A Modification of Classical Conjugate Gradient Method using Strong Wolfe Line Search

摘要

Recently many researches try to develop and improve the Conjugate Gradient (CG) methods because of its convergence properties and low computation costing. In this paper, another CG coefficient (β_k) will be proposed which is categorized as modification in such a way to improve the performance of the classical CG methods. This paper is focused on generating β_k with several desirable properties: (1) generate descent search direction at each iterations; and (2) converge globally by using strong Wolfe line search. Numerical comparisons of three CG methods show the robustness and the efficiency of the new method in solving all given problems.