基于双重经验结合的自适应差分进化算法  

Adaptive differential evolution algorithm based on dual experience combination

作　　者：郭肇禄[1,2] 向传娇 杨火根 张文生[2] GUO Zhaolu;XIANG Chuanjiao;YANG Huogen;ZHANG Wensheng(School of Science,Jiangxi University of Science and Technology,Ganzhou 341000,Jiangxi China;Institute of Automation,Chinese Academy of Sciences,Beijing 100190,China)

机构地区：[1]江西理工大学理学院,江西赣州341000 [2]中国科学院自动化研究所,北京100190

出　　处：《华中科技大学学报（自然科学版）》2024年第6期171-178,共8页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(12161043,61662029);江西省自然科学基金资助项目(20192BAB201007);江西省教育厅科技项目(GJJ160623,GJJ170495);江西理工大学青年英才支持计划项目(2018)。

摘　　要：针对传统差分进化算法在解决复杂优化问题时性能不足的问题,提出了一种双重经验结合的自适应差分进化算法,该算法提出了基于个体经验和集体经验结合的参数自适应机制.在该机制中,每个个体都有自己的缩放因子和杂交概率,并且个体同时利用自身经验和多个成功个体的集体经验来自适应地更新参数值.该机制不仅很好地利用了个体自身的演化信息,还结合了集体的有益信息,有利于生成优秀个体,提高算法性能.此外,自适应差分进化算法设计了一种新的带外部存档的变异策略,该变异策略引入了一个调整变异策略贪婪性的参数,这个参数在进化过程中随着函数评价次数的增加而动态变化,自适应地调整变异策略在不同进化阶段的贪婪性,较好地平衡了算法的勘探和开采,进而提高算法性能.在CEC2017基准集上对算法进行数值实验,并将自适应差分进化算法与多个改进的差分进化算法进行了比较.实验结果表明:自适应差分进化算法取得了较好的求解结果,并在整体上优于其他算法.To improve the performance of the traditional differential evolution for solving some complex optimization problems,an adaptive differential evolution based on dual experience combination was proposed.Adaptive differential evolution based on dual experience combination proposed a parameter adaptive mechanism based on the combination of individual experience and collective experience.In this mechanism,each individual adaptively updated its scaling factor and crossover probability by using its own experience and the collective experience of multiple successful individuals.It not only utilized the individual’s evolutionary information well but also combined the beneficial information of the collective,which facilitated the generation of good individuals and improved the performance.In addition,a new mutation strategy with external archiving was proposed.In this mutation strategy,a parameter was introduced to adaptively adjust its greediness at different evolutionary stages.This parameter changed dynamically as the number of function evaluations increased,which better balanced exploration and exploitation and improved the performance.The experiments were conducted on the CEC2017 benchmark suite,and adaptive differential evolution based on dual experience combination was compared with several improved differential evolution algorithms.Experimental results show that adaptive differential evolution based on dual experience combination achieves better solution results,and outperforms other comparative algorithms overall.

关 键 词：差分进化 个体经验 集体经验 参数自适应机制 变异策略 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]