不等式约束优化问题的可行信赖域滤子法  

A feasible Trust Region Filter Method for Inequality Constrained Optimization Problems

作　　者：王祥玲 左双勇 Wang Xiangling;Zuo Shuangyong(Department of Primary Education,Yichun Early Childhood Teachers College,Yichun Jiangxi,330814)

机构地区：[1]宜春幼儿师范高等专科学校初等教育学院,江西宜春330814

出　　处：《保山学院学报》2018年第5期47-49,共3页JOURNAL OF BAOSHAN UNIVERSITY

基　　金：宜春市社科研究"十三五"规划项目(项目编号:YCSK2018-106;YCSK2018-115)

摘　　要：利用可行信赖域方法和滤子方法的思想,建立了一个新的求解不等式约束优化问题的算法;算法利用修正的二次规划子问题获得可行搜索方向,当迭代点不能被滤子接受时对搜索方向进行修正,并减小信赖域半径。算法既保证了QP子问题的可行性,又在一定程度上简化了算法结构;在合适的条件下,证明了算法具有全局收敛性。Using the idea of feasible trust region method and filter method,a new algorithm forsolving inequality constrained optimization problem is established. The algorithm uses the modi-fied quadratic programming subproblem to obtain the feasible search direction. When the iterationpoint cannot be accepted by the filter, the search direction is corrected and the trust region ra-dius is reduced. The algorithm not only ensures the feasibility of the QP sub-problem, but alsosimplifies the algorithm structure to some extent. Under suitable conditions, the global conver-gence of the algorithm is proved.

关 键 词：不等式约束优化 可行信赖域 滤子 全局收敛性 

分 类 号：O13[理学—数学]