Some hybrid Riemannian conjugate gradient methods with restart strategy

Hybrid nonlinear conjugate gradient (CG) methods are known to be efficient and have less memory requirement for solving optimization problems in Euclidean spaces . The formulations strategically switch between two or more CG parameters to enhance their general structures. In this work, some hybrid Riemannian CG methods are presented, which are sufficiently descent irrespective of the line search technique employed and converges globally using scaled version of the nonexpensive condition. Finally, the numerical experiments are performed to confirm the strength of the schemes by considering some Riemannian optimization problems. The numerical results obtained when compared to some classical Riemannian CG methods indicated the efficiency of the proposed hybrid Riemannian CG methods.