Skip to main navigation Skip to search Skip to main content

Harmonious competition learning for gaussian mixtures

  • Harbin Institute of Technology

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This paper proposes a novel automatic model selection algorithm for learning Gaussian mixtures. Unlike EM, we shall further increase the negative entropy of the posterior of latent variables to exert an indirect effect on model selection. The increase of negative entropy can be interpreted as a competition, which corresponds to an annihilation of those components with insufficient data to support. More importantly, this competition only depends on the data itself. Additionally, we seamlessly integrate parameter estimation and model selection into a single algorithm, which can be applied to any kind of parametric mixture model solved by an EM algorithm. Experiments involving Gaussian mixtures show the efficiency of our approach on model selection.

Original languageEnglish
Title of host publicationIntelligence Science and Big Data Engineering - 4th International Conference, IScIDE 2013, Revised Selected Papers
PublisherSpringer Verlag
Pages385-392
Number of pages8
ISBN (Print)9783642420566
DOIs
StatePublished - 2013
Event4th International Conference on Intelligence Science and Big Data Engineering, IScIDE 2013 - Beijing, China
Duration: 31 Jul 20132 Aug 2013

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8261 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference4th International Conference on Intelligence Science and Big Data Engineering, IScIDE 2013
Country/TerritoryChina
CityBeijing
Period31/07/132/08/13

Keywords

  • Expectation maximization
  • Gaussian mixture model
  • Harmonious competition learning
  • Model selection

Fingerprint

Dive into the research topics of 'Harmonious competition learning for gaussian mixtures'. Together they form a unique fingerprint.

Cite this