Severity of local maxima for the em algorithm: Experiences with hierarchical latent class models

Yi Wang*, Nevin L. Zhang

*Corresponding author for this work

Research output: Chapter in Book/Conference Proceeding/ReportConference Paper published in a bookpeer-review

Abstract

It is common knowledge that the EM algorithm can be trapped at local maxima and consequently fails to reach global maxima. We empirically investigate the severity of this problem in the context of hierarchical latent class (HLC) models. Our experiments were run on HLC models where dependency between neighboring variables is strong. (The reason for focusing on this class of models will be made clear in the main text.) We first ran EM from randomly generated single starting points, and observed that (1) the probability of hitting global maxima is generally high, (2) it increases with the strength of dependency and sample sizes, and (3) it decreases with the amount of extreme probability values. We also observed that, at high dependence strength levels, local maxima are far apart from global ones in terms of likelihoods. Those imply that local maxima can be reliably avoided by running EM from a few starting points and hence are not a serious issue. This is confirmed by our second set of experiments.

Original languageEnglish
Title of host publicationProceedings of the 3rd European Workshop on Probabilistic Graphical Models, PGM 2006
Pages301-308
Number of pages8
Publication statusPublished - 2006
Event3rd European Workshop on Probabilistic Graphical Models, PGM 2006 - Prague, Czech Republic
Duration: 12 Sept 200615 Sept 2006

Publication series

NameProceedings of the 3rd European Workshop on Probabilistic Graphical Models, PGM 2006

Conference

Conference3rd European Workshop on Probabilistic Graphical Models, PGM 2006
Country/TerritoryCzech Republic
CityPrague
Period12/09/0615/09/06

Fingerprint

Dive into the research topics of 'Severity of local maxima for the em algorithm: Experiences with hierarchical latent class models'. Together they form a unique fingerprint.

Cite this