\(LU\)-factorization versus Wiener-Hopf factorization for Markov chains (Q383600)

Consider Markov chains with a denumerable state space. The generator of such a chain is an infinite matrix (typically indexed by integers, however, other options are included as well). The author opens an interesting approach to study such matrices via a particular form of the LU-factorization. It is shown that, up to a Fourier transform, the approach is a generalization of the Wiener-Hopf factorization method for random walks. A probabilistic interpretation of the results is offered and several examples are analyzed. In particular, the method is applied toward the computation of the invariant measure for a positive recurrent Markov chain.