Home

I am Maître de Conférences in Computer Science at Université Paris Sud since September 2018, at Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), within the [A&O or TAU](https://team.inria.fr/tau2/) team (Apprentissage & Optimisation, but overlapping Inria team TAU).

I develop new Deep Learning methods, with various application frameworks:

• physics of structural glasses, in collaboration with the Simons collaboration "cracking the glass problem", particularly in collaboration with Andrea J. Liu (UPenn, Pennsylvania), David Reichman (Columbia, New York), Giulio Biroli (ENS Paris, France), and Olivier Dauchot (ESPCI, France). In this context I use the recently introduced group-equivariant Graph-Neural Networks, in particular, I use equivariance with respect to rotations. Completed suprvision of PhD student F.S.Pezzicoli

• Dynamics of earthquakes or other scale-invariant sytems. In this context we want to develop Scale-Equivariant networks, most likely using wavelet-based methods (see recent works from Mallat et.al.) Ongoing PhD with A.Alvez

• Agro-ecological problems. For now, very simple applications of Computer Vision (identifying grains or identifying insects on pictures). I also want to consider more challenging problems in the future, such as recommender systems for selecting optimal mixes in the context of mixed cropping (the yield depends on the genetics, the soil, the meteorological time series and how they all interact).

A new topic I started recently (oct 2023) is:

• Understanding the learning dynamics, using the tools of statistical physics. This is collaboration with Valentina Ros and Marco Baity-Jesi. Ongoing PhD with F.S.Pezzicoli

About me:

• About my past research, see my Google Scholar
• My CV.
• About ongoing works and perspectives, see also the internships and PhD I propose.

Contact: [\prenom.\nom]@universite-paris-saclay.fr (vous n'êtes pas un bot?) Tél: 01 6915 7643
Laboratoire Interdisciplinaire des Sciences du Numérique

Adresse de travail: OpenStreetMap
À l'ancienne:
arrêt "Moulon" des lignes de bus 9 et 91.06, 91.10. Bât.660 (Digitéo)
rue René Thom / rue Noetzlin
91190 Gif-sur-Yvette

LISN, adresse postale officielle:
Bâtiment 508
rue John Von Neumann
91400 Orsay

Laboratoire Interdisciplinaire des Sciences du Numérique (LISN)
Départements AAC et SD, équipe A&O (notre équipe est à cheval sur 2 départements)
UMR CNRS 9015
CNU section 27