一种改进的求解极大极小问题的非单调滤子法  

A Modified Nonmonotone Filter Method for Minimax Problems

作　　者：苏珂[1] 王晨[1] 李小川 SU Ke;WANG Chen;LI Xiaochuan(College of Mathematics and Information Science,Key Laboratory of Machine Learning and Computational Intelligence,Hebei University,Baoding 071002,China)

机构地区：[1]河北大学数学与信息科学学院,河北省机器学习与计算智能重点实验室,河北保定071002

出　　处：《应用数学》2020年第2期358-372,共15页Mathematica Applicata

基　　金：Supported by the National Natural Science Foundation of China(61572011);Hebei Provience Nature Science Foundation of China(A2018201172);Foundation of Hebei Educational Committee(QN2019142)。

摘　　要：本文提出一种求解极小极大问题的非单调信赖域滤子法.该算法基于滤子技术,放松了试验点的可接受准则,与已有的求解极大极小问题的序列二次规划牛顿法(SQP)相比,我们的方法具有更大的灵活性.在适当的条件下,建立了全局收敛性.最后进行了数值实验.In this paper,we propose a nonmonotone trust region filter method for minimax problems.In the presented algorithm,based on the filter technique,the acceptable criterion of the trial points is relaxed,so compared to the existing Seqential quadratic programming(SQP)Newton-type methods for minimax problems,our method is more flexible.Under some suitable conditions,the global convergence properties are established.The numerical tests are reported in the end.

关 键 词：滤子法 极大极小问题 非单调 信赖域 全局收敛性 

分 类 号：O221.2[理学—运筹学与控制论]