两类变参数梯度法及收敛性

Convergence of Gradient Methods with a Variable Parameter for Unconstrained Optimization

本文在 L MINN方法的基础上 ,提出了两类变参数梯度法 ,然后证明了这两类方法在非精确线性搜索的

In this paper, author present two class of gradient methods with a parameter in the choice of the scalar β k , and the b class and the μ class of these methods are first improved, then two class of gradient methods with a variable parameter are attained and it is shown that these two class of gradient methods are both descent methods and of global convergence when steplength satisfy general Wolfe conditions.