摘要
本文探讨了求解线性约束不可微凸规划极小化问题,给出了一类高阶算法,该算法模仿了变尺度思想,应用了 Kiwiel 聚合次梯度思想,试图改善逼近程度,提高收敛速度,并证明了算法有较好的收敛性.
This paper discusses a class of linear constrained nonsmooth convex programming,an algorithems was given,which was derived by combing kiwiel' s aggregate subgrabient and variable metric method,so as to improve degree of approximation and rate of convergence,also was proved to have good convergence property.
出处
《辽宁大学学报(自然科学版)》
CAS
1993年第1期35-41,共7页
Journal of Liaoning University:Natural Sciences Edition
基金
国家自然科学基金
关键词
不可微凸规划
高阶算法
变尺度法
Nonsmooth convex programming
Higher-order algorithem
Variable metric method
Aggregate subgradient