強化学習の連続値への適用

概要

論文の詳細を見る
The reinforcement learning is drawing attention as a technique that can solve problems automatically and adaptably. With thereinforcement learning, a series of actions is learned by maximizing a rewardsignal through trial and error. In general, the reinforcement learning isused for discrete problems. In order to solve real-world problems thattend to include continuous numbers, a couple of methods which is capableto deal with continuous problems has been proposed so far. In this paper,we discuss how the learning performance will be a?ected by changing thenumber of features within a linear architecture method, a type ofmethod applied for solving continuous problems.
2007-12-21