深度强化学习综述:兼论计算机围棋的发展  

Review of deep reinforcement learning and discussions on the development of computer Go

作　　者：赵冬斌 邵坤[1] 朱圆恒[1] 李栋[1] 陈亚冉[1] 王海涛[1] 刘德荣[2] 周彤[3] 王成红 

机构地区：[1]中国科学院自动化研究所复杂系统管理与控制国家重点实验室,北京100190 [2]北京科技大学自动化学院,北京100083 [3]清华大学自动化系,北京100084 [4]国家自然科学基金委信息科学部,北京100085

出　　处：《控制理论与应用》2016年第6期701-717,共17页Control Theory & Applications

基　　金：国家自然科学基金项目(61273136;61573353;61533017)~~

摘　　要：深度强化学习将深度学习的感知能力和强化学习的决策能力相结合,可以直接根据输入的图像进行控制,是一种更接近人类思维方式的人工智能方法.自提出以来,深度强化学习在理论和应用方面均取得了显著的成果.尤其是谷歌深智(Deep Mind)团队基于深度强化学习方法研发的计算机围棋"初弈号–Alpha Go",在2016年3月以4:1的大比分战胜了世界围棋顶级选手李世石(Lee Sedol),成为人工智能历史上一个新里程碑.为此,本文综述深度强化学习的发展历程,兼论计算机围棋的历史,分析算法特性,探讨未来的发展趋势和应用前景,期望能为控制理论与应用新方向的发展提供有价值的参考.Deep reinforcement learning which incorporates both the advantages of the perception of deep learning and the decision making of reinforcement learning is able to output control signal directly based on input images. This mechanism makes the artificial intelligence much close to human thinking modes. Deep reinforcement learning has achieved remarkable success in terms of theory and application since it is proposed. ‘Chuyihao–Alpha Go＇, a computer Go developed by Google Deep Mind, based on deep reinforcement learning, beat the world＇s top Go player Lee Sedol 4：1 in March2016. This becomes a new milestone in artificial intelligence history. This paper surveys the development course of deep reinforcement learning, reviews the history of computer Go concurrently, analyzes the algorithms features, and discusses the research directions and application areas, in order to provide a valuable reference to the development of control theory and applications in a new direction.

关 键 词：深度强化学习 初弈号 深度学习 强化学习 人工智能 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]