Premium
General framework for multidimensional models
Author(s) -
Jiroušek Radim,
Vejnarová Jiřina
Publication year - 2003
Publication title -
international journal of intelligent systems
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.291
H-Index - 87
eISSN - 1098-111X
pISSN - 0884-8173
DOI - 10.1002/int.10077
Subject(s) - probabilistic logic , computer science , mathematical proof , bayesian network , bayesian probability , theoretical computer science , statistical model , artificial intelligence , mathematics , geometry
This article is an attempt to build a unifying theoretical framework in which both probabilistic and possibilistic multidimensional models can be described. This novel approach is applied to presenting models based on iterative application of operators of composition, which were introduced in previous articles separately for probabilistic and possibilistic distributions. It appears that, although not all of the proofs are elegant, the apparatus is quite efficient, and enables the authors to deduce all the necessary properties in a uniform way. Because the described models are, in the probabilistic setting, fully equivalent to Bayesian networks, our main result describes how to compute marginal distributions of both Bayesian networks and possibilistic belief networks (when represented in the form of generating sequences). © 2003 Wiley Periodicals, Inc.