Accurate solutions of \(M\)-matrix Sylvester equations (Q2428544)

The authors present a relative perturbation theory for an \(M\)-matrix Sylvester equation (MSE). Specifically, the MSE is meant by the matrix equation \(AX + XB = C\) where \(A\) and \(B\) have positive diagonal entries and nonpositive off-diagonal entries; \(P = I_m \otimes A + B^T \otimes I_n\) is a nonsingular \(M\)-matrix; and \(C\) is entry-wise nonnegative. By the authors, it has been proved that the small relative perturbations to the entries of \(A\), \(B\), and \(C\) introduce small relative errors to the entries of \(X\). This is unlike the existing perturbation theory on a (general) Sylvester equation. The authors propose some important modifications to the existing numerical methods for the computation of \(X\). Numerical examples are given to verify their claims as well. Reviewer's remark: A matrix can be considered as an linear ``operator'', and the main concern of this paper is the relative perturbation analysis related to the matrix equations.