Uncertainty Quantification for Physical and Artificial Intelligence systems

Project funding and duration:

The project was funded by the ANITI AI cluster from 2024 to 2025 (€350 000), thanks to the "Starting Chair" instrument (SC).

The project had to end its initial format in 2025, earlier than anticipated, because the PI, François Bachoc, moved to the University of Lille. After 2025, the project is absorbed into the multi-PI Chair TRIAL (Trust and Responsibility in Artificial Intelligence), also funded by ANITI.

Project members:

• François Bachoc, Institut de Mathématiques de Toulouse (principal investigator)
  
• Mélisande Albert, Institut de Mathématiques de Toulouse
  
• Nathalie Bartoli, ONERA Toulouse
  
• Sylvain Dubreuil, ONERA Toulouse
  
• Jérôme Morio, ONERA Toulouse
  
• Paul Novello, IRT Saint Exupery
• Robert Koprinkov, PhD student, Institut de Mathématiques de Toulouse and ONERA Toulouse
• Lucas Monteiro, PhD student, Institut de Mathématiques de Toulouse and ONERA Toulouse

Summary of the project:

This project aims for new methods in uncertainty quantification applicable to both physical and AI systems. The three main axes are the following.

• Sensitivity analysis
• Bayesian active learning
• Real-world applications

How to acknowledge funding from the project:

• Our work has benefitted from the AI Interdisciplinary Institute ANITI. ANITI is funded by the France 2030 program under the Grant agreement n°ANR-23-IACL-0002.