Last Change: Fri Aug 04 07:00 PM 2006 J

Version 0.4.2

pyem is a python module build upon numpy and scipy
(see http://www.scipy.org/) for learning mixtures models
using Expectation Maximization. For now, only Gaussian
Mixture Models are implemented. Included features:
    
 * computation of Gaussian pdf for multi-variate Gaussian
 random vectors (spherical, diagonal and full covariance matrices)
 * Sampling of Gaussian Mixtures Models
 * Confidence ellipsoides with probability (fixed level of 
 0.39 for now)
 * Classic EM for Gaussian Mixture Models
 * K-mean based and random initialization for EM available

Has been tested on the following platforms:

 * Ubuntu dapper, bi Xeon 3.2 Ghz, 2 Go RAM
 python 2.4 + pyrex, numpy 1.0.b2SVN + scipy 0.5.1SVN, uses atlas3-sse2
 * Ubuntu dapper, pentium M 1.2 ghz,. 512 Mo Ram
 python 2.4 + pyrex, numpy 1.0.b2SVN + scipy 0.5.1SVN, uses atlas3-sse2
 * Ubuntu dapper, minimac (ppc G4 1.42 Ghz, 1Gb RAM)
 python 2.4 + pyrex, numpy 1.0.b2SVN + scipy 0.5.1SVN, uses atlas3-sse2
