无罚无滤子的修正非单调不可行QP-free方法及其全局收敛性（英文）  

Global Convergence of a Modified Nonmonotone QP-Free Method Without Penalty Function or Filter

作　　者：苏珂[1,2] 任乐乐 SU Ze;REN Lele(College of Mathematics and Information Science,Hebei University,Baoding 071002,China;Key Laboratory of Machine Learning and Computational Intelligence,Baoding 071002,China)

机构地区：[1]河北大学数学与信息科学学院,河北保定071002 [2]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》2018年第4期856-865,共10页Mathematica Applicata

基　　金：Supported by the National Natural Science Foundation of China(61572011);Hebei Provience Nature Science Foundation of China(A2018201172);the Key Research Foundation of Education Bureau of Hebei Province(ZD2015069)

摘　　要：本文提出一个解决不等式规划问题的无罚无滤子的修正非单调不可行QP-free算法.在每步迭代,只需要解两个或三个相同系数矩阵来获得搜索方向.我们利用修正的非单调技术松弛了试探点的判别准则,相比其他方法,不要求滤子结构也不涉及罚参数的选取,在一定程度上避免了Maratos效应.在合理的条件下,得到算法的全局收敛性.In this paper, an infeasible modified nonmonotone QP-free method without penalty function or filter is presented to solve inequality constrained optimization. At each iteration, it is only necessary to solve two or three linear equations with the same coefficients to obtain the search direction. We relax the acceptable criteria for trial point based on a modified nonmonotone line search technique. Compared with the existed method, the filter structure and penalty parameter are not required in this paper. Moreover, the Maratos effects are avoided to a certain degree. Under suitable conditions, the global convergence of the algorithm is proved.

关 键 词：不等式约束优化 QP-free方法 积极集 全局收敛 非单调 

分 类 号：O221.2[理学—运筹学与控制论]