State transition rate based reinforcement learning

Reinforcement learning is a kind of machine learning that adapts to an environment with special input called a reinforcement signal. An agent using reinforcement learning can obtain purposeful behavior autonomically. However, there are problems in that reinforcement learning takes a long time because it advances while repeating trial-and-error, and an acquired action is not necessarily optimal. We propose reinforcement learning using state transition rates, and compare it with another method. As a result, our method shows the capability of learning purposeful behavior efficiently.