bibtype	C - Conference Paper (international conference)
ARLID	0327312
utime	20240103191906.9
mtime	20090724235959.9
WOS	000268585700049
SCOPUS	69049089935
DOI	10.1007/978-3-642-02906-6_49
title (primary) (eng)	Triangulation Heuristics for BN2O Networks
specification	page_count 12 s.
serial	ARLID cav_un_epca*0327310 ISBN 978-3-642-02905-9 ISSN 0302-9743 title Symbolic and Quantitative Approaches to Reasoning with Uncertainty page_num 566-577 publisher place Berlin name Springer year 2009 editor name1 Sossai name2 C. editor name1 Chemello name2 G.
title (cze)	Heuristiky pro triangulaci sítí typu BN2O
keyword	Bayesian network
keyword	BN2O
keyword	noisy-or
keyword	graphical transformation
keyword	parent divorcing
keyword	tensor rank-one decomposition
author (primary)	ARLID cav_un_auth*0100825 name1 Savický name2 Petr institution UIVT-O full_dept Department of Theoretical Computer Science fullinstit Ústav informatiky AV ČR, v. v. i.
author	ARLID cav_un_auth*0101228 name1 Vomlel name2 Jiří institution UTIA-B full_dept Department of Decision Making Theory fullinstit Ústav teorie informace a automatizace AV ČR, v. v. i.
cas_special	project project_id 1M0545 agency GA MŠk country CZ ARLID cav_un_auth*0203502 project project_id 1ET100300517 agency GA AV ČR ARLID cav_un_auth*0001446 project project_id 1M0572 agency GA MŠk ARLID cav_un_auth*0001814 project project_id 2C06019 agency GA MŠk country CZ ARLID cav_un_auth*0216518 project project_id GEICC/08/E010 agency GA ČR ARLID cav_un_auth*0241637 project project_id GA201/09/1891 agency GA ČR ARLID cav_un_auth*0253175 research CEZ:AV0Z10300504 research CEZ:AV0Z10750506 abstract (eng) A BN2O network is a Bayesian network having the structure of a bipartite graph with all edges directed from one part (the top level) toward the other (the bottom level) and where all conditional probability tables are noisy-or gates. In order to perform efficient inference, graphical transformations of these networks are performed. The complexity of inference is proportional to the total table size of tables corresponding to the cliques of the triangulated graph. Therefore in order to get efficient inference it is desirable to have small cliques in the triangulated graph. We analyze existing heuristic triangulation methods applicable to BN2O networks after transformations using parent divorcing and tensor rank-one decomposition and suggest several modifications. Both theoretical and experimental results confirm that tensor rank-one decomposition yields better results than parent divorcing in randomly generated BN2O networks that we tested. abstract (cze) Síť typu BN2O je Bayesovská síť se strukturou bipartitního grafu, ve kterém jsou všechny hrany orientované z jedné části (horní vrstva) do druhé části (dolní vrstva) a ve kterém jsou všechny pravděpodobnostní tabulky noisy-or. Pro zvýšení efektivnosti inference jsou tyto sítě transformovány z hlediska jejich grafové struktury. Složitost inference je úměrná celkové velikosti tabulek v síti. Proto je cílem transformace dosáhnout co nejmenší součet velikosti tabulek triangulované sítě. Článek testuje existující metody triangulace aplikovatelné na sítě BN2O po transformacích parent divorcing a tensor rank-one decomposition a navrhuje některé modifikace. Jak teoretické, tak experimentální výsledky potvrzují, že tensor rank-one decomposition poskytuje lepší výsledky než parent divorcing na náhodně generovaných sítích BN2O, které byly v experimentu použity. action ARLID cav_un_auth*0252448 name ECSQARU 2009. European Conference /10./ place Verona dates 01.07.2009-03.07. 2009 country IT reportyear 2010 RIV BA permalink http://hdl.handle.net/11104/0174155 arlyear 2009 mrcbU14 69049089935 SCOPUS mrcbU34 000268585700049 WOS mrcbU63 cav_un_epca*0327310 Symbolic and Quantitative Approaches to Reasoning with Uncertainty 978-3-642-02905-9 0302-9743 566 577 Berlin Springer 2009 Lecture Notes in Artificial Intelligence 5590 mrcbU67 Sossai C. 340 mrcbU67 Chemello G. 340