Performance Testing of Propositional Logic Inference Engines

Abstract

Speed is a significant factor in the implementations of rule-based systems, and many inference engines slow dramatically as the size of the problem increases. Test sets such as Waltz and Manners measure the speed of first order logic inference engines. However, to our knowledge, no test sets for propositional logic inference engines have heretofore been identified.This paper proposes and tests two test sets that measure the performance of propositional logic inference engines. The first, Chess, measures the speed at which individual rules are tested in a large test set. The second, the Christmas Tree, tests the speed of the chaining process using a binary tree of configurable depths.