We consider denumerable stochastic processes with (or without) memory. Their evolution is encoded by a finite or infinite rooted tree. The main goal is to compare the entropy rates of a given base process and a second one, to be considered as a perturbation of the former. The processes are described by probability measures on the boundary of the given tree and by corresponding forward transition probabilities at the inner nodes. The comparison is in terms of Kullback-Leibler divergence. We elaborate and extend ideas and results of Böcherer and Amjad. Our extensions involve length functions on the edges of the tree as well as nodes with countably many successors. In particular, in Section V, we consider trees with infinite nonbacktracking paths and random perturbations of a given process.
Fields of Expertise
- Information, Communication & Computing