基于群体适应度均值商的自适应遗传算法

ADAPTIVE GENETIC ALGORITHM BASED ON MEAN QUOTIENT OF POPULATION FITNESS

在系统分析和研究自适应遗传算法特点的基础上,提出一种基于群体适应度均值商的自适应遗传算法。一方面,算法设计了群体早熟收敛的定量计算公式和引入了早熟收敛的判定阈值;另一方面,算法采用了改进的交叉和变异策略,即对陷入局部收敛的群体,依据平均适应度值将其一分为二:对性能优于平均适应度者实施先变异后交叉,而对性能劣于平均适应度者实施先交叉后变异。该算法应用在函数优化中,仿真结果表明其能有效提高全局寻优的性能,且鲁棒性好。

This paper proposed an adaptive genetic algorithm based on mean quotient of population fitness（MQF-AGA） on the basis of studying and analyzing the characteristic of canonical adaptive genetic algorithm（CAGA）.In MQF-AGA,on the one hand,the quantitative computing formula of population premature convergence is constructed and the decision threshold of premature convergence is introduced;on the other hand,the improved strategies of crossover and mutation are adopted,i.e.,the population,which falls into the local optimum,is divided into two parts according to average fitness value.The better ones whose fitness value is superior to the average fitness value are performed the mutation in advance to the crossover,whereas for the ones on the contrary,the crossover is executed before the mutation.Finally,the new algorithm was applied to function＇ s optimization and the simulation results demonstrated that the MQF-AGA can improve the performance of global search efficiently with higher robustness.