Probabilistic approach to computational algorithms for finding stationary distributions of Markov chains (Q1177197)

The aim of this paper is to study computational algorithms for finding stationary distributions of Markov chains using purely probabilistic arguments, based on the semi-regenerative structure of Markov chains. The reason is that this approach gives better insight into formal manipulations of equilibrium equations and provides probabilistic interpretation of the coefficients obtained at each such manipulation. Two theorems arising in connection with Gaussian elimination are derived using semi-regenerative analysis and then applied to an example connected with the birth-death process.