A remark on "Robust machine learning by median-of-means"

TL;DR

This paper critically examines recent claims about robust machine learning using median-of-means, revealing that these results are largely straightforward extensions of existing tournament procedure methods.

Contribution

It clarifies that the recent advances in median-of-means robust learning are essentially natural consequences of established tournament procedure frameworks.

Findings

Recent median-of-means results follow from existing tournament machinery

The paper provides a clarification and perspective on the novelty of these results

It emphasizes the foundational nature of the underlying methods

Abstract

We explore the recent results announced in "Robust machine learning by median-of-means: theory and practice" by G. Lecu\'e and M. Lerasle. We show that these results are, in fact, almost obvious outcomes of the machinery developed in [4] for the study of tournament procedures.