arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2005.04269cs.LGcs.AIstat.ML
收藏

用截断混合连续分布分位数批评控制高估偏差

Controlling Overestimation Bias with Truncated Mixture of Continuous Distributional Quantile Critics

Samsung AI center(三星人工智能中心);National Research University Higher School of Economics(国立研究型大学高等经济学院)

Arsenii Kuznetsov, Pavel Shvechikov, Alexander Grishin, Dmitry Vetrov

摘要:

高估偏差是精确离策略学习的主要障碍之一。本文研究了一种新的方法,以减轻在连续控制设置的高估偏差。我们的方法-截断分位数批评,TQC,-融合了三个想法:分布式表示的批评,截断的批评预测,和合奏的多个批评。分布表示和截断允许任意粒度的高估控制,而集成提供了额外的分数改进。TQC在连续控制基准测试套件的所有环境中的表现都优于当前最先进的技术,在最具挑战性的Humanoid环境中表现出25%的改进。

英文摘要:

The overestimation bias is one of the major impediments to accurate off-policy learning. This paper investigates a novel way to alleviate the overestimation bias in a continuous control setting. Our method---Truncated Quantile Critics, TQC,---blends three ideas: distributional representation of a critic, truncation of critics prediction, and ensembling of multiple critics. Distributional representation and truncation allow for arbitrary granular overestimation control, while ensembling provides additional score improvements. TQC outperforms the current state of the art on all environments from the continuous control benchmark suite, demonstrating 25% improvement on the most challenging Humanoid environment.

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。

↑