基于抽象凸估计选择策略的差分进化算法  

Differential evolution algorithm based on abstract convex underestimate selection strategy

作　　者：周晓根 张贵军[1] 梅珊[1] 明洁[1] 

机构地区：[1]浙江工业大学信息工程学院,浙江杭州310023

出　　处：《控制理论与应用》2015年第3期388-397,共10页Control Theory & Applications

基　　金：国家自然科学基金项目(61075062;61379020);浙江省自然科学基金项目(LY13F030008);浙江省科技厅公益项目(2014C33088);浙江省重中之重学科开放基金资助项目(20120811);杭州市产学研合作项目(20131631E31)资助~~

摘　　要：针对传统差分进化算法计算代价、可靠性及收敛速度的问题,提出一种基于抽象凸估计选择策略的差分进化算法(DEUS).首先,通过提取新个体的邻近个体建立局部抽象凸下界松弛模型;然后,利用下界松弛模型估计目标函数值来指导种群更新,同时利用下界估计区域极值点快速枚举算法系统排除部分无效区域;最后,借助线性拟凸包络的广义下降方向有效地实现局部增强.12个标准测试函数的实验结果表明,所提算法计算代价、可靠性及收敛速度均优于DE及DERL,DELB,Sa DE等改进算法.To solve the problems of computational cost, success rate and convergence speed in the conventional dif- ferential evolution algorithm, we propose a new differential evolution algorithm based on abstract convex underestimate selection strategy （DEUS）. Firstly, the local abstract convex lower relaxed model is constructed by extracting the neigh- boring individuals of the new individual. Then, the underestimate values which are estimated through the lower relaxed model are used to guide the update process of the population, and some invalid regions of the domain where the global optimum cannot be found are systematically excluded by using the fast enumeration algorithm of the local minimum in the underestimate regions. Finally, the generalized descent directions of the linear quasi convex envelope are employed for local enhancement. Experiments results of 12 benchmark functions show that the proposed algorithm is superior to DE, DERL, DELB and SaDE algorithm in terms of computational cost, success rate and convergence speed.

关 键 词：差分进化 全局优化 下界估计 抽象凸 支撑向量 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]