Abstract : Last year, a major flood in the Yvette river basin caused significant damage in the area around Paris-Saclay University. Such rare, high-impact events lie in the tails of probability distributions and may not be adequately described by standard statistical methods. Extreme value analysis (EVA) provides a dedicated framework to model these rare events.

In this seminar, we introduce the theoretical background behind Extreme Value Theory, and discuss how it can be leveraged to derive machine learning algorithms tailored to the analysis of extreme events.