非精确搜索下的超记忆梯度法及其收敛性

应用科学学报 ›› 2003, Vol. 21 ›› Issue (3): 241-243.

非精确搜索下的超记忆梯度法及其收敛性

时贞军^1,2

1 大连理工大学应用数学系辽宁大连 116024;
2 曲阜师范大学运筹与管理学院山东日照 276826

收稿日期:2002-07-03 修回日期:2002-12-02 出版日期:2003-09-10 发布日期:2003-09-10
作者简介:时贞军(1963-),男,山东新泰人,教授,博士
基金资助:
国家自然科学基金资助项目(10171054)

The Supermemory Gradient Method with Inexact Line Searches

SHI Zhen-Jun^1,2

1 Department of Applied Mathematics, Dalian University of Technology, Dalian 116024, China;
2 College of Operations Research and Management, Qufu Normai University, Rizhao 276826, China

Received:2002-07-03 Revised:2002-12-02 Online:2003-09-10 Published:2003-09-10

摘要/Abstract

摘要： 提出一种新的无约束优化超记忆梯度算法,算法在每步迭代中充分利用前面迭代点的信息产生下降方向,采用Armijo搜索产生搜索步长,在较弱的条件下证明了算法的全局收敛性.

关键词: 无约束优化, Armijo线性搜索, 全局收敛性, 超记忆梯度法

Abstract: The paper presents a new supermemory gradient algorithm for unconstrained minimiza-tion problems. The algorithm makes full use of the previous iterative information to generate the search direction at each iteration and uses Armijo's line search to find the stepsize. We can prove its global convergence under mild conditions.

Key words: unconstrained minimization, supermemory gradients, global con-vergence, inexact line search

中图分类号:

O221.2

时贞军. 非精确搜索下的超记忆梯度法及其收敛性[J]. 应用科学学报, 2003, 21(3): 241-243.

SHI Zhen-Jun. The Supermemory Gradient Method with Inexact Line Searches[J]. Journal of Applied Sciences, 2003, 21(3): 241-243.