This talk has been automatically
captioned. You can remove these by pressing CC on the video toolbar.
Name: Matthieu Lerasle
Talk Title: Robust statistical learning
Abstract: I will present two tools that allow to obtain strong deviation properties of estimators when data are both heavy tailed and possibly contaminated. The first tool originally developed by Lugosi and Mendelson leads to a deviation inequality for median of means processes. The second tool is an homogeneity lemma that simplifies the more classical peeling argument when, for example, the loss is convex. I will illustrate the extent of these tools in various statistical learning problems.
This
talk is an invited talk at EVA 2021.