求解一类二阶段有补偿问题的对偶梯度法  

A Dual Gradient Method for Solving a Class of Two-stage Compensating Problems

作　　者：陈志平[1] 徐成贤[1] 

机构地区：[1]西安交通大学数学系

出　　处：《应用数学》1996年第3期266-271,共6页Mathematica Applicata

摘　　要：利用对偶理论，本文给出了求解一类具有简单补偿的非线性二阶段问题的新对偶梯度法．在假设目标函数为可分连续可微凸函数的条件下，在每一选代步可将原二阶段有补偿问题转化为几个一维凸规划问题，大大简化了问题的求解．所给算法简单易行，文中还证明了该算法的全局收敛性．Using the duality theory,a new dual gradient,method is presented to solve a class ofnonlinear two-stage problems with simple recourse. Under the assumption that the objective function is separable,convex and continuously differentiable,the original two-stage compensating problem can be turned into several one-dimensional convex programmings at each step,making it easyto solve. The suggested algorithm is simple in implementation and global convergence property is analysed.

关 键 词：可分性 对偶理论 补偿问题 对偶梯度法 凸规划 

分 类 号：O221.5[理学—运筹学与控制论]