求解无约束极大极小问题的光滑化不精确牛顿算法

A Smoothing Inexact Newton Algorithm for Solving Unconstrained Minimax Problem

提出了求解无约束极大极小问题的光滑化不精确牛顿算法.该算法利用光滑凝聚函数近似不可微的极大值函数,从而得到目标函数的光滑近似,进而再利用不精确牛顿法求解光滑化后的可微的无约束优化问题.在一定的假设条件下,算法具有全局收敛性,初步的数值实验表明,算法是有效的.

This paper presents a smoothing inexact Newton algorithm for solving unconstraint minimax problem.The new method uses aggregate function to approximate max-function,and uses inexact Newton method to solve the approximating smoothing unconstraint problem. Under some assumptions,global convergence of the algorithm is considered. Elementary numerical experiments show that the algorithm is effective.