Modified parameter of Dai Liao conjugacy condition of the conjugate gradient method

Abstract

The conjugate gradient (CG) method is widely used for solving nonlinear unconstrained optimization problems because it requires less memory to implement. In this paper, we propose a new parameter of the Dai Liao conjugacy condition of the CG method with the restart property, which depends on the Lipschitz constant and is related to the Hestenes Stiefel method. The proposed method satisfies the descent condition and global convergence properties for convex and non-convex functions. In the numerical experiment, we compare the new method with CG_Descent using more than 200 functions from the CUTEst library. The comparison results show that the new method outperforms CG Descent in terms of CPU time, number of iterations, number of gradient evaluations, and number of function evaluations.