Abstract: An evaluation function with parameters is designed, and the parameters are optimized by using genetic algorithm. The heuristic information is added to guide the searching and improves the convergence rate of the algorithm. Through introducing the fitness matrix, the crossover and mutation rate matrix, each parameter of the chromosome is considered, the convergence rate is further improved. Sparring algorithm is introduced to guide the training, using gradient training programs to save training time. Experimental result shows the skills in playing Dots-and-Boxes are improved after its evaluation function parameters are optimized.

Key words: genetic algorithm, evaluation function, game

摘要： 设计了一种带参数的评估函数，采用遗传算法对参数进行优化。加入启发式信息指导搜索的进行，使算法的收敛速度得到了提高。引入适应度矩阵，交叉变异率矩阵，对染色体中的每个参数进行分别考虑，进一步提高了收敛速度。引入陪练算法进行训练指导，提出一种梯度训练方案，节省了训练时间。实验结果表明评估函数参数优化后的点点连格棋的棋力得到了提高。

关键词: 遗传算法, 评估函数, 博弈