A comparison theorem for the iterative method with the preconditioner \((I+S_{max})\) (Q1612412)

The authors propose the preconditioner \(P_m=I+S_{\max}\) where \(S_{\max}\) is contructed by using only the largest element of each row of the upper triangular part of the nonsingular diagonally dominant \(\mathbb Z\)-matrix \(A\), that is, \(S_{\max}=(s_{ij}^m)=-a_{ik_i}\) for \(i=1,2,\ldots,n-1\), \(j>i\), and \(0\) otherwise, where \(k_i=\min j\in\{j\mid \max_j|a_{ij}|\}\) for \(i<n\). Without loss of generality, it is assumed that \(A=I-L-U\), where \(I\) is the identity matrix, and \(L\) and \(U\) are strictly lower and upper triangular matrices, respectively. Under the additional assumption that \(a_{ii+1}a_{i+1j}\leq a_{ik_i}a_{k_ij}\), \(1\leq i\leq n-1\), \(j\leq i\), it is shown that \(P_m\) is preferable to the preconditioner \(P_s=I+S\) where \(S=(s_{ij}^m) = -a_{ij}\) for \(j=i+1\), \(i=1,2,\ldots,n-1\), and 0 otherwise. This \(P_s\) was proposed by \textit{A. D. Gunawardena, S. K. Jain} and \textit{L. Snyder} [Linear Algebra Appl. 154-156, 123-143 (1991; Zbl 0731.65016)].