一个无惩罚型两步线性搜索算法  

A penalty-free-type two-step linear search algorithm

作　　者：孙刘平[1] 

机构地区：[1]常州纺织服装职业技术学院,江苏常州213164

出　　处：《徐州师范大学学报（自然科学版）》2009年第4期41-44,共4页Journal of Xuzhou Normal University(Natural Science Edition)

摘　　要：受Ulbrich-Ulbrich方法的启发,提出两步线性搜索算法.在算法中,每次计算一个切方向和一个法方向,再通过后退线搜索技术确定步长.通过要求法向下降量、切向下降量和函数下降量满足一定的关系来保证全局收敛性.该算法不需要使用罚函数,搜索方向的计算量比较小.最后,通过数值试验来验证算法的有效性.Inspired by the work of Ulbrich M and Ulbrich S, a two-step linear search algorithm is presented in which a tangential direction and a normal direction are computed for each time, and the step-length is determined through the backtracing search. To ensure the global convergence, it is required that there must be a certain relationship to be satisfied between the reductions in normal and tangential directions and the reduction of function. In the algorithm the penalty function need not be used, and the amount of computation for searching direction is much smaller. For testing the effectiveness of the algorithm the numerical experiments have been made.

关 键 词：无惩罚型 等式约束 线性搜索 

分 类 号：O221.2[理学—运筹学与控制论]