改进的协同进化遗传算法在机器博弈中的应用

阐述了一种典型的协同进化遗传算法（SANE），在机器博弈中，用前馈神经网络（FNN）表示局面估值函数，该算法采用两个种群合作协同的方式进化该FNN。对上述算法在种群的初始化方面进行了合理改进：用粒子群算法（PSO）先对种群进行预处理。实验表明，在协同进化的过程中，经过预处理的种群会比随机生成的种群效率更高。

Application of Co-evolution and PSO Algorithm in Machine Game

Presents a novel co-evolutionary genetic system(SANE),the algorithm use two populations to co-evolve feedback neural networks(FNN) which evaluate board positions in machine games.Some reasonable improvements which pre-train the population by PSO algorithm are used in the population initialization.In our case,the results of experiments show that pre-training of the population in co-evolution is highly effective in creating stronger game playing strategies than co-evolution with random population.