On a two-level multigrid solution method for finite Markov chains (Q1894507): Difference between revisions

The author presents a two-level algebraic multigrid scheme for computing the stationary distribution of a homogeneous Markov chain with a finite state space. Both, continuous-time and discrete-time Markov chains are discussed. The author reveals his technique to be equivalent to an iterative aggregation-disaggregation technique. Error analysis results are presented, too. The author refines the error analysis of \textit{F. Chatelin} and \textit{W. L. Miranker} [ibid. 43, 17-47 (1982; Zbl 0485.65023)] and \textit{M. Haviv} [SIAM J. Numer. Anal. 24, 952- 966 (1987; Zbl 0637.65147)]. An interpretation of the block underrelaxation smoother in the stochastic context is developed, too.