Journée Laboratoire Hubert Curien-Institut Camille Jordan

Liste des Contributions

7. Ouverture de la journée

12/11/2026 09:30

Allocution des directeurs d'unité

3. Approximation du flot de Willmore par des opérateurs neuronaux

Elie Bretin (INSA de LYON, INstitut Camille Jordan)

12/11/2026 10:00

La méthode de champ de phase constitue un outil numérique particulièrement efficace pour l’approximation des flots géométriques d’interfaces orientées, tels que le mouvement par courbure moyenne ou le flot de Willmore.

Certaines applications, comme la reconstruction d’interfaces à partir de nuages de points, nécessitent toutefois une représentation non orientée des interfaces, pour laquelle...

1. Basics on generalization bounds with a focus on the PAC-Bayesian theory

Emilie Morvant (Laboratoire Hubert Curien, Université Jean Monnet)

12/11/2026 10:35

2. Équations Intégrales et méthodes Physics-Informed Neural Networks (PINNS)

Laurence Grammont (Institut Camille Jordan, Université Jean Monnet)

12/11/2026 11:30

8. Exposés court 1

Filippo Santambrogio (Institut Camille Jordan, UCBL)

12/11/2026 12:05

9. Exposés court 2

Benjamin Girault (Laboratoire Hubert Curien, Université Jean Monnet)

12/11/2026 12:20

10. Exposé court 3

12/11/2026 12:35

11. Exposé court 4

12/11/2026 12:50

4. Formulation of the many-particle scattering problem based on individual particle's quasi-normal modes

Jean-Philippe Banon (Laboratoire Hubert Curien, Université Jean Monnet)

12/11/2026 14:45

Quasi-normal modes, also known as resonant states, are modes of the electromagnetic field satisfying the source free harmonic Maxwell's equations. They generally appear for complex-valued frequencies and extend the notion of eigenmodes from Hermitian to non-Hermitian systems. The real part of the frequency represents the oscillation frequency of the time harmonic field while the imaginary part...

5. Exposé 5

Brice Colombier (Laboratoire Hubert Curien, Université Jean Monnet), Michael Bulois (Institut Camille Jordan (UJM-Saint-Etienne)), Pierre-Louis Cayrel (Laboratoire Hubert Curien, Université Jean Monnet)

12/11/2026 15:20

6. From proximal alternating minimization techniques to composite proximal gradient

Clara Lage (Institut Camille Jordan, École Centrale Lyon)

12/11/2026 15:55

The discrete Mumford–Shah formulation leads to a challenging nonconvex optimization problem involving the joint minimization of two variables, typically addressed through alternating minimization methods such as PALM and its variants.

In this talk, I will present a connection between the discrete Mumford–Shah formulation and a class of composite minimization problems for which...

15. Constraint-Augmented Physics-Informed Neural Networks for Parameter Calibration of Reaction–Diffusion Equations

Jayrah Bena Riñon (Institut Camille Jordan, Université Jean Monnet)

Physics-informed neural networks (PINNs) have emerged as a powerful framework for solving inverse problems governed by partial differential equations. In this work, we study the calibration of reaction–diffusion equations from observational data using PINNs. We show that standard PINN formulations are generally insufficient to recover the complete set of model parameters and that additional...

13. Poster

Ehud Mossa Ockegna

12. Poster: A Kernel Approach to Multivariate Generalized Gamma Convolutions

Léo Gonin

We propose a kernel-based estimation framework for multivariate generalized gamma convolutions (MGGC), a class of probability distributions on Rd +. MGGC are useful, for example, in risk modeling, but their densities are generally not explicit. The construction relies on a new Stein operator for the MGGC class, derived from a partial differential equation satisfied by the characteristic...

14. Poster: Activation-Matched Gibbs Initialization for Neural Networks

Diego Pinto-Suarez

Initialization plays an important role in neural network training. A classical way to address this task relies on Gaussian weights and preactivations, motivated in part by their analytical tractability and their emergence in large-width approximations. We propose an activation-induced initialization based on a statistical-mechanical viewpoint: for activations that arise as the gradient of a...