Armijo搜索下的记忆梯度法及其收敛性  

Convergence of New Memory Gradient Method with Armijo Search

作　　者：张祖华[1] 时贞军[1] 

机构地区：[1]曲阜师范大学运筹与管理学院,山东日照276826

出　　处：《运筹与管理》2007年第1期24-27,共4页Operations Research and Management Science

摘　　要：本文提出一种新的无约束优化记忆梯度算法,在Armijo搜索下,该算法在每步迭代时利用了前面迭代点的信息,增加了参数选择的自由度,适于求解大规模无约束优化问题。分析了算法的全局收敛性。In this paper we present a new memory gradient method for solving unconstrained optimization problems. We use the previous iterative information to generate a new iterate at each iteration with Armijo search rule and add the freedom of some parameters. Therefore it is suitable to solve large scale optimization problems. We analyze the global comvergence under some mild conditions.

关 键 词：无约束优化 记忆梯度法 ARMIJO搜索 全局收敛性 

分 类 号：O221.2[理学—运筹学与控制论]