A modified descent Polak-Ribiere-Polyak conjugate gradient method with global convergence property for nonconvex functions

Following the modification scheme of Dong et al. made on the Hestenes-Stiefel method, we suggest a modified Polak-Ribiere-Polyak technique which satisfies the sufficient descent condition. We show that the method is globally convergent with the Wolfe line search conditions as well as the backtracking Armijo-type line search strategy proposed by Grippo and Lucidi, without convexity assumption on the objective function. Numerical experiments on some test functions of the CUTEr collection show that the method performs promisingly.