改进的拟牛顿非凸族算法的收敛性

The Convergence of Quasi-Newton-Non-Convex Class Methods for Unconstrained Optimization

对于无约束最优化问题min f(x),x∈Rn,提出了一类改进的拟牛顿非凸族算法,并结合一类更广泛的线性搜索,在目标函数凸的条件下,证明了算法的全局收敛性。

Algorithms which are update Quasi-Newton-non-convex methods for unconstrained optimization,minf（x）,x∈Rn,are proposed.The proof of the global and super linearly convergence of the generalized-quasi-Newton methods and the three main theorems are given.