求解不可微凸优化问题的充分下降共轭梯度法  

Sufficient Descent Conjugate Gradient Methods for Solving Nondifferentiable Convex Optimization Problem

作　　者：黄元元[1] HUANG Yuanyuan(Mathematics&Statistics School,Henan University of Science&Technology,Luoyang 471023,China)

机构地区：[1]河南科技大学数学与统计学院,河南洛阳471023

出　　处：《河南科技大学学报（自然科学版）》2021年第2期94-99,M0008,共7页Journal of Henan University of Science And Technology:Natural Science

基　　金：国家自然科学基金项目(12071112,62073124,11701150)。

摘　　要：给出求解不可微凸优化问题的一类具有充分下降条件的共轭梯度法。该类方法是基于Moreau-Yosida正则化策略和传统的共轭梯度法设计的。理论分析表明:该类方法在一定条件下是全局收敛的。A class of conjugate gradient methods with sufficient descent condition for nondifferentiable convex optimization problem was presented.These methods were designed based on the Moreau-Yosida regularization strategy and traditional conjugate gradient methods.The theory analysis demonstrates that these methods are globally convergent under some conditions.

关 键 词：不可微凸优化 共轭梯度法 充分下降条件 Moreau-Yosida正则化 

分 类 号：O224[理学—运筹学与控制论]