首页 | 本学科首页   官方微博 | 高级检索  
     

带有固定步长的非单调信赖域方法
引用本文:莫降涛,刘春燕,颜世翠. 带有固定步长的非单调信赖域方法[J]. 曲阜师范大学学报, 2006, 32(3): 30-34
作者姓名:莫降涛  刘春燕  颜世翠
作者单位:广西大学数学与信息科学学院,530004,广西壮族自治区南宁市;广西大学数学与信息科学学院,530004,广西壮族自治区南宁市;广西大学数学与信息科学学院,530004,广西壮族自治区南宁市
摘    要:该文提出一种新的非单调信赖域方法.当试探步不能被接受时,算法沿着试探步的方向求得下一个迭代点,其中步长利用固定公式计算.这种方法既避免了重复求解信赖域子问题,又减少了线搜索方法计算函数值的次数.该文采用的非单调策略是基于张洪超和Hanger(2004)出的非单调线搜索技术.在适当的条件下,证明了算法的全局收敛性及超线性收敛性.最后给出了初步的数值实验结果.

关 键 词:固定步长  非单调线搜索  非单调信赖域方法  全局收敛性  超线性收敛性
文章编号:1001-5337(2006)07-0030-05
收稿时间:2005-09-19
修稿时间:2005-09-19

A Nonmonotone Trust Region Method with Fixed Stepsize
MO Jiang-tao,LIU Chun-yan,YAN Shi-cui. A Nonmonotone Trust Region Method with Fixed Stepsize[J]. Journal of Qufu Normal University(Natural Science), 2006, 32(3): 30-34
Authors:MO Jiang-tao  LIU Chun-yan  YAN Shi-cui
Affiliation:College of Mathematics and Information Science, Guangxi University, 530004, Nanning, Guangxi, PRC
Abstract:In this paper,a new nonmonotone trust region method is presented.If the trial step is rejected,the method finds a new iteration point along the trial step and the stepsize is computed by a fixed formula.The method not only avoids resolving the subproblem repeatedly but also reduces the numbers of computing function values in the line search algorithm.The nonmonotone technique applied to the method is based on the nonmonotone line search technique proposed by Zhang and Hanger(2004).The global convergence and superlinear convergence results of algorithm are preserved under suitable conditions.Some preliminary numerical results are given.
Keywords:fixed stepsize   nonmonotone line search technique   nonmonotone trust region method   global convergence   superlinear convergence
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号