Reinforcement learning by Improved Kohonen Feature Map Probabilistic Associative Memory based on Weights Distribution using extended eligibility

In this paper, we propose a reinforcement learning method by Improved Kohonen Feature Map Probabilistic Associative Memory based on Weights Distribution (IKFMPAM-WD) using extended eligibility. The proposed method is based on the actor-critic method, and the actor is realized by the IKFMPAM-WD. In the proposed method, the extended eligibility for the pair of the state and the action is defined. The extend eligibility is used for the selection of the action decision method and the reduction of unnecessary area. We carried out a series of computer experiments, and confirmed the effectiveness of the proposed method in the path-finding problem and the pursuit problem.