一个解无约束优化问题的过滤信赖域方法  

A FILTER TRUST-REGION METHOD FOR UNCONSTRAINED OPTIMIZATION

作　　者：缪卫华[1] 孙文瑜[2] 

机构地区：[1]南京审计学院应用数学系,南京210029 [2]南京师范大学数学与计算机科学学院,南京210097

出　　处：《高等学校计算数学学报》2007年第1期88-96,共9页Numerical Mathematics A Journal of Chinese Universities

基　　金：This work was supported by the national natural science foundation of china at grant10231060;Specialized Research Fund for the Doctoral Program of Higher Education(SRFDP)of China No.20040319003;the Graduates' Fund of Jiangsu Province;This author's work was also supported by the project directed by the higher shool'snatural science foundation of jiangsu province at grout 06KJD110086.

摘　　要：本文中，我们考虑一般的无约束极小化问题：minx∈R^nf（x）,（1,1）其中f：R^n→R二次连续可微．N.I.M. Gould, C. Sainvitu and Ph.L. Toint [9] presented a filter technique for unconstrained optimization. Based on this work, we present a new filter trust-region algorithm without judgement of convexity of the trust-region subproblem. We prove that under certan conditions, the algorithm converges globally to a second order stationary point. Numerical results show that the presented algorithm is more efficient than the basic trust-region algorithm.

关 键 词：无约束优化问题 信赖域方法 过滤 极小化问题 连续可微 R^N 

分 类 号：O224[理学—运筹学与控制论]