摘要
针对差分进化算法求解高维复杂优化问题存在的全局搜索能力和收敛速度不足,论文充分利用局部搜索策略、协同进化机制以及多种群进化模式,提出一种改进的多种群协作自适应差分进化(MSDPIDE)算法。MSDPIDE算法按照个体适应度的差异将个体分成不同的子种群,在多种群协同进化过程中采用局部搜索策略和协同进化机制来提高个体多样性来避免早熟收敛,保证个体之间能够进行充分高效的信息交换,以平衡局部搜索能力与全局搜索能力。通过对9个典型的Benchmarks复杂函数进行了测试,并和DE、和CADE算法进行比较,实验结果表明MSDPIDE算法能有效地避免早熟收敛,具有较高的收敛速率、较高的计算精度、较好的稳定性和较强的全局搜索能力。
For the low global searching ability and convergence speed of differential evolution algorithm(DE)in high-dimension complex function optimization,the different searching strategy and parallel evolution mechanism are used,a dynamic multiple populations parallel self-adaptive differential evolution algorithm with multiple strategies(MSDPIDE)is proposed to optimize functions in this paper.In MSDPIDE algorithm,the population is dynamically divided into multi-populations individuals according to the difference of individuals'fitness.Multiple strategies in multiple populations'parallel evolution are used to improve the individuals'diversity for avoiding premature convergence and ensure the efficiency and sufficiency information exchanging among sub-populations.In addition to,self-adaptive adjustment is introduced to automatically adjust the scaling factor and crossover factor during the running time.The MSDPIDE algorithm is tested on ten complex benchmark functions.The experiment results are compared with DE and CADE algorithms.The compared results show that the MSDPIDE algorithm takes on better searching accuracy,convergence speed,stability,remarkable global convergence ability,it is better in the searching precision,convergence speed,stability,remarkable global convergence ability,and it can avoid premature convergence effectively.
作者
周頔
ZHOU Di(Industrial Technological Institute of Intelligent Manufacturing, Sichuan University of Arts and Science, Dazhou 635000)
出处
《计算机与数字工程》
2019年第7期1648-1651,1718,共5页
Computer & Digital Engineering
基金
自然科学基金项目(编号:61304187,61771080)
四川省教育厅科技计划项目(编号:18ZA0415)资助
关键词
差分进化
多种群协作
局部搜索策略
自适应
高维函数优化
differential evolution
multiple populations cooperation
local-search strategy
self-adaptive
High-dimensional function optimization
