自己組織化マップを用いた教示による強化学習の高速化手法の提案(機械力学,計測,自動制御)

概要

論文の詳細を見る
A new pre-teaching method for reinforcement learning using Self-Organizing Map (SOM) is described. The purpose of our study is to increase the learning rate using small number of teaching data generated by a human expert. In our method, the SOM is used to generate initial teaching data for the reinforcement learning agent from a few teaching data. The reinforcement learning function of the agent is initialized by using the teaching data generated by the SOM so as to increase the probability of selecting the optimal actions estimated by the SOM. Because the agent can get high rewards from the start of reinforcement learning, it is expected to increase the learning rate. The results of two computer simulations, mobile robot navigation and pursuit game, showed that the learning rate increased although the human expert had showed only a few teaching data.
社団法人日本機械学会の論文
2004-06-25