arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2005.09814cs.LGcs.AIstat.ML
收藏

镜像下降策略优化

Mirror Descent Policy Optimization

University of Alberta(阿尔伯塔大学);Technion(以色列理工学院);Microsoft Research(微软研究院);Google Research(谷歌研究院)

Manan Tomar, Lior Shani, Yonathan Efroni, Mohammad Ghavamzadeh

摘要:

镜像下降法(Mirror Descent,MD)是约束凸优化中的一阶算法,近年来被证明是分析强化学习(Reinforcement Learning,RL)中信赖域算法的重要工具。然而,这些理论分析的算法和实际使用的算法之间仍然存在相当大的差距。受此启发,我们提出了一种高效的RL算法,称为{ \em镜像下降策略优化}(MDPO)。MDPO通过{ \em approximately}求解信赖域问题来迭代更新策略,该信赖域问题的目标函数由两个项组成:标准RL目标的线性化和限制两个连续策略彼此接近的邻近项.每次更新通过在该目标函数上采取多个梯度步骤来执行这种近似。我们推导出{ \em的政策}和{ \em的政策} MDPO的变体,同时强调重要的设计选择,激励现有理论的MD在RL。我们强调的政策MDPO和两个流行的信任区域RL算法:TRPO和PPO之间的连接,并表明,明确执行的信任区域约束实际上是{ \em不} TRPO的高性能增益的必要性。然后,我们展示了如何流行的软演员-评论家(SAC)算法可以通过轻微的修改,关闭政策MDPO。总体而言,MDPO源自MD原则,提供了一种统一的方法来查看许多流行的RL算法,并且在许多连续控制任务中表现优于或等同于TRPO,PPO和SAC。代码可在\url{https://github.com/manantomar/Mirror-Descent-Policy-Optimization}上获得。

英文摘要:

Mirror descent (MD), a well-known first-order method in constrained convex optimization, has recently been shown as an important tool to analyze trust-region algorithms in reinforcement learning (RL). However, there remains a considerable gap between such theoretically analyzed algorithms and the ones used in practice. Inspired by this, we propose an efficient RL algorithm, called {\em mirror descent policy optimization} (MDPO). MDPO iteratively updates the policy by {\em approximately} solving a trust-region problem, whose objective function consists of two terms: a linearization of the standard RL objective and a proximity term that restricts two consecutive policies to be close to each other. Each update performs this approximation by taking multiple gradient steps on this objective function. We derive {\em on-policy} and {\em off-policy} variants of MDPO, while emphasizing important design choices motivated by the existing theory of MD in RL. We highlight the connections between on-policy MDPO and two popular trust-region RL algorithms: TRPO and PPO, and show that explicitly enforcing the trust-region constraint is in fact {\em not} a necessity for high performance gains in TRPO. We then show how the popular soft actor-critic (SAC) algorithm can be derived by slight modifications of off-policy MDPO. Overall, MDPO is derived from the MD principles, offers a unified approach to viewing a number of popular RL algorithms, and performs better than or on-par with TRPO, PPO, and SAC in a number of continuous control tasks. Code is available at \url{https://github.com/manantomar/Mirror-Descent-Policy-Optimization}.

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。

↑