Machine Learning for Physical Simulation challenge: and the winners are…

This contest invited the competing teams to address the challenge of improving baseline solutions of the Airfoils design use case by building new ML-based surrogate models, improving the trade-off accuracy/computational time while satisfying basic physical constraints.

After 4 months of intense competition, IRT SystemX announced on March 26th the five winning teams:

• 1st prize: Fabien Casenave, Brian Staber, Xavier Roynard, Abbas Kabalan (Safran Tech, Département des Sciences et Technologies Numériques) – “MMGP: A Mesh Morphing Gaussian Process based machine learning method for regression of physical problems under nonparametrized geometrical variability”
• 2d prize: Pierre Serrano, Nawar Ammari (Framatome (DTIM)) – “Piecewise metamodel : a combined GNN and FC approach”
• 3rd prize: Xavier Bertrand, Frédéric TOST, Giovanni Catalani (Airbus) – “Multiscale Implicit Neural Representations”
• 4th prize: Anthony Kalaydjian (ENSTA Paris – EPFL), Anton Balykov (EPFL), Adrien CHAN-HON-TONG (ONERA) – “Subsampled Bi-transformer”
• 5th prize: Henri Durliat, Hamza AHFIDI (Framatome DTIPL) – “NeurEco based MLP”

Congratulations to all! This competition is the first step of a series of other challenges to come… Stay tuned!