Klaszterezési módszeren alapuló fuzzy szabálybázis redukálás a FRIQ-learning rendszerben

Abstract

Fuzzy logic-based reinforcement learning methods represent system knowledge in the form of a fuzzy rule-base. The number of rules in the rule-base, in other words, the size of the knowledge base, determines the complexity and computational demand of the RL system. Consequently, optimizing the size of the rule-base and identifying potentially redundant rules is crucial for enhancing the efficiency of these fuzzy logic-based learning methods. A large sized rule-base can increase computational demand and reduce the transparency of the decision-making process. The aim of this paper is to introduce a fuzzy rule-base reduction method based on a clustering technique in the FRIQ-learning reinforcement learning system. This method is suitable for reducing the size of the fuzzy rule-base by identifying relevant rules while ensuring that the decision-making capability of the system remains unchanged.