Performance Analysis of a Modified Conjugate Gradient Algorithm for Optimization Models

Abstract

The Conjugate gradient (CG) algorithms is very important and widely used in solving optimization models. This is due to its simplicity as well as global convergence properties. Various line search procedures as usually employ in the analysis of the CG methods. Recently, many studies have been done aimed at improving the CG method. In this paper, an alternative formula for conjugate gradient coefficient has been proposed which possesses the global convergence properties under exact minimization condition. The result of the numerical computation has shown that this new coefficient performs better than the existing CG methods.