Iterative algorithms for large stochastic matrices (Q808164)

Among the best-known iterative methods are the power method, the Jacobi method, and the Gauss-Seidel method. None of these methods can be proved to present, a priori, better performance than the others. This paper aims at providing a practical criterion which allows one to compare the convergence rates of these methods. This measure is an approximation of the modules of the second largest eigenvalue of the iteration matrix and can be determined a priori.