabilistic nonnegative matrix factorization </li> <li> PSMF - Probabilistic sparse matrix factorization </li> <li> SNMF - Sparse nonnegative matrix factorization based on alternating nonnegativity constrained least squares </li> <li> SNMNMF - Sparse network-regularized multiple nonnegative matrix factorization </li> </ul> <p>Initialization Methods </p> <ul> <li> Random </li> <li> Fixed </li> <li> NNDSVD </li> <li> Random C </li> <li> Random VCol </li> </ul> <p>Quality Measures </p> <ul> <li> Distance </li> <li> Residuals </li> <li> Connectivity matrix </li> <li> Consensus matrix </li> <li> Entropy of the fitted NMF model </li> <li> Dominant basis components computation </li> <li> Explained variance </li> <li> Feature score computation representing its specificity to basis vectors </li> <li> Computation of most basis specific features for basis vectors </li> <li> Purity </li> <li> Residual sum of squares (rank estimation) </li> <li> Sparseness </li> <li> Cophenetic correlation coefficient of consensus matrix (rank estimation) </li> <li> Dispersion </li> <li> Factorization rank estimation </li> <li> Selected matrix factorization method specific </li> </ul> <p>Utils </p> <ul> <li> Fitted factorization model tracker across multiple runs </li> <li> Residuals tracker across multiple factorizations / runs </li> </ul></html>Marinka Zitnik, Blaz ZupanThu, 22 Mar 2012 02:38:18 -0000 matrix factorizationinitialization methodsquality measures