An efficient hybrid conjugate gradient method with sufficient descent property for unconstrained optimization. (3rd September 2022)

• Record Type: Journal Article
• Authors: Lotfi, M.; Hosseini, S. Mohammad
• Is Part Of: Optimization methods and software. Volume 37:Number 5(2022)
• Page Start: 1725
• Page End: 1739
• Publication Date: 2022-09-03
• Subjects: Conjugate gradient method -- sufficient descent -- unconstrained optimization -- global convergence; 65K05 -- 90C30; Mathematical optimization -- Periodicals; Algorithms -- Periodicals; 519.7
• DOI: 10.1080/10556788.2021.1977808
• Languages: English
• ISSNs: 1055-6788

Abstract

Abstract : In order to take advantage of the strong theoretical properties of the FR method and computational efficiency of the P R P + method, we present a new hybrid conjugate gradient method based on the convex combination of these methods. In our method, the search directions satisfy the sufficient descent condition independent of any line search. Under some standard assumptions, we established global convergence property of our proposed method for general functions. Numerical comparisons on some test problems from the CUTEst library illustrate the efficiency and robustness of our proposed method in practice.