arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2105.06136cs.AIcs.LG
收藏

AlphaZero-like深度强化学习中的自适应热启动MCTS

Adaptive Warm-Start MCTS in AlphaZero-like Deep Reinforcement Learning

Leiden Institute of Advanced Computer Science, Leiden University(莱顿大学高级计算机科学研究所)

Hui Wang, Mike Preuss, Aske Plaat

摘要:

AlphaZero在深度强化学习方面取得了令人印象深刻的表现,它利用了一种架构,将搜索和神经网络的训练结合在一起。许多研究人员正在寻找方法来复制和改善其他游戏/任务的结果。然而,该架构旨在从头开始学习,白板,接受自我游戏中的冷启动问题。最近,提出了一种用于蒙特卡罗树搜索的热启动增强方法,以改善自启动阶段。它采用一个固定的参数$I^\prime$来控制热启动长度。据报道,在小型棋盘游戏中的表现有所改善。在本文中,我们提出的结果与自适应开关方法。实验表明,我们的方法比固定的$I^\prime$,特别是“深”,战术,游戏(奥赛罗和连接四)。我们推测$I^\prime$的自适应值也受游戏大小的影响,并且平均而言,$I^\prime$将随着游戏大小而增加。我们得出结论,AlphaZero式深度强化学习受益于基于自适应推出的热启动,就像15年前快速行动价值估计对基于推出的强化学习所做的那样。

英文摘要:

AlphaZero has achieved impressive performance in deep reinforcement learning by utilizing an architecture that combines search and training of a neural network in self-play. Many researchers are looking for ways to reproduce and improve results for other games/tasks. However, the architecture is designed to learn from scratch, tabula rasa, accepting a cold-start problem in self-play. Recently, a warm-start enhancement method for Monte Carlo Tree Search was proposed to improve the self-play starting phase. It employs a fixed parameter $I^\prime$ to control the warm-start length. Improved performance was reported in small board games. In this paper we present results with an adaptive switch method. Experiments show that our approach works better than the fixed $I^\prime$, especially for "deep," tactical, games (Othello and Connect Four). We conjecture that the adaptive value for $I^\prime$ is also influenced by the size of the game, and that on average $I^\prime$ will increase with game size. We conclude that AlphaZero-like deep reinforcement learning benefits from adaptive rollout based warm-start, as Rapid Action Value Estimate did for rollout-based reinforcement learning 15 years ago.

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。

↑