Distinguishing Hidden Markov Chains

被引:5
|
作者
Kiefer, Stefan [1 ]
Sistla, A. Prasad [2 ]
机构
[1] Univ Oxford, Oxford OX1 2JD, England
[2] Univ Illinois, Chicago, IL 60680 USA
来源
PROCEEDINGS OF THE 31ST ANNUAL ACM-IEEE SYMPOSIUM ON LOGIC IN COMPUTER SCIENCE (LICS 2016) | 2016年
基金
英国工程与自然科学研究理事会;
关键词
Hidden Markov chains; Labelled Markov chains; monitors; MODELS;
D O I
10.1145/2933575.2933608
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
Hidden Markov Chains (HMCs) are commonly used mathematical models of probabilistic systems. They are employed in various fields such as speech recognition, signal processing, and biological sequence analysis. Motivated by applications in stochastic runtime verification, we consider the problem of distinguishing two given HMCs based on a single observation sequence that one of the HMCs generates. More precisely, given two HMCs and an observation sequence, a distinguishing algorithm is expected to identify the HMC that generates the observation sequence. Two HMCs are called distinguishable if for every epsilon > 0 there is a distinguishing algorithm whose error probability is less than epsilon. We show that one can decide in polynomial time whether two HMCs are distinguishable. Further, we present and analyze two distinguishing algorithms for distinguishable HMCs. The first algorithm makes a decision after processing a fixed number of observations, and it exhibits two-sided error. The second algorithm processes an unbounded number of observations, but the algorithm has only one-sided error. The error probability, for both algorithms, decays exponentially with the number of processed observations. We also provide an algorithm for distinguishing multiple HMCs.
引用
收藏
页码:66 / 75
页数:10
相关论文
共 50 条
  • [31] Limit Theorems for the Sample Entropy of Hidden Markov Chains
    Han, Guangyue
    2011 IEEE INTERNATIONAL SYMPOSIUM ON INFORMATION THEORY PROCEEDINGS (ISIT), 2011, : 3009 - 3013
  • [32] FINITE DIMENSIONAL PREDICTORS FOR HIDDEN MARKOV-CHAINS
    AGGOUN, L
    ELLIOTT, RJ
    SYSTEMS & CONTROL LETTERS, 1992, 19 (05) : 335 - 340
  • [33] An Approach of Diagnosis Based On The Hidden Markov Chains Model
    Bouamrane, Karim
    Djebbar, Amel
    Atmani, Baghdad
    COMPUTER SCIENCE JOURNAL OF MOLDOVA, 2008, 16 (02) : 256 - 268
  • [34] Exploring the state sequence space for hidden Markov and semi-Markov chains
    Guedon, Yann
    COMPUTATIONAL STATISTICS & DATA ANALYSIS, 2007, 51 (05) : 2379 - 2409
  • [35] Detection and identification of changes of hidden Markov chains: asymptotic theory
    Dayanik, Savas
    Yamazaki, Kazutoshi
    STATISTICAL INFERENCE FOR STOCHASTIC PROCESSES, 2022, 25 (02) : 261 - 301
  • [36] Analyticity of entropy rate in families, of hidden Markov chains (II)
    Han, Guangyue
    Marcus, Brian
    2006 IEEE INTERNATIONAL SYMPOSIUM ON INFORMATION THEORY, VOLS 1-6, PROCEEDINGS, 2006, : 103 - +
  • [37] Modeling and Unsupervised Classification of Multivariate Hidden Markov Chains With Copulas
    Brunel, N. J-B.
    Lapuyade-Lahorgue, Jerome
    Pieczynski, Wojciech
    IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 2010, 55 (02) : 338 - 349
  • [38] Contextual estimation of Hidden Markov Chains with application to image segmentation
    Derrode, S.
    Ben Youssef, L.
    Pieczynski, W.
    2006 IEEE International Conference on Acoustics, Speech and Signal Processing, Vols 1-13, 2006, : 1937 - 1940
  • [39] A COMPACT FRAMEWORK FOR HIDDEN MARKOV CHAINS WITH APPLICATIONS TO FRACTAL GEOMETRY
    Ruiz, Victor
    JOURNAL OF APPLIED PROBABILITY, 2008, 45 (03) : 630 - 639
  • [40] Copulas in vectorial hidden Markov chains for multicomponent image segmentation
    Brunel, N
    Pieczynski, W
    Derrode, S
    2005 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH, AND SIGNAL PROCESSING, VOLS 1-5: SPEECH PROCESSING, 2005, : 717 - 720