On HSS-Based Iteration Methods for Two Classes of Tensor Equations

Abstract

HSS-based iteration methods for large systems of tensor equations $\mathcal{T}$($x$) = $b$ and $Ax$ = $\mathcal{T}$($x$) + $b$ are considered and conditions of their local convergence are presented. Numerical experiments show that for the equations $\mathcal{T}$($x$) = $b$, the Newton-HSS method outperforms the Newton-GMRES method. For nonlinear convection-diffusion equations the methods based on HSS iterations are generally more efficient and robust than the Newton-GMRES method.