Researcher in applied mathematics & machine learning
Samuel Hurault
CR CNRS · MMSID team, LIGM, Université Gustave Eiffel
About Me
I am a CNRS researcher in the MMSID team at LIGM (Université Gustave Eiffel, France). My research focuses on generative modeling, image inverse problems, and optimization, with a particular interest in the theoretical analysis of denoising diffusion models. I am also a core developer of the Python library DeepInv, which provides a unified framework for solving inverse problems with deep neural networks.
Previously, I was a postdoctoral researcher at ENS Paris with Gabriel Peyré, and I obtained my Ph.D. from Université de Bordeaux, co-supervised by Nicolas Papadakis and Arthur Leclaire.
Research interests
- Diffusion models & Flow Matching
- Drifting models
- Image inverse problems
- Proximal algorithms
- Large Language Models
Selected work
Recent publications
Geometry-aware discretization error of diffusion models
arXiv preprint arXiv:2605.08392, 2026
Seminars & conferences
Recent talks
Luminy, Marseille, France
Learning and Optimization in Luminy (LOL)
Talk Geometry-Aware Sampling Error of Diffusion Models
Caen, France
Summer School AI for Science
Lecture Mathematics of Diffusion Models and Their Application for Image Inverse Problems
Grenoble, France
Grenoble AI4Science Workshop
Talk Diffusion Models, Flow Matching and Their Applications for Image Inverse Problems
Edinburgh, United Kingdom
SIAM Conference on Optimization
Talk Geometry-Aware Discretization Error of Diffusion Models