視覚情報を用いた状態・行動空間の自律的生成

抄録

To apply reinforcement learnig in the real world, we need to process sensor data adequately for action learning. Since it is difficult to construct state space and to learn the appropreate action simultaneously, we assume that an evaluation is given to each step of action. Evaluations are binary signals that mean actions are good or bad. Under this condition, we propose a method of dividing and clustering the state space. The TRN (Topology Representing Networks) algorithm is a vector quantization algorithm, and it can preserve topology in the input space. We apply the TRN algorithm to our problem with dynamically increasing nodes and the radial basis function.