Effective search methods for pattern matching inferencing using specific similarity measures

Abstract

Pattern matching inferencing (PMI) is one of the ways of approximating the compositional rule of inference (CRI) as proposed by L. A. Zadeh (1973). PMI is a generic algorithm to create different approximate inferencing algorithms. In particular, approximate analogical reasoning, approximate deductive reasoning and approximate analogical and deductive reasoning are under the class of PMI. PMI as extended by C. Lucas and I. G. Turksen (1990) and the search methods currently used in PMI are considered. Several similarity measures are shown to have some desired properties to make the search process to fire rules in PMI more effective. Using these properties, two new search strategies are proposed instead of the commonly used exhaustive search.<>