Number of the records: 1
Rank of tensors of l-out-of-k functions: an application in probabilistic inference
- 1.
SYSNO ASEP 0361630 Document Type J - Journal Article R&D Document Type Journal Article Subsidiary J Článek ve WOS Title Rank of tensors of l-out-of-k functions: an application in probabilistic inference Author(s) Vomlel, Jiří (UTIA-B) RID, ORCID Number of authors 1 Source Title Kybernetika. - : Ústav teorie informace a automatizace AV ČR, v. v. i. - ISSN 0023-5954
Roč. 47, č. 3 (2011), s. 317-336Number of pages 20 s. Language eng - English Country CZ - Czech Republic Keywords Bayesian network ; probabilistic inference ; tensor rank Subject RIV BB - Applied Statistics, Operational Research R&D Projects 1M0572 GA MŠMT - Ministry of Education, Youth and Sports (MEYS) GA201/09/1891 GA ČR - Czech Science Foundation (CSF) GEICC/08/E010 GA ČR - Czech Science Foundation (CSF) CEZ AV0Z10750506 - UTIA-B (2005-2011) UT WOS 000293207900002 EID SCOPUS 83455262599 Annotation We study the problem of efficient probabilistic inference with Bayesian networks when some of the conditional probability tables represent deterministic or noisy l-out-of-k functions. These tables appear naturally in real-world applications when we observe a state of a variable that depends on its parents via an addition or noisy addition relation. We provide a lower bound of the rank and an upper bound for the symmetric border rank of tensors representing l-out-of-k functions. We propose an approximation of tensors representing noisy l-out-of-k functions by a sum of r tensors of rank one, where r is an upper bound of the symmetric border rank of the approximated tensor. We applied the suggested approximation to probabilistic inference in probabilistic graphical models. Numerical experiments reveal that we can get a gain in the order of two magnitudes but at the expense of a certain loss of precision. Workplace Institute of Information Theory and Automation Contact Markéta Votavová, votavova@utia.cas.cz, Tel.: 266 052 201. Year of Publishing 2012
Number of the records: 1