1: \begin{abstract}

2: We present novel bounds for

3: estimating discrete probability distributions under the

4: $\ell_\infty$ norm.

5: These are nearly optimal in various precise senses,

6: including a kind of instance-optimality.

7: Our data-dependent convergence guarantees for the maximum likelihood estimator

8: significantly improve upon 

9: the

10: currently known results. 

11: A variety of techniques are utilized

12: and innovated upon, including

13: Chernoff-type inequalities and empirical Bernstein bounds. 

14: We illustrate our results in synthetic and real-world experiments. Finally, we apply our proposed framework to a basic selective inference problem, where we estimate the most frequent probabilities in a sample. 

15: \end{abstract}

16: