一种新的非单调梯度路径线搜索方法(英文)

A New Nonmonotone Gradient-Path Algorithm for Unconstrained Optimization

通过近似处理割线方程提出一种解无约束优化问题的单调梯度路径算法.其中,非单调技术用于加速目标函数的收敛过程.理论分析给出了算法的弱全局收敛性,数值结果表明了算法的有效性.

This paper presents a nonmonotone gradient-path algorithm by approximating the secant equation for unconstrained optimization problem. The nonmonotone criterion is used to speed up the convergence progress of objective function. Theoretical analysis is given which proves that the proposed algorithm is weakly globally convergent. The results of numerical experiments are reported to show the effectiveness of the proposed algorithm.