不等式约束优化拟正则条件下的超线性收敛FSSLE算法  

A SUPERLINEARLY CONVERGENT FSSLE ALGORITHM FOR INEQUALITY CONSTRAINED OPTIMIZATION WITH QUASI-REGULARITY

作　　者：张新华[1] 李学全[2] 

机构地区：[1]南京农业大学工学院,江苏南京210031 [2]湖南第一师范学院,湖南长沙410083

出　　处：《经济数学》2009年第3期85-94,共10页Journal of Quantitative Economics

基　　金：江苏省农机局科研基金(GXZ06014);湖南省教育厅科学研究项目(05B070)

摘　　要：对不等式约束优化问题,提出一个可行序列线性方程组(FSSLE)算法.该算法每次迭代只需求解两个具有相同系数矩阵的线性方程组,因而计算量较小.在一定条件下,算法具有全局收敛性.在没有严格互补条件、比强二阶充分条件弱的拟正则条件下,证明了算法具有超线性收敛性并用数值试验表明其有效性.A feasible sequential system of linear equations （FSSLE） algorithm was proposed for solving inequality constrained optimization problems. It is observed that the algorithm is merely necessary to solve two systems of linear equations with the same coefficient matrix. Under some suitable conditions, we show that this algorithm is globally convergence. Without assuming the strict complementarity, the superlinear conver gence rate is obtained under the quasi-regularity, which is weaker than the strong second-order sufficiency condition. Numerical experiments show that the method is effective.

关 键 词：非线性规划 FSSLE 全局收敛 超线性收敛 

分 类 号：O221.2[理学—运筹学与控制论]