Reduction and Reconstruction of Control Knowledge in Rule Based Systems. Part 1: Reduction Theory

摘要

A model of rule based systems in which control knowledge can be represented explicitly, is introduced. The control structure of this Programmed Rule System (PRS) determines the rule to be chosen next, depending on the applicability of the current rule, and can be viewed as a straightforward generalization of the Markov control structure for production systems. The relation between control structure and the behavior of a PRS is analyzed. Properties of systems whose observable behaviors (traces) are identical, while their control structures might differ, are investigated. It is shown that, for every PRS M, a most efficient PRS M' can be obtained such that M and M' generate the same observable behavior. Such a PRS M' is called a minimal reduction of M.