Artigo Acesso aberto Revisado por pares

Recursive unsupervised learning of finite mixture models

2004; IEEE Computer Society; Volume: 26; Issue: 5 Linguagem: Inglês

10.1109/tpami.2004.1273970

ISSN

2160-9292

Autores

Zoran Živković, Ferdinand van der Heijden,

Tópico(s)

Machine Learning and Algorithms

Resumo

There are two open problems when finite mixture densities are used to model multivariate data: the selection of the number of components and the initialization. In this paper, we propose an online (recursive) algorithm that estimates the parameters of the mixture and that simultaneously selects the number of components. The new algorithm starts with a large number of randomly initialized components. A prior is used as a bias for maximally structured models. A stochastic approximation recursive learning algorithm is proposed to search for the maximum a posteriori (MAP) solution and to discard the irrelevant components.

Referência(s)