Concentration Inequalities for Dependent Random Variables on Bayesian Networks

Abstract

The thesis presents a theoretical study of the concentration results for the function defined on the random variables on a Bayesian Network. In this work, we provide several concentration inequality results under the assumption that the function is Lipshitz or bounded difference. In addition, we illustrate about the concentration of the maximum likelihood estimator of some learning models. We also show the optimality of certain results and the comparison to the results in other relevant literature.