函数优化的遗传算法策略优选  

Genetic algorithm strategy selection for function optimization

作　　者：洪玮[1] 崔杜武[1] 

机构地区：[1]西安理工大学计算机科学与工程学院,陕西西安710048

出　　处：《计算机工程与设计》2010年第13期3043-3046,共4页Computer Engineering and Design

基　　金：国家自然科学基金项目(60873035);陕西省自然科学基金项目(2006F43)

摘　　要：为了提高函数优化的准确性和效率,提出一种基于表达式构造的函数聚类和策略优选的方法。使用英国Sheffield大学开发的Matlab遗传算法工具箱(GATBX)设计不同的算法策略,对随机选取的3种常见的函数构造因子按不同比例组合得到的不同模式进行了策略试算,以收敛率、平均截止代数及截止代数分布熵作为由主到次的性能评价指标来优选策略,并归纳出规则。最后利用4个具有试验模式的数值函数验证了规则的有效性。To improve the accuracy and effectiveness of function optimization,a method of clustering functions and selecting algorithm strategies,is proposed based on the expression construction.In the experiment,different algorithm strategies are designed using the Matlab Genetic Algorithm ToolBox（GATBX）,which is developed by Sheffield University in England.Three common function components are chosen randomly,and generated several patterns with different combination,and they are tested with different strategies respectively.The convergence rate,the average stopped generation,and the distribution entropy of stopped generations are used to evaluate the performance of algorithm strategies as indices from primary to subordinate.Then some rules are summed up from these experimental results.At last,the validity of the rules is verified by four numerical functions with experimented patterns.

关 键 词：遗传算法 函数优化 表达式构造 聚类 策略优选 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]