Effectiveness and limitation of a method for control of chaos with reinforcement learning and its application to a simple chaotic neural network

摘要

In this article, the effectiveness of a method for control of chaos with reinforcement learning proposed by Gadaleta et el. is examined for several chaotic systems. We find that a simpler reinforcement learning than the Gadaleta's method can stabilize relatively simple chaotic maps. It is shown that the Gadaleta's method succeeds to stabilize a simple chaotic neural network. However, an example is given that it is hard to stabilize relatively complex map by the Gadaleta's method.