摘要
针对传统差分进化算法在求解高维优化问题时,耗时长、精度不高的问题,提出一种基于多种群机制的混合策略的云差分进化算法。将种群划分成若干子种群,各子种群采用不同的策略并行进化;种群进化若干代数后,按拓扑结构进行个体迁移,增加多样性,提高寻优机率。利用MapReduce模型,将子种群分发到集群上并行,提高求解速度。仿真结果表明,该算法在求解1000维的13个优化问题时,能取得较好的精度,提高求解效率。
Generally,traditional differential evolution algorithm in solving high dimension optimization problem is time-consuming and low accuracy.A cloud differential evolution algorithm with hybrid strategy was proposed by the use of multiple population mechanism.The population was divided into several sub populations,and each sub population adopted different strategies to evolve in parallel.To increase the diversity of the population and to improve the probability on obtaining optimum,the indivi-duals were migrated between sub-populations according to topology structure in several generations.Through MapReduce mo-del,the sub populations were distributed to cluster in parallel to improve the speed.The simulation results show that the algorithm is more efficient and it can achieve better precision in solving 13 different types of optimization problems with 1000 dimension.
作者
袁斯昊
邓长寿
董小刚
范德斌
殷超
YUAN Si-hao;DENG Chang-shou;DONG Xiao-gang;FAN De-bin;YIN Chao(School of Information Science and Technology,Jiujiang University,Jiujiang 332005,China)
出处
《计算机工程与设计》
北大核心
2018年第9期2792-2799,共8页
Computer Engineering and Design
基金
国家自然科学基金项目(61364025
61662038
61763019)
江西省教育厅科技基金项目(GJJ151081
GJJ161072
GJJ161076)
关键词
差分进化
高维优化
多种群
多策略
云计算
differential evolution
high dimensional optimization
multiple population
multiple strategy
cloud computing
