arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2107.01348cs.LGcs.AIcs.ROcs.SYeess.SY
收藏

检验强化学习中的平均和折扣奖励最优性准则

Examining average and discounted reward optimality criteria in reinforcement learning

School of Information Technology and Electrical Engineering(信息技术与电气工程学院);University of Queensland(昆士兰大学)

Vektor Dewanto, Marcus Gallagher

摘要:

在强化学习(RL)中,目标是获得最优策略,其中最优性准则至关重要。两个主要的最优准则是平均和折扣奖励。虽然后者更受欢迎,但在没有内在折扣概念的环境中应用是有问题的。这促使我们重新审视a)动态规划中最优性准则的进展,b)人为折扣因子的合理性和复杂性,以及c)直接最大化平均奖励准则的好处,这是无折扣的。我们的贡献包括对平均奖励和折扣奖励之间关系的彻底研究,以及对RL中它们的利弊的讨论。我们强调,平均报酬强化学习方法具有将一系列无折扣最优性准则(Veinott,1969)应用于强化学习的成分和机制。

英文摘要:

In reinforcement learning (RL), the goal is to obtain an optimal policy, for which the optimality criterion is fundamentally important. Two major optimality criteria are average and discounted rewards. While the latter is more popular, it is problematic to apply in environments without an inherent notion of discounting. This motivates us to revisit a) the progression of optimality criteria in dynamic programming, b) justification for and complication of an artificial discount factor, and c) benefits of directly maximizing the average reward criterion, which is discounting-free. Our contributions include a thorough examination of the relationship between average and discounted rewards, as well as a discussion of their pros and cons in RL. We emphasize that average-reward RL methods possess the ingredient and mechanism for applying a family of discounting-free optimality criteria (Veinott, 1969) to RL.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑