A study of online and blockwise updating of the EM algorithm for Gaussian mixtures

Einbeck, Jochen; Bonetti, Daniel

A study of online and blockwise updating of the EM algorithm for Gaussian mixtures

Einbeck, Jochen; Bonetti, Daniel

Authors

Professor Jochen Einbeck jochen.einbeck@durham.ac.uk
Professor

Daniel Bonetti

Contributors

Thomas Kneib
Editor

Fabian Sobotka
Editor

Jan Fahrenholz
Editor

Henriette Irmer
Editor

Abstract

A variant of the EM algorithm for the estimation of multivariate Gaussian mixtures, which allows for online as well as blockwise updating of sequentially obtained parameter estimates, is investigated. Several dierent update schemes are considered and compared, and the benets of articially performing EM in batches, even though all data are available, are discussed.

Citation

Einbeck, J., & Bonetti, D. (2014). A study of online and blockwise updating of the EM algorithm for Gaussian mixtures. In T. Kneib, F. Sobotka, J. Fahrenholz, & H. Irmer (Eds.), 29th International Workshop on Statistical Modelling, 14-18 July 2014, Göttingen, Germany ; proceedings (35-38)

Conference Name	29th International Workshop on Statistical Modelling
Conference Location	Göttingen
Publication Date	Jul 18, 2014
Deposit Date	Sep 29, 2014
Publicly Available Date	Oct 7, 2014
Volume	2
Pages	35-38
Series Title	Proceedings of the 29th International Workshop on Statistical Modelling. Göttingen, Germany, 14-18 July 2014
Book Title	29th International Workshop on Statistical Modelling, 14-18 July 2014, Göttingen, Germany ; proceedings.
Keywords	Multivariate Gaussian mixtures, Maximum Likelihood, Incremental EM.
Public URL	https://durham-repository.worktribe.com/output/1154424
Related Public URLs	http://www.statmod.org/workshops_archive_proceedings_2014.htm