Title: On concentration for (regularized) empirical risk minimization

Abstract: Rates of convergence for empirical risk minimizers have been well studied in the literature. In this paper we complement the rates by showing that after normalization, the empirical minimizer concentrates on a single point. Such results have first been established by \cite{chatterjee2014new} in the least squares context. We generalize this to other loss functions such minus log-likelihood for exponential families. We furthermore allow for a strictly convex regularization penalty. The results are based on concentration inequalities for maxima of empirical processes.