Knowledge Representation, Belief Revision, and the Challenge of Optimality

Abstract

The fields of KR and BR are closely related, and abstractly circumscribe the two requirements of articulating and consistently accumulating knowledge. The application to particular problem solving tasks provides further constraints on the articulation and accumulation of knowledge, many of which are complex, conflicting, and difficult to formalize.

Beginning with the idea that belief revision provides the most general framework for accumulating knowledge, we review recent experience in optimization problem solving, where the difficulties include specifying the problem, the objective function for solution, and the knowledge of how to search a large search space.

The experience reveals ideas for a general optizmation problem solving framework, in which belief revision, constraint programming, and heuristic optimzation all work together.