Decision Network Semantics of Branching Constraint Satisfaction Problems
Abstract: Branching Constraint Satisfaction Problems (BCSPs) have been introduced to model dynamic resource allocation subject to constraints and uncertainty. We give BCSPs a formal probability semantics by showing how they can be mapped to a certain class of Bayesian decision networks. This allows us to describe logical and probabilistic constraints in a uniform fashion. We also discuss extensions to BCSPs and decision networks suggested by the relationship between the two formalisms.
Loading