一类超线性收敛的既约变尺度法  

A CLASS OF SUPERLINEAR CONVERGENT REDUCED-VARIABLE METRIC ALGORITHM

作　　者：简金宝[1] 

机构地区：[1]广西大学数学与信息科学系

出　　处：《系统科学与数学》1996年第1期78-80,共3页Journal of Systems Science and Mathematical Sciences

基　　金：广西青年自然科学基金;广西教委科学基金

摘　　要：本文将既约梯度法与Ｈｕａｎｇ族变尺度法相结合，给出标准型线性约束规划问题的一类既约变尺度法．在较温和的假设下，算法具有全局收敛性和超线性收敛速度，最后指出本文算法包含和改进几个己有的有效算法．This paper prestents a class of reduced variable metric algorithm for programming problems with standard linear constraints with the help of reduced gradient and Huang's variable metric methods. Under some milder assumptions, the algorithm is proved to possess global and superlinear convergence. Finally, it is pointed out that our algorithm contains and improves some effective algorithms.

关 键 词：既约梯度法 变尺度法 线性规划 超线性收敛性 

分 类 号：O221.1[理学—运筹学与控制论]