Weighted versions of Gl-FOM and Gl-GMRES for solving general coupled linear matrix equations (Q904386)

The paper deals with solving the general coupled linear matrix equations \[ \sum_{j=1}^p A_{ij}X_jB_{ij}=C_i, \;\;\;i=1,\dots,p, \] where \(A_{ij}\in \mathbb R^{m\times m}\), \(B_{ij}\in \mathbb R^{n\times n}\), and \(C_{i}\in \mathbb R^{m\times n}\), are given matrices and \(X_{j}\in \mathbb R^{m\times n}\) are unknowns, \(i,j=1,\dots,p\). Two algorithms based on the Krylov subspace method are proposed: the WGl-FOM (weighted global full orthogonalization method) and the WGl-GMRES (weighted global generalized minimal residual) method. These algorithms are based on a new inner product and the corresponding matrix product that is used in the weighted global Arnoldi process. Some links between weighted and non-weighted algorithms are established. The numerical experiments show on two simple examples that the weighted variants of the algorithms are more efficient. The paper is intelligible although it contains typo mistakes (already in the formulation of the problem). The weighted versions of the algorithms are natural generalizations of the non-weighted ones and the main idea is clearly explained in the paper. On the other hand, the numerical experiments are not convincing: the solved problems are small and the computational times are relatively large.