arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 2406.04081cs.LGcs.AI
收藏

强化学习中的期望分位数自举

Bootstrapping Expectiles in Reinforcement Learning

CMAP, Ecole Polytechnique(巴黎综合理工学院CMAP);INRIA Paris, HeKA(法国国家信息与自动化研究所巴黎分所HeKA团队);SUPAERO(法国航空航天高等学院);Cohere(Cohere公司)

Pierre Clavier, Emmanuel Rachelson, Erwan Le Pennec, Matthieu Geist

AI总结:

本文提出 ExpectRL,用期望分位数损失替代评论家的 $L_2$ 损失以在强化学习中引入悲观主义,在过估计问题和鲁棒强化学习基准上均优于经典方法,并扩展了自动选择悲观程度的机制。

摘要:

许多经典的强化学习(RL)算法依赖于贝尔曼算子,该算子涉及对下一状态的期望,从而产生了自举的概念。为了引入一种悲观主义形式,我们提出用期望分位数(expectile)来替代这一期望。在实践中,这可以通过将评论家(critic)的 $L_2$ 损失替换为更一般的期望分位数损失来非常简单地实现。在强化学习中引入悲观主义出于多种原因是可取的,例如解决过估计问题(其经典解决方案是双重 Q 学习或 TD3 的双评论家方法)或鲁棒强化学习(其中状态转移是对抗性的)。我们对这两种情况进行了实证研究。对于过估计问题,我们表明所提出的方法 ExpectRL 比经典的双评论家方法提供了更好的结果。在涉及环境变化的鲁棒强化学习基准上,我们表明我们的方法比经典强化学习算法更具鲁棒性。我们还引入了 ExpectRL 的一个变体,将其与域随机化相结合,该变体与最先进的鲁棒强化学习智能体具有竞争力。最后,我们还将 ExpectRL 扩展为具有自动选择期望分位数值(即悲观程度)的机制。

英文摘要:

Many classic Reinforcement Learning (RL) algorithms rely on a Bellman operator, which involves an expectation over the next states, leading to the concept of bootstrapping. To introduce a form of pessimism, we propose to replace this expectation with an expectile. In practice, this can be very simply done by replacing the $L_2$ loss with a more general expectile loss for the critic. Introducing pessimism in RL is desirable for various reasons, such as tackling the overestimation problem (for which classic solutions are double Q-learning or the twin-critic approach of TD3) or robust RL (where transitions are adversarial). We study empirically these two cases. For the overestimation problem, we show that the proposed approach, ExpectRL, provides better results than a classic twin-critic. On robust RL benchmarks, involving changes of the environment, we show that our approach is more robust than classic RL algorithms. We also introduce a variation of ExpectRL combined with domain randomization which is competitive with state-of-the-art robust RL agents. Eventually, we also extend \ExpectRL with a mechanism for choosing automatically the expectile value, that is the degree of pessimism

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。

↑