0452538 - ÚTIA 2016 RIV CH eng C - Conference Paper (international conference)
Grim, Jiří - Pudil, P.
Mixtures of Product Components versus Mixtures of Dependence Trees.
Computational Intelligence. Cham: Springer, 2016, s. 365-382. Studies in Computational Intelligence, 620. ISBN 978-3-319-26393-9.
[IJCCI 2014 - International Joint Conference on Computational Intelligence (Rome/Italy). Rome (IT), 22.10.2014-24.10.2014]
R&D Projects: GA ČR(CZ) GA14-02652S
Grant - others:GA ČR(CZ) GAP403/12/1557
Institutional support: RVO:67985556
Keywords : Product mixtures * Mixtures of Dependence Trees * EM algorithm
Subject RIV: BD - Theory of Information
Result website:
http://library.utia.cas.cz/separaty/2016/RO/grim-0452538.pdf
DOI: https://doi.org/10.1007/978-3-319-26393-9_22

Mixtures of product components assume independence of variables given the index of the component. They can be efficiently estimated from data by means of EM algorithm and have some other useful properties. On the other hand, by considering mixtures of dependence trees, we can explicitly describe the statistical relationship between pairs of variables at the level of individual components and therefore approximation power of the resulting mixture may essentially increase. However, we have found in application to classification of numerals that both models perform comparably and the contribution of dependence-tree structures to the log-likelihood criterion decreases in the course of EM iterations. Thus the optimal estimate of dependence-tree mixture tends to reduce to a simple product mixture model.

Permanent Link: http://hdl.handle.net/11104/0257056