Alfons Juan, José García-Hernández, and Enrique Vidal.  
EM Initialisation for Bernoulli Mixture Learning.  
In A. Fred et al., editor, Proc. of the SSPR-SPR04, LNCS 3138, 
Lecture Notes in Computer Science, 3138, pages 635-643, 
Lisbon (Portugal), August 2004. Springer.

Abstract

Mixture modelling is a hot area in pattern recognition. This paper focuses 
on the use of Bernoulli mixtures for binary data and, in particular, for 
binary images. More specifically, six EM initialisation techniques are 
described and empirically compared on a classification task of handwritten 
Indian digits. Somehow surprisingly, we have found that a relatively good 
initialisation for Bernoulli prototypes is to use slightly perturbed 
versions of the hypercube centre.