Cluster: An Unsupervised Algorithm for Modeling Gaussian Mixtures

Charles A. Bouman
bouman@purdue.edu
School of Electrical and Computer Engineering
Purdue University
West Lafayette, IN 47907-1285

Cluster Software
Cluster is an unsupervised algorithm for modeling Gaussian mixtures that is based on the expectation (EM) algorithm and the minimum discription length (MDL) order estimation criteria. This program clusters feature vectors to produce a Gaussian mixture model. It also estimates the number of clusters directly from the data. The parameters of the Gaussian mixture model are store in an easy-to-read ASCII file format. The package also includes simple routines for performing ML classification and unsupervised clustering with classification with the resulting Gaussian mixture models.
README file
manual.pdf - Manual explaining how to use software and explaining theory
cluster-3.6.6.zip - Directory containing ANSI C-code designed to compile on a variety of platforms.
cluster-old - Previous releases of the software.
bibtex citation - If you would like to cite this software in you publications, you can use this bibtex citation.

Matlab cluster algorithm - New Matlab version of cluster available!!!