arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1906.04556cs.LGstat.ML
收藏

利用优势函数的符号学习连续域中的确定性策略

Exploiting the Sign of the Advantage Function to Learn Deterministic Policies in Continuous Domains

UM-SJTU Joint Institute(上海交通大学密西根学院)

Matthieu Zimmer, Paul Weng

摘要:

在连续域中学习确定性策略的背景下,我们重新审视了一种方法,该方法首先在连续演员批评学习自动机(CACLA)中提出,后来在神经拟合演员批评(NFAC)中扩展。该方法基于与确定性策略梯度(DPG)不同的策略更新。以前的工作已经观察到其出色的性能经验,但缺乏理论依据。为了填补这一空白,我们提供了一个理论解释,以激励这一非正统的政策更新,将其与另一个更新,并明确后者的目标函数。我们还深入讨论了这些更新的属性,以更深入地了解整体方法。此外,我们扩展它,并提出了一个新的信任域算法,惩罚NFAC(PeNFAC)。最后,我们在几个经典的控制问题的实验证明,它超越了国家的最先进的算法来学习确定性的政策。

英文摘要:

In the context of learning deterministic policies in continuous domains, we revisit an approach, which was first proposed in Continuous Actor Critic Learning Automaton (CACLA) and later extended in Neural Fitted Actor Critic (NFAC). This approach is based on a policy update different from that of deterministic policy gradient (DPG). Previous work has observed its excellent performance empirically, but a theoretical justification is lacking. To fill this gap, we provide a theoretical explanation to motivate this unorthodox policy update by relating it to another update and making explicit the objective function of the latter. We furthermore discuss in depth the properties of these updates to get a deeper understanding of the overall approach. In addition, we extend it and propose a new trust region algorithm, Penalized NFAC (PeNFAC). Finally, we experimentally demonstrate in several classic control problems that it surpasses the state-of-the-art algorithms to learn deterministic policies.

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。

↑