Исследование и реализация алгоритмов обратного вывода в распределенных интеллектуальных системах

Abstract

The amount of resources necessary for functioning of intelligent systems is not always fully taken into account while creating, maintaining and operating large knowledge bases of these systems. As a result, the processes of working with knowledge bases, in particular finding out the truth of hypotheses, turn out to be very resource-intensive. The improvement of logical inference algorithms allows to increase the efficiency of intellectual systems of production type. An example is the approach of the relevant backward inference (LP-inference), which significantly reduces the number of calls to external sources of information. This reduces the resource consumption of inference. One of the stages of this process is to calculate special parameters called relevance scores. In case of distributed intelligent systems these strategies should take into consideration the distributed nature of knowledge bases. This research presents the basic concepts of algorithms of distributed LP-inference and considers the peculiarities of strategies for calculating relevance scores for distributed LP-structures, the use of which significantly increases the efficiency of distributed intelligent systems.