Using a new class quasi-Newton equation for solving unconstrained optimization problems

Abstract

The quasi-Newton plays a basic role in constructing the Hessian approximation, and quasi-Newton algorithms are built based on it. In this paper, a new class quasi-Newton equation is presented. The proposed method always generates a new descent search direction. Under appropriate conditions, the new method is proved to possess global convergence. Comparative results show the computational efficiency of the proposed method.