Snob - http://www.cs.monash.edu.au/~dld/Snob.html

Snob (Wallace and Boulton, 1968) was probably the first (Bayesian) program to do clustering (or unsupervised learning or mixture modelling), and was also in the first published paper using the Minimum Message Length (MML) principle. The use of MML means that Snob is statistically consistent, converging to any true underlying model (or as closely as possible to any such thiug). The original (1968) Snob dealt with mixtures of Normal (Gaussian) and multi-state distributions. Snob has recently been extended (Wallace and Dowe,1994, 1996, 1997) to also permit mxitures of von Mises circular distributions (as found in proteins) and Poisson distributions.
Contributed by: David Dowe dld@cs.monash.edu.au

Note: This info converted from the original "The Data Mine" pages and pre-dates June 2001. Please remove this note if you update or check the info