Fast change point detection in switching dynamics using a hidden Markov model of prediction experts

J. Kohlmorgen, S. Lemm, Klaus Muller, S. Liehr, K. Pawelzik

Research output: Chapter in Book/Report/Conference proceedingConference contribution

9 Citations (Scopus)

Abstract

We present a framework for modeling switching dynamics from a time series that allows for a fast on-line detection of dynamical mode changes. The method is based on a hidden Markov model (HMM) of prediction experts. The predictors are trained by Expectation Maximization (EM) and by using an annealing schedule for the HMM state probabilities. This leads to a segmentation of the time series into different dynamical modes and a simultaneous specialization of the prediction experts on the segments. In a second step, an input-density estimator is generated for each expert. It can simply be computed from the data subset assigned to the respective expert. In conjunction with the HMM state probabilities, this allows for a very fast on-line detection of mode changes: change points are detected as soon as the incoming input data stream contains sufficient information to indicate a change in the dynamics.

Original languageEnglish
Title of host publicationIEE Conference Publication
Place of PublicationStevenage, United Kingdom
PublisherIEE
Pages204-209
Number of pages6
Volume1
Edition470
ISBN (Print)0852967217
Publication statusPublished - 1999 Dec 1
Externally publishedYes
EventProceedings of the 1999 the 9th International Conference on 'Artificial Neural Networks (ICANN99)' - Edinburgh, UK
Duration: 1999 Sep 71999 Sep 10

Other

OtherProceedings of the 1999 the 9th International Conference on 'Artificial Neural Networks (ICANN99)'
CityEdinburgh, UK
Period99/9/799/9/10

Fingerprint

Hidden Markov models
Time series
Annealing

ASJC Scopus subject areas

  • Electrical and Electronic Engineering

Cite this

Kohlmorgen, J., Lemm, S., Muller, K., Liehr, S., & Pawelzik, K. (1999). Fast change point detection in switching dynamics using a hidden Markov model of prediction experts. In IEE Conference Publication (470 ed., Vol. 1, pp. 204-209). Stevenage, United Kingdom: IEE.

Fast change point detection in switching dynamics using a hidden Markov model of prediction experts. / Kohlmorgen, J.; Lemm, S.; Muller, Klaus; Liehr, S.; Pawelzik, K.

IEE Conference Publication. Vol. 1 470. ed. Stevenage, United Kingdom : IEE, 1999. p. 204-209.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Kohlmorgen, J, Lemm, S, Muller, K, Liehr, S & Pawelzik, K 1999, Fast change point detection in switching dynamics using a hidden Markov model of prediction experts. in IEE Conference Publication. 470 edn, vol. 1, IEE, Stevenage, United Kingdom, pp. 204-209, Proceedings of the 1999 the 9th International Conference on 'Artificial Neural Networks (ICANN99)', Edinburgh, UK, 99/9/7.
Kohlmorgen J, Lemm S, Muller K, Liehr S, Pawelzik K. Fast change point detection in switching dynamics using a hidden Markov model of prediction experts. In IEE Conference Publication. 470 ed. Vol. 1. Stevenage, United Kingdom: IEE. 1999. p. 204-209
Kohlmorgen, J. ; Lemm, S. ; Muller, Klaus ; Liehr, S. ; Pawelzik, K. / Fast change point detection in switching dynamics using a hidden Markov model of prediction experts. IEE Conference Publication. Vol. 1 470. ed. Stevenage, United Kingdom : IEE, 1999. pp. 204-209
@inproceedings{654a19f93b934e71bb49979eefe25282,
title = "Fast change point detection in switching dynamics using a hidden Markov model of prediction experts",
abstract = "We present a framework for modeling switching dynamics from a time series that allows for a fast on-line detection of dynamical mode changes. The method is based on a hidden Markov model (HMM) of prediction experts. The predictors are trained by Expectation Maximization (EM) and by using an annealing schedule for the HMM state probabilities. This leads to a segmentation of the time series into different dynamical modes and a simultaneous specialization of the prediction experts on the segments. In a second step, an input-density estimator is generated for each expert. It can simply be computed from the data subset assigned to the respective expert. In conjunction with the HMM state probabilities, this allows for a very fast on-line detection of mode changes: change points are detected as soon as the incoming input data stream contains sufficient information to indicate a change in the dynamics.",
author = "J. Kohlmorgen and S. Lemm and Klaus Muller and S. Liehr and K. Pawelzik",
year = "1999",
month = "12",
day = "1",
language = "English",
isbn = "0852967217",
volume = "1",
pages = "204--209",
booktitle = "IEE Conference Publication",
publisher = "IEE",
edition = "470",

}

TY - GEN

T1 - Fast change point detection in switching dynamics using a hidden Markov model of prediction experts

AU - Kohlmorgen, J.

AU - Lemm, S.

AU - Muller, Klaus

AU - Liehr, S.

AU - Pawelzik, K.

PY - 1999/12/1

Y1 - 1999/12/1

N2 - We present a framework for modeling switching dynamics from a time series that allows for a fast on-line detection of dynamical mode changes. The method is based on a hidden Markov model (HMM) of prediction experts. The predictors are trained by Expectation Maximization (EM) and by using an annealing schedule for the HMM state probabilities. This leads to a segmentation of the time series into different dynamical modes and a simultaneous specialization of the prediction experts on the segments. In a second step, an input-density estimator is generated for each expert. It can simply be computed from the data subset assigned to the respective expert. In conjunction with the HMM state probabilities, this allows for a very fast on-line detection of mode changes: change points are detected as soon as the incoming input data stream contains sufficient information to indicate a change in the dynamics.

AB - We present a framework for modeling switching dynamics from a time series that allows for a fast on-line detection of dynamical mode changes. The method is based on a hidden Markov model (HMM) of prediction experts. The predictors are trained by Expectation Maximization (EM) and by using an annealing schedule for the HMM state probabilities. This leads to a segmentation of the time series into different dynamical modes and a simultaneous specialization of the prediction experts on the segments. In a second step, an input-density estimator is generated for each expert. It can simply be computed from the data subset assigned to the respective expert. In conjunction with the HMM state probabilities, this allows for a very fast on-line detection of mode changes: change points are detected as soon as the incoming input data stream contains sufficient information to indicate a change in the dynamics.

UR - http://www.scopus.com/inward/record.url?scp=0033323548&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0033323548&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0033323548

SN - 0852967217

VL - 1

SP - 204

EP - 209

BT - IEE Conference Publication

PB - IEE

CY - Stevenage, United Kingdom

ER -