Information, Apprentissage et Numérique
PRESENTATIOn of the team
The research of team IAN, for Information, leArning, and computatioN (team 3 of the LPENSL), combines two fields: on one hand, it develops research and methodological development in information processing (from signal and image processing to machine learning), and on the other hand, it covers the activities about the physics of information -- and in the sense of information in (statistical) physics and the role of information sciences in physics. The team, affiliated both to CNRS--Physics and CNRS--Informatics, aims at contributing to research at the highest level, both in the physics of information and in information processing.
The research aligns with three major challenges identified by the CNRS: (1) the impact and use of Artificial Intelligence in all its dimensions, particularly in signal and image processing, as well as in physics -- including computational physics. (2) the New challenges for numerical methods in physics. (3) The various aspects of the Physics of Complex Systems, including information physics, stochastic thermodynamics, and modeling or data analysis of social systems, geographic data, genomics, and human health.
NEWS and HIGHLIGHTS
[09/2026] Workshop on Learning-based Methods for Inverse problem at ENSL, Sept. 14-15: Information, Registration & Schedule. The workshop is organize at the occasion of the HDR defense of Julian Tachella. Congratulations to him as he now defended successfully.
[08/2026] Meet many members of the team in Bruges at EUSIPCO 2026. See you there in front of the posters or for the orals.
[07/2026] Two new ANR projects are developped in the team: the ANR PRME VAST "Velocity Augmented Stochastic Thermodynamics" (PI : Ludovic Bellon) and the ANR PRC DeepScale "Multiscale Deep Model-based Neural Networks for Image Restoration" (PI : Nelly Pustelnik).
[05/2026] We are pleased to welcome Giuseppe Petrillo who arrived in the IAN team of the LPENSL on the Chaire de Professeur Junior CNRS about the Dynamique des Décisions Collectives (DyDéCo).
[03/2026] Congratulations to Jérémy Scanvic for his Honourable Mention for his PhD work in the team, in the Arqus Prize on Open Science rewards, for his work in the DeepInverse library. More information on Jérémy's post on linkedin.com
[03/2026] The DeepInverse libray will be supporter by the P16 program of the Programme IA of the Agence de Programme du Numérique : https://p16.inria.fr/
[02/2026] The team has three new papers acccepted at ICLR 2026 (International Conference on Learning Representations) that will be presented in April: (1) Victor Sechaud, Jérémy Scanvic, Quentin Barthelemy, Patrice Abry et Julian Tachella, Equivariant Splitting: Self-supervised learning from incomplete data, (2) Andrea Combette, Antoine Venaille, Nelly Pustelnik, A new initialisation to Control Gradients in Sinusoidal Neural network, and (3) Matthieu Terris, Samuel Hurault, Maxime Song et Julian Tachella, Reconstruct Anything Model: a lightweight foundational model for computational imaging. Congratulations to all the authors and meet them in April at ICLR !
[01/2026] The teams of the laboratory change their names. The "Signals, Systems and Physics'' (SISYPHE) team continues its activity mostly as the IAN : "Information, Apprentissage et Numérique" team, including more themes in Physics of Information, Computational, Physics and Model-based AI for Science. This new webpage is now online.
[Award 2024] Julián Tachella receives the Open Science prize for DeepInverse. More details here (in french).
[Award 2024] Congratulations to Patrice Abry, lauréat du Grand Prix Annuel IMT de l'Académie des Sciences for his studies on scaling phenomena in signal and image processing . See more details : https://www.ens-lyon.fr/PHYSIQUE/presentation/news/prix-academie-2024
[Conference EUSICO 2024 in Lyon] We hosted the EUSIPCO 2024 Conference in Lyon, 26-30 August 2024. EUSIPCO is the major European Signal Processing Conference [EUSIPCO 2024 Website] of EURASIP.