广义Wolfe线搜索下一类修正的Fletcher-Reeves方法的收敛性  

Convergence of a Class of Modified Fletcher-Reeves Methods Under the General Wolfe Line Search

作　　者：刘金魁[1] 

机构地区：[1]重庆三峡学院数学与统计学院,404100

出　　处：《应用数学学报》2013年第6期1109-1117,共9页Acta Mathematicae Applicatae Sinica

基　　金：重庆市教委(KJ121112)资助项目

摘　　要：本文在Fletcher-Reeves共轭梯度法的基础上提出一类修正的Fletcher-Reeves共轭梯度法,其在广义Wolfe线搜索下具有充分下降性和全局收敛性.大量数值试验表明,新方法对于给定的测试函数在数值效果上明显优于Fletcher-Reeves共轭梯度法.In this paper, a class of modified Fletcher-Reeves conjugate gradient methods is proposed on the basic of Fletcher-Reeves conjugate gradient method, which has the sufficient descent property and global convergence under the general Wolfe line search. Numerical experiments show that the new methods substantially outperform Fletcher-Reeves method for the given test problems.

关 键 词：无约束最优化问题 共轭梯度法 充分下降性 广义Wolfe线搜索 全局收敛性 

分 类 号：O224[理学—运筹学与控制论]