基于维度分组交叉的自适应差分进化算法  

Self-adaptive differential evolution algorithm based on dimensionality group cross

作　　者：拓守恒[1] 周涛[2,3] 

机构地区：[1]陕西理工学院计算机系,陕西汉中723000 [2]宁夏医科大学理学院,宁夏银川750004 [3]陕西理工学院数学系,陕西汉中723000

出　　处：《计算机工程与设计》2011年第9期3174-3177,3181,共5页Computer Engineering and Design

基　　金：国家863高技术研究发展计划基金项目(2008AA01A303);陕西省教育厅科研基金项目(2010JK459;2010JK466)

摘　　要：为了解决求解高维复杂问题时存在收敛速度慢、求解精度低、容易陷入局部搜索等问题,提出维度分组交叉算法和自适应多模式并行差分进化算法。算法利用种群中各维变量之间的相关性进行维度划分,通过k个种群分别在k个维度划分上进行自适应差分进化,在种群间按维度划分进行交叉操作,从而较好实现了多种群的协同差分进化。最后对4个benchmark高维函数进行算法验证,实验结果表明,该算法在提高算法的寻优速度的同时,还有效地克服了早熟收敛等问题。In order to overcome the shortcoming that high-dimensional multimodal optimization problems has a low resolving speed,low convergence accuracy and traps into local optima easily,a self-adaptive differential evolution algorithm based on dimensionality group cross is proposed to resolve this problem.Firstly,using the relationship between each two dimensions in population to split the solution vector into the smaller vectors.Secondly,crossover operation is proceed in each population according to the dimensions division,then,cooperation differential evolution is realized.Finally,4 benchmark functions are used to test this algorithm.Experimental result illustrate that the proposed algorithm has some advantages in convergence velocity,solution precision,and stabilization.

关 键 词：维度分组 交叉 自适应 差分进化算法 高维 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]