Entropy and large deviations for discrete-time Markov chains (Q1812373)

The authors propose a new method to prove a weak Sanov theorem for a homogeneous discrete-time Markov chain with denumerable state space, which is not necessarily ergodic or irreducible. In fact, they show how a careful analysis of the entropy function combined with change of measures leads to a new proof of a weak large deviation principle for the law of the pair empirical measure.