Théorique, numérique
This is a CNRS-funded PhD at LJLL, Sorbonne Université, at the interface between Mathematics, Machine Learning, Statistical Physics, and Computational Chemistry. The project is co-advised by Pierre Monmarché (LJLL applied maths) and Jérôme Hénin (CNRS, Laboratoire de Biochimie théorique, computational Biophysics). Our goals are to develop better descriptions of high-dimension free energy landscapes (for example, of biological molecules) and algorithms to sample them in numerical simulations. One of the applications is to explore and sample the "loss landscape" of deep neural networks.