Search:
Match:
1 results

Analysis

This paper investigates the application of the Factorized Sparse Approximate Inverse (FSAI) preconditioner to singular irreducible M-matrices, which are common in Markov chain modeling and graph Laplacian problems. The authors identify restrictions on the nonzero pattern necessary for stable FSAI construction and demonstrate that the resulting preconditioner preserves key properties of the original system, such as non-negativity and the M-matrix structure. This is significant because it provides a method for efficiently solving linear systems arising from these types of matrices, which are often large and sparse, by improving the convergence rate of iterative solvers.
Reference

The lower triangular matrix $L_G$ and the upper triangular matrix $U_G$, generated by FSAI, are non-singular and non-negative. The diagonal entries of $L_GAU_G$ are positive and $L_GAU_G$, the preconditioned matrix, is a singular M-matrix.