Emmanuel Menier, PhD student within the HSA project and the IA2 program at IRT SystemX, will defend his thesis on January 25, 2024,at 9:30 am in the theses room of building 650 at LISN (Gif-sur-Yvette), on the following topic: “Deep Learning for Reduced Order Modeling”.
>> Link to access to the online thesis defense <<
Abstract:
Dynamical systems are generally modeled using Partial Differential Equations (PDE). These models are intricately linked to the way scientists observe the world and, as such, they are limited by our understanding of the behavior of the systems under study. For example, models such as the Navier Stokes equations only account for the local interactions in fluid systems, and ignore the underlying phenomena that drive the system as a whole. This often leads to a poor understanding of the dynamical problems under study and excessive computational costs associated with the numerical resolution of PDE-based models. In this thesis, we discuss the way dynamical data can be exploited to derive better representation spaces for physical systems as well as computationally efficient models, called reduced order models. We then propose to leverage the approximation power of neural networks to derive novel, improved reduced order modeling methods. The modeling techniques proposed in this thesis are built around the concept of hybridization between physical and data driven modeling. We leverage pre-existing knowledge of dynamical systems into theoretically grounded, accurate, and interpretable dynamical models to address the computational costs issues associated with standard physical modeling, while avoiding complete reliance on data.
Jury composition:
- Gianluigi ROZZA – Professor, SISSA, Trieste – Rapporteur & Examinateur
- Amaury HABRARD – Professor, Université Jean Monnet, Saint-Etienne – Rapporteur & Examinateur
- Nicolas THOME – Professor, Sorbonne Université, Paris – Examinateur
- Phaedon-Stelios KOUTSOURELAKIS – Professor, Technical University of Munich – Examinateur
- Johannes BRANDSTETTER – Assistant Professor, Johannes Kepler University, Linz – Examinateur
- Taraneh SAYADI – Chargé de Recherche, CNAM, Paris – Examinateur
Partner laboratory:
LISN
