An improved differential evolution algorithm with dual mutation strategies collaboration
An improved differential evolution algorithm with dual mutation strategies collaboration
复制标题
一种改进的双突变策略协作差分进化算法
DOI：
10.1016/j.eswa.2020.113451
复制
发表时间：
2020-09-01
影响因子：
8.5
通讯作者：
Yang, Bo
中科院分区：
文献类型：
--
作者：
Li, Yuzhen;Wang, Shihao;Yang, Bo
为了减少变异策略和控制参数的选择对差分进化算法性能的影响，提出了一种改进的双变异策略协同差分进化算法（DMCDE），并对DMCDE进行了两方面的改进.首先，DMCDE引入了精英指导机制，提出了两个新的变种的经典DE/兰德/2和DE/best/2变异策略，我们称之为DE/e-rand/2和DE/e-best/2。它们采用从上级精英群体中随机选取的个体作为基础向量和差异向量的第一向量，从而在不丧失随机性的前提下为个体变异提供更清晰的指导。其次，采用双变异策略协作机制，在算法的全局探索和局部利用之间取得平衡。通过使用常用的测试函数以及现实世界的优化问题来评估DMCDE的性能。结果表明，DMCDE能显著提高DE的优化性能，优于同类竞争者的上级性能。(C)2020爱思唯尔有限公司版权所有。
To reduce the effect of the selections of mutation strategies and control parameters on the performance of differential evolution (DE), this paper proposes an improved differential evolution algorithm with dual mutation strategies collaboration (DMCDE), in which two main improvements are presented. First, DMCDE introduces an elite guidance mechanism to propose two new variants of the classical DE/rand/2 and DE/best/2 mutation strategies, which we call DE/e-rand/2 and DE/e-best/2 respectively. They use the individuals randomly chosen from superior elite population as the base vector and the first vector of difference vectors, thereby providing clearer guidance for individual mutation without losing randomness. Second, a mechanism of dual mutation strategies collaboration is utilized to obtain a trade-off between global exploration and local exploitation of the algorithm. The performance of DMCDE is evaluated by using the commonly used test functions as well as a real-world optimization problem. The results show that DMCDE can significantly improve the optimization performance of DE, and is superior to the comparative competitors. (C) 2020 Elsevier Ltd. All rights reserved.