AIAE'23 - Sciencesconf.org

8:30 - 9:00 (30min)

Welcome coffee

9:00 - 9:10 (10min)

Opening words

Yves Balkanski, Institut Pascal Director

9:10 - 9:30 (20min)

Opening session

Marc Schoenauer, Scientific Coordinator, IA2 program – IRT SystemX, Research Director Inria Saclay – Ile-de-France

9:30 - 10:30 (1h)

Keynote - Nils Thuerey, Technical University of Munich (TUM) - Probabilistic Simulations: Diffusion Models & Differentiable Solvers

Nils Thuerey, Technical University of Munich (TUM)

10:30 - 11:00 (30min)

Coffee break and posters

11:00 - 11:25 (25min)

Contributed talks (Session 1) - A Meta-Generation framework for Industrial System Generation

Fouad Oubari, Raphael Meunier, Rodrigue Décatoire, Mathilde Mougeot (ENS-Paris Saclay, Michelin) - A Meta-Generation framework for Industrial System Generation

11:25 - 11:50 (25min)

Contributed talks (Session 1) - Conditional and Unconditonal Generation of Seismic Signals using Diffusion Model .

Hugo Gabrielidis, Filippo Gatti, Stéphane Vialle (LMPS, LISN) - Conditional and Unconditonal Generation of Seismic Signals using Diffusion Model .

11:50 - 12:15 (25min)

Contributed talks (Session 1) - Speeding HPC simulation with AI : An Introduction to Physics-informed Machine Learning with NVIDIA Modulus

Francois Courteille (NVIDIA) - Speeding HPC simulation with AI : An Introduction to Physics-informed Machine Learning with NVIDIA Modulus

12:15 - 13:45 (1h30)

Lunch break and posters

13:45 - 14:10 (25min)

Contributed talks (Session 2) - Approximately well-balanced Discontinuous Galerkin methods using bases enriched with Physics-Informed Neural Networks

Victor Michel-Dansac, Laurent Navoret, Emmanuel Franck (TONUS-Inria Nancy, IRMA) - Approximately well-balanced Discontinuous Galerkin methods using bases enriched with Physics-Informed Neural Networks

14:10 - 14:35 (25min)

Contributed talks (Session 2) - A self-supervised learning framework to identify the permeability field from images based on physics-informed and convolutional neural networks

John Hanna, José V. Aguado, Sebastien Comas-Cardona, Yves Le Guennec (Ecole Centrale de Nantes, IRT Jules Vernes) - A self-supervised learning framework to identify the permeability field from images based on physics-informed and convolutional neural networks

14:35 - 15:00 (25min)

Contributed talks (Session 2) - GNN-based Schwarz Preconditioner for Large Scale Problem Solving

Matthieu Nastorg, Marc Schoenauer, Guillaume Charpiat, Thibault Faney, Jean-Marc Gratien, Michele Alessandro Bucci (Inria, IFPEN, Safran Tech) - GNN-based Schwarz Preconditioner for Large Scale Problem Solving

15:00 - 15:25 (25min)

Contributed talks (Session 2) - Deep Learning Acceleration of Flash Algorithms to Simulate Multicomponent Multiphase Systems.

Paul Mcginn, Chaouki Habchi, Julien Bohbot, Thibault Faney, Bruno Delhom, Jean-Charles De Hemptinne (IFPEN) - Deep Learning Acceleration of Flash Algorithms to Simulate Multicomponent Multiphase Systems.

15:25 - 15:55 (30min)

Coffee break and posters

15:55 - 16:20 (25min)

Contributed talks (Session 3) - Sample efficient fluid flow control using neuroevolution guided deep reinforcement learning

Tarun Singh, Laurent Cordier, Ronan Fablet (PPRIME, Lab-STICC) - Sample efficient fluid flow control using neuroevolution guided deep reinforcement learning

16:20 - 16:45 (25min)

Contributed talks (Session 3) - Solving source- and parameter-dependent PDEs with a multi-input Fourier Neural Operator

Fanny Lehmann, Filippo Gatti, Michaël Bertin, Didier Clouteau (CEA-DAM, LMPS) - Solving source- and parameter-dependent PDEs with a multi-input Fourier Neural Operator

16:45 - 17:10 (25min)

Contributed talks (Session 3) - Learning operators via Deep Learning models.

Joel Soffo (Airbus, Centrale Nantes) - Learning operators via Deep Learning models.

17:10 - 17:20 (10min)

Conclusion

Paul Labrogère, CEO, IRT SystemX

17:20 - 17:30 (10min)

Closing cocktail

Program