二次Lagrange插值模型直接搜索法的收敛性  

ON THE CONVERGENCE OF DERIVATIVE FREE METHODS WITH QUADRATIC LAGRANGE INTERPOLATION MODELS

作　　者：刘浩[1] 倪勤[2] 

机构地区：[1]南京工业大学理学院,南京210009 [2]南京航空航天大学理学院,南京210016

出　　处：《高等学校计算数学学报》2011年第2期109-121,共13页Numerical Mathematics A Journal of Chinese Universities

基　　金：supported by the National Natural Science Foundation of China(10471062) ;Natural Science Foundation of Jiangsu Province(BK2006184)

摘　　要：A class of derivative-free trust region methods is proposed for solving unconstrained optimization problems.These methods use quadratic Lagrange interpolation models.Under mild conditions,the global convergence of this type of methods is proved without any information on the gradient of the objective function.A class of derivative-free trust region methods is proposed for solving unconstrained optimization problems. These methods use quadratic Lagrange interpolation models. Under mild conditions, the global convergence of this type of methods is proved without any information on the gradient of the objective function.

关 键 词：derivative free optimization quadratic interpolation model Lagrange interpolation model A-poisedness trust region method 

分 类 号：O221.2[理学—运筹学与控制论]