EM Initialisation for Bernoulli Mixture Learning

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. Key words: Mixture Models, EM Algorithm, Multivariate Bernoulli Distribution, Initialisation Techniques, Binary Data, Indian Digits

• Type: Conference
• Year: 2004
• Where: SSPR
• Authors: Alfons Juan, José García-Hernández, Enrique Vidal
• Keywords: Binary Data, Indian Digits, Initialisation Techniques, Pattern Recognition, SSPR 2004