Goals of the chair
Over the last few decades, uncertainty quantification has proved highly effective for analyzing physical systems, in fields such as aeronautics, space and energy.
More recently, it has also been demonstrated that uncertainty quantification methods can be advantageously transferred from physical applications to artificial intelligence (AI) systems, in particular for the analysis of explainability and fairness.
The UQPhysAI Chair will improve fundamental understanding and develop robust and widely applicable algorithms for two cornerstones of uncertainty quantification: sensitivity analysis and active learning with Bayesian models.
For sensitivity analysis, the focus will be on causal analysis and the large size and complexity of inputs and outputs.
With regard to Bayesian models and active learning, the Chair will focus on nested and coupled environments, and on exploiting new and more diverse techniques for quantifying uncertainties.
Methodological developments will be fed by and applied to realistic problems and data for physical and artificial intelligence systems, in collaboration with industrial partners: ONERA, Airbus, Liebherr and Vitesco.
Co-chair
- Nathalie Bartoli (DR ONERA)
- Jérôme Morio (DR ONERA)
- Mélisande Albert (MCF IMT INSA)
Participants
- Paul Novello (IR IRT Saint Exupéry)
- Disentangling causality from dependency in global sensitivity analysis
- Calculating sensitivity indices for complex, high-dimensional inputs
- Robustness of sensitivity analysis with respect to structural choices and dependency structure
- Bayesian models for nested and coupled systems
- More diverse uncertainty quantification for more effective active learning
- Multidisciplinary design and optimization
- Explainable and fair artificial intelligence
- IFP Energies Nouvelles: Bayesian calibration of geophysical computer models, PhD thesis by A. Barry, under the supervision of F. Bachoc and C. Prieur, since Jan 2022
- CIROQUO Consortium 6 industrial institutes and 8 academic institutes: research on quantification of uncertainties, since Jan 2021
- IRSNPrediction of complex outputs for nuclear safety, PhD thesis by F. Gossard, supervised by F. Bachoc and J. Baccou, since Nov 2023
- ONERAActive learning in nested and coupled settings, PhD thesis by Robert Koprinkov, supervised by F.Bachoc and N. Bartoli, since Sept 2024
- ONERANew challenges in sensitivity analysis: dependence, causality and high-dimension, PhD thesis by Lucas Monteiro, supervised by F. Bachoc and J. Morio, since Jan 2025
- group, LIEBHERR and VITESCOAI4SAVE project.
- Pôle d'Expertise de la Régulation Numérique and INRIA Regalia pilot project : Explicable and Fair Artificial Intelligence, Sep 2024
- Large-scale, more informative global sensitivity analysis
- Highly applicable Bayesian active learning with improved mathematical guarantees
English 
French