CONVERGENCE OF NONLINEAR CONJUGATE GRADIENT METHODS

• 来源期刊: 计算数学(英文版)
• 学科: 数学
• 关键词: Unconstrained optimization Conjugate gradient (generalized) Line search Global convergence.
• 年，卷（期）: 2001,（5）
• 页码范围: 539-548
• 页数: 10页
• 分类号: O1
• 语种: 英文

摘要

This paper proves that a simplified Armijo-type line search can ensure the global con vergences of the Fletcher-Reeves method and the Polak-Ribiére-Polyak method for un constrained optimization. Although it seems not possible to verify that the PRP method using the generalized Armijo line search converges globally for generally problems, it can be shown that in this case the PRP method always solves uniformly convex problems.