Belief revision of product-based causal possibilistic networks

Abstract

Belief revision is an important task for designing intelligent systems. In the possibility theory framework, considerable work has addressed revising beliefs in a possibilistic logic framework while only few works have addressed a possibilistic revision process in graphical-based frameworks. In particular, belief revision of causal product-based possibilistic networks which are the possibilistic counterparts of probabilistic causal networks has not yet been addressed. This paper is concerned with revising causal possibilistic networks in presence of two kinds of information: observations and interventions (which are external actions forcing some variables to some specific values). It contains two contributions: we first propose an efficient method for integrating and accepting new observations by directly transforming the initial graph. Then we highlight important issues related to belief revision of causal networks with sets of observations and interventions. © 2010 Springer-Verlag Berlin Heidelberg.