一类带筛选策略的改进遗传算法及其性能分析

A class of improved genetic algorithm with sifting strategy and its performance analysis

为避免遗传算法的早熟收敛,增强算法的全局搜索和局部趋化能力,在传统保优GA中引入筛选策略,即基于种群性能和种群地域差别删去一些性能相对差的冗余个体,进而维持种群的多样性.基于典型复杂函数的数值仿真结果表明,所提算法的全局收敛速度和命中全局最优的几率相对传统方法大大提高,并对参数具有较好的鲁棒性.

To avoid premature convergence of genetic algorithm (GA) and to enhance the exploration and exploitation abilities, the sifting strategy is incorporated into classic elitist GA to maintain the population diversity. That is, some bad redundant individuals are deleted from the population according to the difference of population performance and location. Numerical simulation results based on benchmark complex functions show that the convergence rate and hitting probability on global optima of the proposed algorithm are greatly better than that of the classic method, and the improved algorithm is robust on its parameters as well.