差异演化算法及其改进形式的综述  

Overview of differential evolution algorithm and its improved algorithms

作　　者：陈良[1] 戴光明[1] 张全元[1] 谢柏桥[1] 

机构地区：[1]中国地质大学计算机学院,湖北武汉430074

出　　处：《计算机工程与设计》2008年第1期131-134,255,共5页Computer Engineering and Design

基　　金：湖北省自然科学基金项目(2003ABA045)

摘　　要：差异演化算法是一种基于群体差异的演化算法,群体中每个向量代表问题的一个候选解,该算法利用向量之间的差异扰动整个种群,求解问题的最优解。综述了差异演化算法的基本原理、常用测试函数和算法优缺点,讨论了多种改进手段,如:三角变异、混沌理论、逆向计算、均匀设计表等,并给出了未来可能的研究方向。Differential evolution （DE） algorithm is an evolution algorithm based on population differences. Each vector of population represents one candidate solution of the problem optimized. The algorithm finds the best result by perturbing the whole population with the differences between vectors. The basic principle, benchmark test functions, strong points and shortcomings of DE are reviewed. Some improved methods, such as trigonometric mutation, opposition-based learning, uniform design method, are discussed. And future research issues are also given.

关 键 词：差异演化 全局最优 三角变异 逆向差异演化 混沌 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]