线性约束LC^1凸优化问题的内点信赖域算法  

An Interior Trust Region Method for Linear Constrained LC^1 Convex Optimization Problems

作　　者：莫降涛[1] 张可村[1] 

机构地区：[1]西安交通大学理学院

出　　处：《工程数学学报》2006年第6期1009-1016,共8页Chinese Journal of Engineering Mathematics

摘　　要：本文提出一种解线性约束凸规划的数值方法。通过将问题的KKT系统转化成一个约束方程,算法在每步迭代只需解一个线性方程组即可得到搜索方向。算法运用了信赖域方法利内点技术。在较弱的条件下,我们证明了算法的全局收敛性。This paper presents an interior trust region method for linear constrained LC^1 convex optimization problems. By reformulating the KKT system as a constrained equation, the algorithm generates the search direction by solving a linear equation at each iteration. The algorithm combines trust region method and interior technique. Under mild conditions, we show that the algorithm is globally convergent.

关 键 词：信赖域 KKT系统 下降方向 全局收敛性 

分 类 号：O221.1[理学—运筹学与控制论]