Modified Gauss-Seidel type methods and Jacobi type methods for Z-matrices (Q1587902)

The authors discuss the iterative solution of linear systems by modified Gauss-Seidel (MGS) type methods and modified Jacobi (MJ) type methods. The convergence of these algorithms is analyzed and many properties for a matrix splitting of the coefficient matrix of the system are presented. Some recent results are improved. The main tool for a comparison of the different Gauss-Seidel type methods used by the authors is the spectral radius of the iteration matrix of the method. The authors prove that if the coefficient matrix of the system is a non-singular M-matrix, the MGS method converges for all parameters in \([0,1]\) and the convergence rates are better than those of the corresponding Gauss-Seidel type methods. No numerical experiments.