On the use of incomplete semiiterative methods for singular systems and applications in Markov chain modeling. (Q1855124)

The authors describe a semiiterative algorithm for solving singular systems of linear equations. The method is based on the classical definition proposed by \textit{R. S. Varga} in his well known book on iterative methods [Matrix iterative analysis (Prentice-Hall Series in Automatic Computation, Prentice-Hall, Englewood Cliffs, New Jersey) (1963; Zbl 0133.08602)], but uses a special definition of the polynomial sequences used in the construction of approximations. The authors prove convergence of their algorithm even when the spectral radius of the iteration matrix is greather than 1. This gives them the possibility to apply the method for computing the stationary probability vector of the Markov chain.