A Probabilistic-Statistical Model of Optimization of the Search for Solutions in Productional Expert Systems

摘要

An examination is made of a model of a productional expert system that is based on descending search for true hypotheses. The hypothesis with the greatest degree of certainty is considered the true one. An approximation of the density of probability of the degrees of certainty of the hypotheses is constructed on the basis of the statistics of the functioning of the system. Formulas for the conditional and unconditional probabilities of truth of the hypotheses are obtained. With the aid of these formulas, an ordering is made of the hypotheses. This ordering enables us to find, with given probability, the true hypothesis and to accelerate the process of finding solutions in the expert system.