Generalized divergences from generalized entropies

Abstract

Several quantifiers of information, also known as entropies, have been introduced in different contexts and from different motivations. For almost each one of these entropies, a measure of the loss (or gain) of information has been introduced. In this work we introduce generalized weighted divergences associated with an arbitrary entropy. The resulting measures are closely related to Bregman divergences. We study the main formal properties of the resulting divergences, we extend them to weighted probability distributions and we apply some of them to the analysis of simulated and real time series.