Online learning in discrete hidden Markov models

Roberto C. Alamino; Nestor Caticha

doi:10.1063/1.2423274

Online learning in discrete hidden Markov models

Roberto C. Alamino^*, Nestor Caticha

^*Corresponding author for this work

College of Engineering and Physical Sciences

Research output: Chapter in Book/Published conference output › Conference publication

Abstract

We present and analyze three different online algorithms for learning in discrete Hidden Markov Models (HMMs) and compare their performance with the Baldi-Chauvin Algorithm. Using the Kullback-Leibler divergence as a measure of the generalization error we draw learning curves in simplified situations and compare the results. The performance for learning drifting concepts of one of the presented algorithms is analyzed and compared with the Baldi-Chauvin algorithm in the same situations. A brief discussion about learning and symmetry breaking based on our results is also presented.

Original language	English
Title of host publication	Bayesian inference and maximum entropy methods In science and engineering
Editors	Ali Mohammad-Djafari
Publisher	AIP
Pages	187-194
Number of pages	8
ISBN (Print)	978-0-7354-0371-6
DOIs	https://doi.org/10.1063/1.2423274
Publication status	Published - 29 Dec 2006
Event	Bayesian inference and maximum entropy methods In science and engineering - Paris, France Duration: 8 Jul 2006 → 13 Jul 2006

Publication series

Name	AIP conference proceedings
Publisher	AIP
Volume	872
ISSN (Print)	0094-243X
ISSN (Electronic)	1551-7616

Conference

Conference	Bayesian inference and maximum entropy methods In science and engineering
Country/Territory	France
City	Paris
Period	8/07/06 → 13/07/06

Bibliographical note

Keywords

Bayesian algorithm
generalization error
HMMs
online algorithm

Access to Document

10.1063/1.2423274

Online learning in discrete hidden Markov models
© 2007 The Authors
Accepted author manuscript, 670 KB

Cite this

@inproceedings{46c8a4a27eb84c10b7294c50f7bd2458,

title = "Online learning in discrete hidden Markov models",

abstract = "We present and analyze three different online algorithms for learning in discrete Hidden Markov Models (HMMs) and compare their performance with the Baldi-Chauvin Algorithm. Using the Kullback-Leibler divergence as a measure of the generalization error we draw learning curves in simplified situations and compare the results. The performance for learning drifting concepts of one of the presented algorithms is analyzed and compared with the Baldi-Chauvin algorithm in the same situations. A brief discussion about learning and symmetry breaking based on our results is also presented.",

keywords = "Bayesian algorithm, generalization error, HMMs, online algorithm",

author = "Alamino, {Roberto C.} and Nestor Caticha",

note = "{\textcopyright} 2007 The Authors; Bayesian inference and maximum entropy methods In science and engineering ; Conference date: 08-07-2006 Through 13-07-2006",

year = "2006",

month = dec,

day = "29",

doi = "10.1063/1.2423274",

language = "English",

isbn = "978-0-7354-0371-6",

series = "AIP conference proceedings",

publisher = "AIP",

pages = "187--194",

editor = "Ali Mohammad-Djafari",

booktitle = "Bayesian inference and maximum entropy methods In science and engineering",

}

Alamino, RC & Caticha, N 2006, Online learning in discrete hidden Markov models. in A Mohammad-Djafari (ed.), Bayesian inference and maximum entropy methods In science and engineering. AIP conference proceedings, vol. 872, AIP, pp. 187-194, Bayesian inference and maximum entropy methods In science and engineering, Paris, France, 8/07/06. https://doi.org/10.1063/1.2423274

TY - GEN

T1 - Online learning in discrete hidden Markov models

AU - Alamino, Roberto C.

AU - Caticha, Nestor

PY - 2006/12/29

Y1 - 2006/12/29

N2 - We present and analyze three different online algorithms for learning in discrete Hidden Markov Models (HMMs) and compare their performance with the Baldi-Chauvin Algorithm. Using the Kullback-Leibler divergence as a measure of the generalization error we draw learning curves in simplified situations and compare the results. The performance for learning drifting concepts of one of the presented algorithms is analyzed and compared with the Baldi-Chauvin algorithm in the same situations. A brief discussion about learning and symmetry breaking based on our results is also presented.

AB - We present and analyze three different online algorithms for learning in discrete Hidden Markov Models (HMMs) and compare their performance with the Baldi-Chauvin Algorithm. Using the Kullback-Leibler divergence as a measure of the generalization error we draw learning curves in simplified situations and compare the results. The performance for learning drifting concepts of one of the presented algorithms is analyzed and compared with the Baldi-Chauvin algorithm in the same situations. A brief discussion about learning and symmetry breaking based on our results is also presented.

KW - Bayesian algorithm

KW - generalization error

KW - HMMs

KW - online algorithm

UR - http://www.scopus.com/inward/record.url?scp=33845739779&partnerID=8YFLogxK

UR - http://scitation.aip.org/content/aip/proceeding/aipcp/10.1063/1.2423274

U2 - 10.1063/1.2423274

DO - 10.1063/1.2423274

M3 - Conference publication

AN - SCOPUS:33845739779

SN - 978-0-7354-0371-6

T3 - AIP conference proceedings

SP - 187

EP - 194

BT - Bayesian inference and maximum entropy methods In science and engineering

A2 - Mohammad-Djafari, Ali

PB - AIP

T2 - Bayesian inference and maximum entropy methods In science and engineering

Y2 - 8 July 2006 through 13 July 2006

ER -

Online learning in discrete hidden Markov models

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

The evolution of learning systems: to Bayes or not to be

Cite this