吉林大学学报(理学版)

• 数学 • 上一篇    下一篇

基于自适应Barzilai-Borwein步长的直接搜索共轭梯度法

刘加会1, 刘红卫1, 杨善学1,2   

  1. 1. 西安电子科技大学 数学与统计学院, 西安710126;  2. 西安财经学院 统计学院, 西安 710100
  • 收稿日期:2016-06-21 出版日期:2017-05-26 发布日期:2017-05-31
  • 通讯作者: 刘加会 E-mail:liujiahui05@163.com

Direct Search Conjugate Gradient Method Based onAdaptive BarzilaiBorwein StepSize

LIU Jiahui1, LIU Hongwei1, YANG Shanxue1,2   

  1. 1. Shool of Mathematics and Statistics, Xidian University, Xi’an 710126, China;2. School of Statistics, Xi’an University of Finance and Economics, Xi’an 710100, China
  • Received:2016-06-21 Online:2017-05-26 Published:2017-05-31
  • Contact: LIU Jiahui E-mail:liujiahui05@163.com

摘要: 利用最新迭代点附近的函数值信息, 估计该点的单纯形梯度, 并计算当前点的BarzilaiBorwein(BB)步长, 提出一种基于自适应BB步长的网格步长更新策略, 有效解决了网格步长下降过快的问题, 同时结合新的正基更新策略提出一种新的直接搜索算法. 数值结果表明, 该算法在稳定性和效率上有较大改进.

关键词: 共轭梯度, 直接搜索, 自适应BarzilaiBorwein(BB)步长, 单纯形梯度

Abstract: Using the information of the function value near the latest iteration point, we estimated simplex gradient of the point and calculated BarzilaiBorwein (BB) stepsize of the current point. We proposed an updating strategy of the grid stepsize based on adaptive BB stepsize, which could effectively solve the problem that the grid stepsize dropped too fast. Combined with the new updating strategy, we proposed a new direct search algorithm. The numerical results show that the algorithm has a great improvement in stability and efficiency.

Key words: direct search, simplex gradient, conjugate gradient, adaptive BarzilaiBorwein (BB) stepsize

中图分类号: 

  • O224