Zero knowledge hidden Markov model inference

J. M. Schwier, R. R. Brooks, Christopher Griffin, S. Bukkapatnam

Research output: Contribution to journalArticle

12 Citations (Scopus)

Abstract

Hidden Markov models (HMMs) are widely used in pattern recognition. HMM construction requires an initial model structure that is used as a starting point to estimate the model's parameters. To construct a HMM without a priori knowledge of the structure, we use an approach developed by Crutchfield and Shalizi that requires only a sequence of observations and a maximum data window size. Values of the maximum data window size that are too small result in incorrect models being constructed. Values that are too large reduce the number of data samples that can be considered and exponentially increase the algorithm's computational complexity. In this paper, we present a method for automatically inferring this parameter directly from training data as part of the model construction process. We present theoretical and experimental results that confirm the utility of the proposed extension.

Original languageEnglish (US)
Pages (from-to)1273-1280
Number of pages8
JournalPattern Recognition Letters
Volume30
Issue number14
DOIs
StatePublished - Oct 15 2009

Fingerprint

Hidden Markov models
Model structures
Pattern recognition
Computational complexity

All Science Journal Classification (ASJC) codes

  • Software
  • Signal Processing
  • Computer Vision and Pattern Recognition
  • Artificial Intelligence

Cite this

Schwier, J. M. ; Brooks, R. R. ; Griffin, Christopher ; Bukkapatnam, S. / Zero knowledge hidden Markov model inference. In: Pattern Recognition Letters. 2009 ; Vol. 30, No. 14. pp. 1273-1280.
@article{7301416fe7e941198bde074a7aa3b80b,
title = "Zero knowledge hidden Markov model inference",
abstract = "Hidden Markov models (HMMs) are widely used in pattern recognition. HMM construction requires an initial model structure that is used as a starting point to estimate the model's parameters. To construct a HMM without a priori knowledge of the structure, we use an approach developed by Crutchfield and Shalizi that requires only a sequence of observations and a maximum data window size. Values of the maximum data window size that are too small result in incorrect models being constructed. Values that are too large reduce the number of data samples that can be considered and exponentially increase the algorithm's computational complexity. In this paper, we present a method for automatically inferring this parameter directly from training data as part of the model construction process. We present theoretical and experimental results that confirm the utility of the proposed extension.",
author = "Schwier, {J. M.} and Brooks, {R. R.} and Christopher Griffin and S. Bukkapatnam",
year = "2009",
month = "10",
day = "15",
doi = "10.1016/j.patrec.2009.06.008",
language = "English (US)",
volume = "30",
pages = "1273--1280",
journal = "Pattern Recognition Letters",
issn = "0167-8655",
publisher = "Elsevier",
number = "14",

}

Schwier, JM, Brooks, RR, Griffin, C & Bukkapatnam, S 2009, 'Zero knowledge hidden Markov model inference', Pattern Recognition Letters, vol. 30, no. 14, pp. 1273-1280. https://doi.org/10.1016/j.patrec.2009.06.008

Zero knowledge hidden Markov model inference. / Schwier, J. M.; Brooks, R. R.; Griffin, Christopher; Bukkapatnam, S.

In: Pattern Recognition Letters, Vol. 30, No. 14, 15.10.2009, p. 1273-1280.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Zero knowledge hidden Markov model inference

AU - Schwier, J. M.

AU - Brooks, R. R.

AU - Griffin, Christopher

AU - Bukkapatnam, S.

PY - 2009/10/15

Y1 - 2009/10/15

N2 - Hidden Markov models (HMMs) are widely used in pattern recognition. HMM construction requires an initial model structure that is used as a starting point to estimate the model's parameters. To construct a HMM without a priori knowledge of the structure, we use an approach developed by Crutchfield and Shalizi that requires only a sequence of observations and a maximum data window size. Values of the maximum data window size that are too small result in incorrect models being constructed. Values that are too large reduce the number of data samples that can be considered and exponentially increase the algorithm's computational complexity. In this paper, we present a method for automatically inferring this parameter directly from training data as part of the model construction process. We present theoretical and experimental results that confirm the utility of the proposed extension.

AB - Hidden Markov models (HMMs) are widely used in pattern recognition. HMM construction requires an initial model structure that is used as a starting point to estimate the model's parameters. To construct a HMM without a priori knowledge of the structure, we use an approach developed by Crutchfield and Shalizi that requires only a sequence of observations and a maximum data window size. Values of the maximum data window size that are too small result in incorrect models being constructed. Values that are too large reduce the number of data samples that can be considered and exponentially increase the algorithm's computational complexity. In this paper, we present a method for automatically inferring this parameter directly from training data as part of the model construction process. We present theoretical and experimental results that confirm the utility of the proposed extension.

UR - http://www.scopus.com/inward/record.url?scp=68749114961&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=68749114961&partnerID=8YFLogxK

U2 - 10.1016/j.patrec.2009.06.008

DO - 10.1016/j.patrec.2009.06.008

M3 - Article

AN - SCOPUS:68749114961

VL - 30

SP - 1273

EP - 1280

JO - Pattern Recognition Letters

JF - Pattern Recognition Letters

SN - 0167-8655

IS - 14

ER -