Temporal Parallelization of Inference in Hidden Markov Models

Hassan, Sakira, Särkkä, Simo and Garcia-Fernandez, angel ORCID: 0000-0002-6471-8455 (2021) Temporal Parallelization of Inference in Hidden Markov Models IEEE Transactions on Signal Processing, 69 (48). pp. 4875-4887. ISSN 1053-587X, 1941-0476

Text Hassan21_accepted.pdf - Author Accepted Manuscript Download (3MB) | Preview

Abstract

This paper presents algorithms for parallelization of inference in hidden Markov models (HMMs). In particular, we propose parallel backward-forward type of filtering and smoothing algorithm as well as parallel Viterbi-type maximum-a-posteriori (MAP) algorithm. We define associative elements and operators to pose these inference problems as parallel-prefix-sum computations in sum-product and max-product algorithms and parallelize them using parallel-scan algorithms. The advantage of the proposed algorithms is that they are computationally efficient in HMM inference problems with long time horizons. We empirically compare the performance of the proposed methods to classical methods on a highly parallel graphical processing unit (GPU).

Item Type:	Article
Additional Information:	submitted to the IEEE transactions on Signal Processing
Uncontrolled Keywords:	cs.DC, cs.DC, stat.ML
Divisions:	Faculty of Science & Engineering > School of Electrical Engineering, Electronics and Computer Science
Depositing User:	Symplectic Admin
Date Deposited:	23 Aug 2021 08:04
Last Modified:	22 May 2026 16:40
DOI:	10.1109/TSP.2021.3103338
Related Websites:	Author Publisher
URI:	https://livrepository.liverpool.ac.uk/id/eprint/3134375
Disclaimer:	The University of Liverpool is not responsible for content contained on other websites from links within repository metadata. Please contact us if you notice anything that appears incorrect or inappropriate.