一般约束最优化超线性与二次收敛的序列线性方程组算法  

Algorithm of Sequential Systems of Linear Equations with Superlinear and Quadratical Convergence for General Constrained Optimization

作　　者：简金宝[1] 朱志斌[2] 

机构地区：[1]广西大学理学院,南宁530004 [2]桂林电子工业学院计算科学与应用物理系,桂林541004

出　　处：《工程数学学报》2003年第3期24-30,共7页Chinese Journal of Engineering Mathematics

基　　金：国家自然科学基金(10261001);广西自然科学基金(0236001;0249003);广西"十百千人才工程"资金(99214)联合资助.

摘　　要：讨论了一般等式和不等式约束优化问题,利用序列线性方程组技术和广义投影技巧,建立问题的一个"可行下降"算法,每次迭代只需解一个线性方程组和计算一次广义投影。在适当条件下,证明算法超线性和二次收敛于原问题的K T点。Optimization problems with general equality and inequality constraints are discussed. By using the techuique of sequential systems of linear equations and generalized projection, a feasible descent algorithm is presented, at each iteration of the algorithm, only one linear system and a generalized projection need to be solved and computed. Under suitable assumptions, the algorithm is proved to converge superlinearly and quadratically to a KT point of the problem. 

关 键 词：一般约束最优化 序列线性方程组 算法 超线性收敛 二次收敛 

分 类 号：O221.1[理学—运筹学与控制论]