Matrix analysis
Index splitting for the Drazin inverse and the singular linear system
Applied Mathematics and Computation
Convergence properties of Krylov subspace methods for singular linear systems with arbitrary index
Journal of Computational and Applied Mathematics
Hi-index | 0.48 |
There are several methods for solving linear fixed point problem x = Tx + c, where T ∈ CN,N is a square matrix and I - T is possibly singular. Such problems arise if one splits the coefficient matrix of a singular system Ax = b of algebraic equations according to A= M - N (M nonsingular) which leads to x = M-1Nx + M-1b =: Tx +c. The basic iteration x0 ∈ CN, xm = Txm-1 + c (m ≥ 1) requires the modulus of every eigenvalue of the iteration matrix T except 1 is less than 1 and q = index(I-T), the index of I - T is equal to 1 for convergence. In this paper, we try to use the incomplete semiiterative methods (ISIM) to solve x = Tx + c when c ∈ R(I - T)q. Usually the special semiiterative methods are convergent even when the spectral radius of the iteration matrix is greater than 1 and q ≥ 1. Then the use of the ISIM in the Markov chain modeling is considered. Finally, numerical examples are reported.