Distribution estimation under the infinity norm (Q6887384)

This interesting paper studies distribution estimation under the \(l_{\infty}\) norm. To set the scene, denote by \(p\), a discrete probability distribution over the natural numbers and let \(X^n\) be a sample of \(n\) independent observations from the distribution \(p\). The authors are interested in estimating the distribution \(p\) from the sample \(X^n\), more specifically obtaining bounds under the \(l_{\infty}\) norm. They obtain novel bounds using a selection of techniques such as Chernoff-type inequalities and empirical Bernstein bounds. The authors illustrate their results on synthetic data and with real-world experiments. They also apply their results to a selective inference problem, where they estimate the most frequent probabilities in a given sample.\N\NThe paper is well written and has a good reference list.