摘要
针对进化策略算法收敛速度较慢、容易早熟的问题,提出一种新的基于双种群的改进进化策略算法。将种群划分为规模较小的精英子群和规模较大的普通子群。精英子群用于存放种群中最优秀的个体,普通子群用于存放种群中的普通个体。对不同的子群采用不同的变异策略,使种群在解空间具有尽可能分散的全局搜索能力的同时在局部具有尽可能精细的局部搜索能力。通过理论分析证明了算法的正确性,对几个典型的函数应用该算法进行模拟进化实验,也取得了良好的效果。
Aiming at the problems of premature convergence and slow convergence of traditional evolution strategy, a Modified Evolutionary Strategies （MES） algorithm based on bi-group was proposed. In the new algorithm, the group was divided into two sub-groups, general sub-group and elite sub-group. The size of general sub-group was larger than the size of elite sub-group, in which the best individuals were stored. Evolution of the two sub-groups was parallel performed with different mutation strategies respectively, and then the group could not only explore the solution space separately, but also searched the local part in detail. Performance of this algorithm was analyzed in theory. Experimental results demonstrate that the MES algorithm is more efficient to improve convergence speed and avoid premature convergence than classical evolution strategies.
出处
《计算机应用》
CSCD
北大核心
2009年第5期1254-1256,1260,共4页
journal of Computer Applications
关键词
进化策略
双种群
变异
精英子群
普通子群
evolutionary strategy
bi-group
mutation
elite sub-group
general sub-group
