正交Scatter Search算法  

Orthogonal scatter search algorithm

作　　者：谢柏桥[1] 戴光明[1] 谢超成[2] 陈良[1] 

机构地区：[1]中国地质大学计算机学院,湖北武汉430074 [2]西北师范大学电子与信息工程学院,甘肃兰州730070

出　　处：《计算机工程与设计》2008年第9期2266-2268,2389,共4页Computer Engineering and Design

基　　金：湖北省自然科学基金项目(2003ABA045)

摘　　要：针对Scatter Search算法在初始种群的分布欠均匀,难以搜索到全局最优,目标函数评价次数多等方面的缺点,把正交设计方法同Scatter Search算法结合起来,提出了Orthogonal Scatter Search(OSS)算法,以改进其不足,满足实际工程的需要。实验证明OSS算法优于SS算法,不仅目标函数评价次数少,收敛速度快,而且易于搜索到全局最优或者非常接近于全局最优,尤其是高维空间效果更加明显。There were some shortcomings in initial population which scattered less evenly over the feasible space, the poor ability of finding global optimal solutions and the large number of objective function evaluations in scatter search algorithm （SS）, so that combining an orthogonal design method with the SS algorithm is proposed to improve its shortcomings and satisfy the need of practical projects, the result is an hybrid algorithm called orthogonal scatter search algorithm （OSS）. The experimental results prove that the proposed algorithm is superior to the SS algorithm. It is not only the number of its objective function evaluations less and the velocity of its converge faster than the SS algorithm, but also more easy to find global optimal or extremely close-to-optimal solutions. In particularly, the performance of OSS algorithm is superior to SS algorithm in high dimension space.

关 键 词：分散搜索算法 正交设计 正交分散搜索算法 演化算法 种群多样性 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]