摘要
提出一类带调正搜索方向的有限内存变尺度方法,并且证明这类方法对一致凸目标函数具有线性收敛速度,对连续可微函数具有整体收敛性。
In this paper, we propose a class of limited memory BFGS methods with revised search direction. We prove that these mothods are globally convergent for continuously differentiable functions, and they are linear convex for uniformly convex functions.
基金
国家自然科学基金!19371055
关键词
非线性规划
变尺度方法
搜索方向
凸函数
收敛性
non-linear programming, BFGS method, search direction, convergence (mathematics), uniformly convex