Modified iterative methods for consistent linear systems (Q808156)

Given an M-matrix A, certain elementary row operations are performed before applying the Gauss-Seidel or Jacobi iteration in order to solve \(Ax=b.\) The essential idea is to eliminate the entries in the upper part of the matrix next to the diagonal. In some examples the convergence rate is substantially reduced. \{Reviewer's remark: There may be a connection with the incomplete LU- decomposition which is known to lead often to better convergence.\}