arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2306.03362cs.LGcs.AI
收藏

利用动作偏好查询提升离线强化学习

Boosting Offline Reinforcement Learning with Action Preference Query

Tsinghua University(清华大学)

Qisen Yang, Shenzhi Wang, Matthieu Gaetan Lin, Shiji Song, Gao Huang

AI总结:

本文提出无交互的 OAP 训练方案,通过查询预收集动作与学习动作间的偏好来自适应调节策略约束,纠正离线 RL 的错误估计,并在 D4RL 上显著优于现有算法。

摘要:

训练实用智能体通常涉及离线和在线强化学习(RL),以平衡策略性能与交互成本。特别是,在线微调已成为一种常用方法,用于纠正在离线训练阶段学到的分布外数据的错误估计。然而,在医疗保健和自动驾驶等高风险场景中,即使是有限的在线交互也可能无法实现或造成灾难性后果。在这项工作中,我们提出了一种无交互训练方案,称为 Offline-with-Action-Preferences(OAP)。其核心洞见是,与在线微调相比,查询预收集动作与学习动作之间的偏好,对于错误估计问题可能同样有帮助,甚至更有帮助。通过根据动作偏好自适应地鼓励或抑制策略约束,OAP 能够区分高估与有益的策略改进,从而获得对未见数据更准确的评估。在理论上,我们证明了 OAP 为行为策略带来的性能提升下界。此外,在 D4RL 基准和最先进算法上的综合实验表明,OAP 取得了更高的分数(平均提高 29%),尤其是在具有挑战性的 AntMaze 任务上(提高 98%)。

英文摘要:

Training practical agents usually involve offline and online reinforcement learning (RL) to balance the policy's performance and interaction costs. In particular, online fine-tuning has become a commonly used method to correct the erroneous estimates of out-of-distribution data learned in the offline training phase. However, even limited online interactions can be inaccessible or catastrophic for high-stake scenarios like healthcare and autonomous driving. In this work, we introduce an interaction-free training scheme dubbed Offline-with-Action-Preferences (OAP). The main insight is that, compared to online fine-tuning, querying the preferences between pre-collected and learned actions can be equally or even more helpful to the erroneous estimate problem. By adaptively encouraging or suppressing policy constraint according to action preferences, OAP could distinguish overestimation from beneficial policy improvement and thus attains a more accurate evaluation of unseen data. Theoretically, we prove a lower bound of the behavior policy's performance improvement brought by OAP. Moreover, comprehensive experiments on the D4RL benchmark and state-of-the-art algorithms demonstrate that OAP yields higher (29% on average) scores, especially on challenging AntMaze tasks (98% higher).

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。

↑