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.
作者
苏珂
王晨
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)
出处
《应用数学》
CSCD
北大核心
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)。
关键词
滤子法
极大极小问题
非单调
信赖域
全局收敛性
Filter method
Minimax problem
Nonmonotone
Trust region
Global convergence
