一种基于种群划分的分布式自适应差分进化算法  

An Adaptive Differential Evolution Algorithm based on Distributed Framework

作　　者：黄颖[1] 周犇犇 李宝磊 幸亮 王达 HUANG Ying;ZHOU Benben;LI Baolei;XING Liang;WANG Da(School of Mathematics and Computer Science,Gannan Normal University,Ganzhou 341000,China)

机构地区：[1]赣南师范大学数学与计算机科学学院,江西赣州341000

出　　处：《赣南师范大学学报》2024年第3期23-29,共7页Journal of Gannan Normal University

基　　金：国家自然科学基金(62366003);江西省自然科学基金(20232BAB202046,20224BAB212022);江西省教育厅科技项目(GJJ211435)。

摘　　要：种群划分技术因其计算并行性在处理复杂优化问题时往往具有较高的性能,分布式差分进化算法DDE通过将整个种群划分为3个子种群,即精英种群、普通种群和劣势种群,从而构建了一个分布式框架,每个子种群独立进化并相互联系,大大提高了差分进化算法DE在处理大规模复杂优化问题时的性能,使得DDE成为一种很有前途的全局优化方法.因此,本文提出了一种基于种群划分的分布式自适应差分进化算法DE-MDF来提高DE的性能.算法性能是根据CEC 2017测试套件中所有的基准函数进行评估的.实验结果表明,与其他先进自适应DE变体相比,DE-MDF在解的质量和收敛速度方面具有明显的优势.The increasingly complex optimization problems make distributed differential evolution algorithm(DDE)a promising global optimization method.DDE is an evolutionary algorithm based on population division,which means that individuals in the population are divided into several subpopulations due to their own differences,and each subpopulation evolves independently and does not interfere with each other.Population partitioning techniques tend to have high performance when dealing with complex optimization problems due to their computational parallelism.DDE constructs a distributed framework by dividing the entire population into three subpopulations,namely elite,common,and inferior populations,with each subpopulation evolving independently and interconnected,greatly improving the performance of differential evolution algorithms(DEs)when dealing with large-scale complex optimization problems.Therefore,a distributed adaptive differential evolution algorithm(DE-MDF)based on population division is proposed to improve the performance of DE.Algorithm performance is evaluated against all benchmark functions in the CEC 2017 test suite.Experimental results show that DE-MDF has obvious advantages in solution quality and convergence speed compared with other advanced adaptive DE variants.

关 键 词：自适应 分布式 差分进化 种群划分 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]