Representation Shifts Using Knowledge Compilation

Abstract

Tractability versus expressiveness is a problem underlying the current generation of problem solvers. It is desirable to specify the problem, along with knowledge required for its solution, in the most expressive means possible. However, from the perspective of the solver, the most efficient representation is the least expressive one. One solution to this problem is to devise a means of transforming an expressive problem description into an efficient problem solver. Knowledge compilation is the process of transforming a domain theory into a form specialized for the solution of a given problem set. This paper will present OPCOMP, a system devised to compile a first order quantifier free predicate calculus domain theory into a problem reduction solver.