摘要
针对差分算法(differential evolution,DE)在解决高维优化问题时参数设置复杂、选择变异策略困难的现象,提出了广义逆向学习方法的自适应差分进化算法(self-adaptive DE algorithm via generalized opposition-based learning,SDE-GOBL)。利用广义的逆向学习方法(generalized opposition-based learning,GOBL)来进行多策略自适应差分算法(Self-adaptive DE,Sa DE)的初始化策略调整,求出各个候选解的相应逆向点,并在候选解和其逆向点中选择所需要的最优初始种群,然后再进行自适应变异、杂交、选择操作,最后通过CEC2005国际竞赛所提供的9个标准测试函数对SDE-GOBL算法进行验证,结果证明该算法具有较快的收敛速度和较高的求解精度。
The problem related to defects of complex parameter setting and difficult selection of mutation strategies existing in the differential evolution( DE) algorithm when solving high-dimensional optimization problem is studied.This paper proposed a new self-adaptive DE algorithm based on generalized opposition-based learning( SDEGOBL). The generalized opposition-based learning( GOBL) is utilized for the adjustment of initiation strategy on multi-strategy self-adaptive DE( Sa DE) algorithm. The corresponding reverse points of each candidate solution are figured out. In addition,the necessary optimal initial population is selected among the candidate solutions and its reverse points. Next,the self-adaptive mutation,hybridization and selection operations are conducted. Finally,nine standard test functions provided in CEC2005 International Competition are applied for demonstrating SDE-GOBL algorithm. The result showed that the algorithm has fast convergence speed and high solution precision.
出处
《智能系统学报》
CSCD
北大核心
2015年第1期131-137,共7页
CAAI Transactions on Intelligent Systems
基金
国家自然科学基金资助项目(61034008
61203099
61225016)
北京市自然科学基金资助项目(4122006)
教育部博士点新教师基金资助项目(20121103120020)
北京市科技新星计划资助项目(Z131104000413007)
关键词
差分算法
优化
自适应
逆向学习
收敛速度
精度
高维
初始化
differential evolution
optimization
generalized opposition-based learning
convergencespeed
accuracy
highdimension
initialization