Host: Japan Society for Fuzzy Theory and Intelligent Informatics (SOFT)
Chain Form Reinforcement Learning (CFRL) was proposed for a reinforcement learning agent using low memory. However, we hold unused information in the memory. In this paper, in order to make memory small dramatically, we introduce Turning Spot Learning (TSL). TSL is a method which imitates human perceptions. If we are asked direction, we often tell a suitable turning spot. Based on this, a TSL agent learns only suitable turning spots. In order to select an action, a TSL agent uses distance (the number of actions). Our method was made a comparison to Q-Learning and CFRL in two kinds of goal search problem. We examined performance and discussed the best usage environment.
