关于DFP算法的全局收敛性

On the Global Convergence of DFP Method

本文讨论求解无约束最优化问题的DFP算法的全局收敛性问题。设步长满足Armijo非精确搜索条件,证明了对严格凸二次函数最小值问题,DFP算法具有全局收敛性,并且收敛速度为超线性。

This paper deals with the global convergence problem of DFP method for unconstrained optimization problems. For the case of Armijo line searches,we prove the global convergence property of DFP method used for the minimization problem of strickly quadratically convex function. The superlinear convergence property is also obtained.