Extending the CLP Engine for Reasoning under Uncertainty

Abstract

We show how the amalgamation of Logic Programming with probabilistic reasoning enhances its capabilities for intelligent reasoning. Unlike current approaches we use concepts from Constraint Logic Programming in order to achieve this. In particular, we use the constraint store for storing probabilistic information and inference, and finite domains as sets of basic elements over which distributions can be defined. We describe a new language, Probabilistic finite domains and show how it can be used to code code two examples. First the Monty Hall problem is coded and the extensional means of simulating intelligence within our system are described. Second, we illustrate the benefits of the probabilistic information over the crisp finite domains in solving a simple encoding scheme. Aspects of a prototype implementation, a Prolog meta-interpreter, are discussed.