混合互补问题牛顿型算法的二阶收敛性  

Quadratic Convergence of Newton-type Algorithms for Mix-complementarity Problems

作　　者：马昌凤[1] 陈新美[1] 

机构地区：[1]长沙电力学院数学与计算机系,长沙410077

出　　处：《数学物理学报（A辑）》2000年第2期145-151,共7页Acta Mathematica Scientia

基　　金：湖南省自然科学基金资助课题! (98JJY2 0 53);长沙电力学院科学基金资助项目

摘　　要：在凸规划理论中 ,通过 KT条件 ,往往将约束最优化问题归结为一个混合互补问题来求解 .该文就正则解和一般解两种情形分别给出了求解混合互补问题牛顿型算法的二阶收敛性的充分性条件 。In convex programming theory, a constrained optimization problem, by KT conditions, is usually converted into a mixed nonlinear complementarity problem. Accor-ding to regular solution and general solution, we in the paper describe and establish a sufficient condition under which the Newton-type algorithm possesses quadratic convergence property when it is applied to solving mix-complementarity problems. In addition, we also show that, when the step-size is suitably chosen, the inexact Newton's method and the discrete Newton's method converge quadratically.

关 键 词：混合互补问题 牛顿型算法 二阶收敛性 最优化 

分 类 号：O224[理学—运筹学与控制论]