Block iterative algorithms for stochastic matrices

AbstractThe theory of regular splittings for singular M-matrices is used to derive the necessary and sufficient conditions for the convergence of iterative decomposition and aggregation techniques in the computation of the Perron-Frobenius eigenvector of a stochastic irreducible matrix. These conditions turn out to be less restrictive than those which must be satisfied by the block Jacobi or the block Gauss-Seidel method