增强学习与人工神经网络结合在对弈中的应用  

Application of Combining Reinforcement Learning With Artificial Neural Network to Playing Chess

作　　者：谢磊[1] 鲁汉榕[2] 徐新[1] 

机构地区：[1]空军雷达学院研究生队,湖北武汉4300199 [2]空军雷达学院指挥自动化工程系,湖北武汉430019

出　　处：《空军雷达学院学报》2004年第4期37-39,共3页Journal of Air Force Radar Academy

摘　　要：为了得到高智能的对弈程序,鉴于增强学习应用于对弈程序时状态数太多而难以建表,提出将增强学习与人工神经网络结合起来的方法,通过应用该方法的五子棋程序与一般的五子棋程序的比较,表明该方法更优越.该对弈程序可应用于模拟军事对战,为实际作战提供参考.The purpose of this article is to research the highly intelligent playing chess program. When applying reinforcement learning to play chess, there is too much states to set up tables. This paper presents a method of com- bining reinforcement learning with artificial neural network to solve the problem. By comparing this method with the general one, it proves that this method is much better. The program also can be applied to simulating the military battle, and help to make decision.

关 键 词：增强学习 Q学习 人工神经网络 反向传播算法 目标函数 

分 类 号：O234[理学—运筹学与控制论]