NeurIPS 2020

Robust and Heavy-Tailed Mean Estimation Made Simple, via Regret Minimization


Meta Review

This paper makes a connection between outlier-robust mean estimation and mean estimation with subgaussian rates for heavy tailed distributions and show an equivalence between natural estimators for both these estimation tasks. As a consequence, one derives the algorithms for outlier-robust mean estimation can be used to solve the heavy-tailed mean estimation problem. The reviewers found this work interesting in its algorithmic contributions and insights on two natural tasks in robust estimation. I recommend acceptance of this paper to the NeurIPS program.