优化博弈问题评估函数参数的自适应遗传算法  

Adaptive genetic algorithm for evaluation function parameters optimization in game problem

作　　者：王秀坤[1] 刘健男[1] 

机构地区：[1]大连理工大学计算机系,辽宁大连116023

出　　处：《计算机工程与应用》2009年第20期42-44,51,共4页Computer Engineering and Applications

摘　　要：使用自适应遗传算法对五子棋博弈问题中的评估函数参数进行训练优化。引入陪练算法对训练进行指导。给出了一种新的适应度函数计算方法,该方法避免了在训练过程中种群个体之间的大量竞赛,从而节省了训练时间。实验结果表明训练得到的参数组成的评估函数优于陪练算法中的评估函数。Adaptive Genetic Algorithm（AGA） is used to optimize the parameters of evaluation function in the gobang game problem.The tutorial algorithm is introduced to guide the training.A new fitness function is given which can avoid large amount of competitions among individuals during the training process so as to save training time.The experiment results show that the evaluation functions composed of parameters obtained from training are better than those in tutorial algorithms.

关 键 词：自适应遗传算法 博弈 评估函数 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]