Generative AI for Extreme events

Overview

This event will focus on addressing the opportunities and current challenges of using Generative models to simulate Extreme events.

The aim is to federate the Statistical Learning / Generative modelling community with the Extreme values statisticians and Climate / Financial Risk assessment researchers, to start a conversation about recent advances in each field and their interaction or lack thereof.

The presentations will be given by leading experts of both domains that have already a contribution in the intersection between both disciplines. Registration is free but mandatory.

Program

Morning – Theoretical / Fundamental Session

9:00 – 9:30

Welcome coffee and opening session

09:30 – 10:30

Juliette Legrand

Heavy tails and extreme events

10:30 – 11:30

Arthur Stéhanovitch

Theory of generative models (SGMs, etc.)

11:30 – 12:30

Stéphane Girard

GANs and Extreme Value theory

12:30 – 14:00

Lunch break

Afternoon – Applied Session & Actuarial Perspectives

14:00 – 15:00

Philippe Naveau

VAEs and Extreme Value theory: models for climate

15:00 – 16:00

Umut Şimşekli

SGMs and Extreme Value theory: generative models for extreme events

16:00 – 16:30

Coffee break

16:30 – 17:15

Jakiw Pidstrigach

Concrete application or real-world use case

17:15 – 18:00

Daniel Nkameni

Application of Extreme Value theory to a parametric insurance product

18:00 – 18:30

Closing

Venue

The venue of the workshop is located at Sorbonne Université, Campus Pierre et Marie Curie, 4 place de Jussieu, 75005 Paris. The sessions will take place in Barre 44–45 (1er étage, niveau 2), accessible via Tour 44, in the salles du CICSU.

Getting there

Metro: Jussieu (Lines 7 & 10)
Nearby Metro: Place Monge (Line 7), Cardinal Lemoine (Line 10)
RER: Gare d’Austerlitz (RER C) then ~15–20 min walk (or metro connection)
Buses: 63, 67, 86, 87, 89

GenAI4EV