A Hybrid Riemannian Conjugate Gradient Method for Nonconvex Optimization Problems

Tang Chunming,Rong Xianglin,Jian Jinbao,Xing Shajie

DOI: https://doi.org/10.1007/s12190-022-01772-5

2022-01-01

Abstract:In this paper, we develop and analyze a hybrid Riemannian conjugate gradient (CG) method to the problem of minimizing a smooth objective function on a Riemannian manifold, based on a Riemannian generalization of the hybrid formula created by Jian et al. (Appl. Math. Model. 39: 1281-1290, 2015) in Euclidean space. This method can produce descent directions independent of any stepsize strategy on Riemannian manifolds. Moreover, under the Riemannian Wolfe conditions, the proposed method is proved to be globally convergent without assuming the objective function to be convex. Finally, the new method is compared numerically with the existing methods for solving some large-scale optimization problems on different manifolds. Numerical experiments and their performance profiles are reported, which show that the proposed methods are efficient.