On contraction properties of Markov kernels (Q1400842)

The authors study general properties of contractions of Markov kernels without assumptions on the existence of an invariant probability measure. In their previous paper [in: Séminaire de Probabilités XXXIV. Lect. Notes Math. 1729, 1-145 (2000; Zbl 0963.60040)] they have shown that the system is forgetting its initialization without nevertheless converging in large times to an invariant distribution. This was the motivation for the present paper. The Lipschitz contraction properties of general Markov kernels are studied seen as operators on spaces of probability measures equipped with entropy-like ``distances''. Strong contraction properties in Orlicz spaces for relative densities are proved under more restrictive mixing assumptions. Contraction bounds in the entropy sense around arbitrary probability measures are described by introducing a suitable Dirichlet form and the corresponding modified logarithmic Sobolev constants. The example of inhomogeneous Gaussian chains illustrates the interest in these bounds.