《郑州大学学报(工学版)》[ISSN:1671-6833/CN:41-1339/T]
卷:39卷
期数:2018年03期
页码:15-21
栏目:出版日期:2018-05-10
Title:A Survey of Evolutionary Algorithms for Large Scale Optimization Problem
作者:
1.郑州大学电气工程学院,河南郑州,450001;2.中原工学院电子信息学院,河南郑州,450007
Author(s):
Liang Jing1; Liu Rui1; Qu Boyang2; Yue Caitong1
1. School of Electrical Engineering, Zhengzhou University, Zhengzhou, Henan, 450001; 2. School of Electronic Information, Zhongyuan Institute of Technology, Zhengzhou, Henan, 450007
关键词:
大规模优化问题; 进化算法; 协同进化; 种群初始化; 基准测试函数
Keywords:
large scale optimization problem; evolutionary algorithm; cooperative coevolution; population initialization; benchmark function
DOI:
10.13705/j.issn.1671-6833.2017.06.016
文献标志码:
A
摘要:
针对大规模问题的特点,对已有的大规模优化算法进行了简单的分析.主要介绍算法的初始化方法、不分组策略、静态分组策略、动态分组策略、自适应分组策略、大规模优化算法测试函数集以及算法结果的对比等方面;侧重描述优化算法的搜索策略、更新策略、突变策略和协同进化策略,并列出大规模优化算法测试函数集的特点及优化算法的评价方法;最后,给出了目前大规模优化问题的几个研究难点。
Abstract:
Based on the characterisities of large-scale problems, lager-scale optimization were grossly analyzed. This paper introduced some methods for lager-scale problems.The methods included the initialization method, decomposition strategy, updating strategy and so on. This paper mainly focued on the search strategy, update strategy, mutation strategy and cooperative coevolution. Meanwhile, the characteristics of lager-scale optimization algorithm testing function set and evaluation method were listed. Finally, the future research directions were given.
笔记
LSGO大规模全局优化问题
种群初始化方法:随机、定值设定、两步式、混合、具体应用法
不分组策略,子代产生策略,基于广义反向学习的DE算法GOBL generalized opposition-based learning
新的变异策略,连续差分进化邻域搜索算法sequential differential evolution enhanced by neighborhood search SDENS
自适应策略,广义自适应差分进化算法generalized adaptive DE GaDE
局部搜索策略,基于局部搜索的动态多种群粒子群优化算法dynamic multi-swarm particle swarm optimizer with local search DMS-L-PSO
社会学习的粒子群优化算法social learning particle swarm optimization algorithm SL-PSO
小种群搜索策略minimum population search MPS
分组策略
协同进化策略,由分治策略进化,逐步分解cooperative coevolution CC
结论
分组优化式主要方法 但决策变量之间是否可分限制了协同进化策略 判断是否可分的代价随着变量增多而迅速加大
全部可分和全部不可分,可分的对于每个维度单独优化比较好,但对于不可分的分组方式是难题
不平衡特性的测试函数cec2013特定设计