改进的BFGS算法在一种新搜索下的收敛性

Global Convergence of a Class of Modified BFGS Algorithm with a New Line Search

李正峰提出了一类改进的BFGS算法,并在目标函数为一致凸的条件下,证明了带有Wolf线搜索的该算法具有全局收敛性。刘光辉提出了一种与回追搜索(backtracking linesearch)有关的可行线性搜索。证明在这种新搜索下,改进的BFGS算法亦县有全局收敛性。

A class of modified BFGS algorithm is proposed this paper, and the algorithm with Wolf＇s line search is globally convergent under the conditiont that the objective function is uniformly convex. A＇new line search technique is also shown in the paper ,and the global convergence of the modified BFGS algorithm with the new line search is proved in this paper.