Application of FCM-type Co-clustering with Exclusive Partition to Reinforcement Learning

Abstract

Recently, search space explosion and increasing learning time have been a severe problem in reinforcement learning. Several methods and algorithms are introduced to learning agents in order to respond to the problem. In this paper, we introduce FCCM clustering technique into the Q-learning. Agent is also able to learn from compressed information given by another FCCM clustering agent. We demonstrate the effect of the technique with a simple example.