Fast sampling from a Hidden Markov Model posterior for large data

Rasmus Bonnevie, Lars Kai Hansen

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

Abstract

Hidden Markov Models are of interest in a broad set of applications including modern data driven systems involving very large data sets. However, approximate inference methods based on Bayesian averaging are precluded in such applications as each sampling step requires a full sweep over the data. We show that Approximate Bayesian Computation offers an interesting alternative for approximate sampling from the posterior distribution. In particular we use recent advances in moment based methods for HMM estimation to generate summary statistics for Approximate Bayesian Computation for large data sets offering fast access to approximate posterior samples. In a specific example we see that the new scheme is a hundred times faster than conventional Markov Chain Monte Carlo sampling using the Forward-backward method.
Original languageEnglish
Title of host publicationProceedings of the 2014 IEEE International Workshop on Machine Learning for Signal Processing (MLSP)
EditorsMamadou Mboup, Tülay Adali , Éric Moreau, Jan Larsen
Number of pages6
PublisherIEEE
Publication date2014
ISBN (Print)978-1-4799-3694-6
DOIs
Publication statusPublished - 2014
Event24th IEEE International Workshop on Machine Learning for Signal Processing - Reims Centre des Congrès, Reims, France
Duration: 21 Sep 201424 Sep 2014
Conference number: 24
http://mlsp2014.conwiz.dk/home.htm
http://mlsp2014.conwiz.dk/home.htm

Conference

Conference24th IEEE International Workshop on Machine Learning for Signal Processing
Number24
LocationReims Centre des Congrès
CountryFrance
CityReims
Period21/09/201424/09/2014
Internet address

Keywords

  • Bioengineering
  • Communication, Networking and Broadcast Technologies
  • Computing and Processing
  • Engineering Profession
  • Signal Processing and Analysis
  • Approximate Bayesian Computation
  • Approximation methods
  • Bayes methods
  • Computational modeling
  • Hidden Markov Models
  • Markov Chain Monte Carlo
  • Markov processes
  • Moment based learning
  • Monte Carlo methods
  • Proposals

Fingerprint Dive into the research topics of 'Fast sampling from a Hidden Markov Model posterior for large data'. Together they form a unique fingerprint.

Cite this