四川轻化工大学学报（自然科学版）2025，Vol.38Issue(2)：47-53,7.DOI:10.11863/j.suse.2025.02.06
随机差量学习优化算法
Random Difference Learning Optimization Algorithm
摘要
Abstract
A random difference learning(RDL)algorithm is proposed in the present study,using the differential evolution(DE)algorithm to generate cross components from individual differences in a population.Firstly,two individuals are randomly selected from a population,after subtracting them,random weights are assigned to each dimensional difference to obtain the random learning difference of the new individual.Then,one of the selected individuals is fused with the difference to generate a new individual.The global random reachable mutation mechanism is introduced into the RDL,and the probability of new individuals obtained through mutation operation being distributed to any region in the search feasible domain during the search process is setted greater than 0,ensuring the global performance of the algorithm.At the same time,the elite retention strategy is added to form a new alternative population from parent and child individuals,and sorting method is used for survival of the fittest evolution to retain the elite and generate the next generation,improving the convergence speed of the algorithm.Commonly used benchmark testing functions are adopted to test the RDL algorithm,the test results are compared with various excellent algorithms.The results show that the proposed RDL algorithm has at least one order of magnitude higher convergence accuracy than other algorithms,and the number of iterations required to achieve the target accuracy is about 25%of other algorithms.关键词
随机差量学习/差分进化算法/全域随机可达变异/精英保留Key words
random difference learning/differential evolution algorithm/global random reachable mutation/elite retention分类
通用工业技术引用本文复制引用
杜正淼,谭飞,王志强,王梦莎..随机差量学习优化算法[J].四川轻化工大学学报（自然科学版）,2025,38(2):47-53,7.基金项目
国家自然科学基金项目(61902268) （61902268）
四川省科技项目(21ZDYF4052 （）
2020YFH0124 （）
2021YFSY0060) （）
四川轻化工大学创新计划项目(cx2023193 （）
cx2023195 （）
cx2023198) （）
