计算机工程与应用 ›› 2019, Vol. 55 ›› Issue (10): 36-44.DOI: 10.3778/j.issn.1002-8331.1812-0268

• 热点与综述 • 上一篇    下一篇

强化学习与生成式对抗网络结合方法研究进展

吴宏杰1,3,戴大东1,傅启明1,2,4,陈建平1,2,4,陆卫忠1   

  1. 1.苏州科技大学 电子与信息工程学院,江苏 苏州 215009
    2.苏州科技大学 江苏省建筑智慧节能重点实验室,江苏 苏州 215009
    3.苏州大学 江苏省计算机信息处理技术重点实验室,江苏 苏州 215006
    4.苏州市移动网络技术与应用重点实验室,江苏 苏州 215009
  • 出版日期:2019-05-15 发布日期:2019-05-13

Research on Combination of Reinforcement Learning and Generative Adversarial Networks

WU Hongjie1,3, DAI Dadong1, FU Qiming1,2,4, CHEN Jianping1,2,4, LU Weizhong1   

  1. 1.College of Electronics and Information Engineering, Suzhou University of Science and Technology, Suzhou, Jiangsu 215009, China
    2.Jiangsu Province Key Laboratory of Intelligent Building Energy Efficiency, Suzhou University of Science and Technology, Suzhou, Jiangsu 215009, China
    3.Jiangsu Provincial Key Lab for Information Processing Technologies, Soochow University, Suzhou, Jiangsu 215006, China
    4.Suzhou Key Laboratory of Mobile Network Technology and Application, Suzhou, Jiangsu 215009, China
  • Online:2019-05-15 Published:2019-05-13

摘要: 强化学习和生成式对抗网络是近年来人工智能领域的两个热门主题,在众多领域表现非常出色。近期出现较多关于两者结合的工作与报道,将强化学习交互式学习的优点与生成式对抗网络的启发自博弈思想相互融合。对两者结合的最新进展进行了梳理、比较与实验分析。对强化学习与生成式对抗网络的理论进行了概述;从强化学习改进生成式对抗网络、生成式对抗网络改进强化学习两个研究方向进行了阐述与比较,通过实验方式分析了这些方法在自然语言、机器控制领域的应用情况;展望了可能的发展趋势。

关键词: 强化学习, 生成式对抗网络, 深度学习, 人工智能

Abstract: Reinforcement learning and generative adversarial networks are hot research topics in the field of artificial intelligence in recent years, and have achieved excellent performance in many fields. The combination of the two fields has begun to appear in the recent future, which makes the interactive learning characteristics of reinforcement learning and the heuristics of generative adversarial networks improve each other. This paper first summarizes the theory of reinforcement learning and generative adversarial networks. Secondly, this paper elaborates on the latest research directions of using reinforcement learning to improve generative adversarial networks, and using generative adversarial networks to improve reinforcement learning. Finally, this paper summarizes the successful application in the field of natural language and machine control, and gives the future development trend.

Key words: reinforcement learning, generative adversarial networks, deep learning, artificial intelligence