摘要
混洗蛙跳算法(SFLA)具有算法简单、控制参数少、易于实现等优点,但在高维难优化问题中算法容易早熟收敛且求解精度不高。导致该缺陷的主要原因是在进化后期种群多样性迅速下降,且缺乏局部细化搜索能力。借鉴差分进化算法(DE)中DE/best/1/bin版本具有全局搜索能力较强、种群多样性较好的优点,将SFLA与DE有机融合,形成混合优化算法(SFL-DE),以克服SFLA容易早熟收敛的缺陷。给出了6个30维benchmark问题数值对比实验,结果表明,在给定的较小进化代数内,SFL-DE的寻优效率、计算精度、鲁棒性等性能优于SFLA和基本DE(DE/best/1/bin和DE/rand/1/bin),不足之处是其耗时更长。
Shuffled Leaping Frog Algorithm（SFLA） is characterized by simplicity,few control parameters required,and easily be used,but has the disadvantages of premature convergence and low precision for hard high-dimensional optimization prob- lems,due to its rapid loss of the population diversity and the lack of local refined search abilities at the later stages of generations.In order to overcome the easy premature or early convergence of SFLAs,this paper hybridizes the SFLA and the Differential Evolution（DE） algorithm to form a hybrid optimization algorithm,namely SFL-DE,which borrows the idea from DE/best/1/bin strategy that has the advantages of strong global search ability and better population diversity.Comparisons are presented to test performances of the new algorithm employing 6 benchmark 30-dimensional functions.Compared with SFLA and standard DE（i.e.,DE/best/1/bin and DE/rand/1/bin schemes） algorithms,the experimental results in terms of the global optimization efficiency,the solution accuracy and the computation robustness demonstrate that the SFL-DE algorithm is a better tool for solving some benchmark optimization problems within a few fixed generations,but takes a longer run time.
出处
《计算机工程与应用》
CSCD
北大核心
2011年第18期4-8,共5页
Computer Engineering and Applications
基金
四川省应用基础研究计划项目(No.2008JY0163)
泸州市重点科技计划项目(No.2010-S-21(2/7))
关键词
混洗蛙跳算法
差分进化算法
混合优化
连续优化问题
Shuffled Frog Leaping Algorithm（SFLA）
Differential Evolution（DE） algorithm
hybrid optimization
continuous optimization problem
