Recursive unsupervised learning of finite mixture models

    Research output: Contribution to journalArticleAcademicpeer-review

    202 Citations (Scopus)

    Abstract

    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.
    Original languageUndefined
    Pages (from-to)651-656
    Number of pages6
    JournalIEEE transactions on pattern analysis and machine intelligence
    Volume26
    Issue number5
    DOIs
    Publication statusPublished - 2004

    Keywords

    • finite mixtures
    • Online (recursive) estimation
    • unsupervised learning
    • EWI-17484
    • METIS-220268
    • IR-48638
    • Model selection
    • EM-algorithm

    Cite this