An overview of classical robust statistics and generalizations to the future

Po-Ling Loh (University of Cambridge) https://simons.berkeley.edu/talks/po-ling-loh-university-cambridge-2024-08-28 Modern Paradigms in Generalization Boot Camp Robust statistics aims to provide methods for reliable estimation and inference when data are generated from a distribution with some form of contamination. In this talk, we will provide a self-contained overview of some key concepts in classical robust statistics. We will then discuss a few recent results where classical concepts have proven useful in more modern settings, including heterogenous distributions, new forms of contamination, and private hypothesis testing.