J4

• 数学 • 上一篇    下一篇

一种无约束优化的非单调拟牛顿信赖域算法

孙中波, 段复建   

  1. 桂林电子科技大学 数学与计算科学学院, 广西 桂林 541004
  • 收稿日期:2008-10-06 修回日期:1900-01-01 出版日期:2009-05-26 发布日期:2009-06-23
  • 通讯作者: 孙中波

Nonmonotone QuasiNewton Trust Region Algorithm forUnconstraint Optimization

SUN Zhongbo, DUAN Fujian   

  1. School of Mathematics and Computational Science, Guilin University of Electronic Technology, Guilin 541004,Guangxi Zhuang Autonomous Region, China
  • Received:2008-10-06 Revised:1900-01-01 Online:2009-05-26 Published:2009-06-23
  • Contact: SUN Zhongbo

摘要: 提出一种新的信赖域算法, 使子问题产生的试探步 dk始终保持在信赖域中, 在每次迭代过程中, 试探步dk均能得到校正, 当试探步不被接受时, 采用非单调线搜索技术, 无需重解子问题. 在适当的条件下, 证明了算法的全局收敛性.

关键词: 非单调线搜索, 拟牛顿法, 信赖域算法, 全局收敛性

Abstract: he subproblem generates trail step size dk always in trust region. dk is updated at every iter ation. Nonmonotone line search is used to solve subproblem when dk is not accepted. Global convergence is proved under some suitable conditions.

Key words: nonmonotone line search, quasiNewton method, trust region algorithm, global convergence

中图分类号: 

  • O224.2