Abstract
Motivated by various modification approaches applied to the well-known Fletcher–Reeves (FR) algorithm—one of the pioneering classical techniques for large-scale optimization—this study proposes a revised Conjugate Gradient (CG) parameter that satisfies the sufficient descent condition under Wolfe line search criteria. Specifically, the proposed CG parameter can be regarded as a hybridization of several recent modifications of the FR formula, designed to benefit from the Polak–Ribière–Polyak (PRP) parameter, which is not only computationally efficient but also structurally close to the FR parameter. The proposed algorithm possesses sufficient descent property, which is widely recognized as a key condition for ensuring global convergence as well. To evaluate the effectiveness of the suggested parameter, computational experiments are carried out on a set of CUTEr test functions, as well as on selected noisy image restoration cases. The numerical results demonstrate the efficiency of the algorithm, particularly in terms of running time.