中文摘要：

针对目前在中国象棋计算机博弈中广泛采用人工设置的评估函数,提出了基于激励学习与神经网络相结合的评估函数自学习方法,基于此模型实现了一个能自学习的中国象棋博弈程序.该方法避免了人工设置评估函数,解决了传统程序深层搜索博弈树消耗的时间和运行空间均很大的问题,也适用于其他的计算机博弈程序设计.实验结果表明,该方法是一种有效的自适应学习方法.

英文摘要：

The artificial evaluation function is widely used in Chinese chess game on computer currently,so a self-teaching method of evaluation function based on reinforced learning,which combined with neural network,is described.Based on this,a program of Chinese chess is implemented.This method avoids the shortcomings of imprecision by artificial factors,this method can consume less time and space than traditional methods,it can be applied to other computer games.The experimental results show that it is an effective self-teaching method.