Research scientist at CNRS, Nice, France
Contact
Samuel Vaiter
Laboratoire J.A. Dieudonné
CNRS UMR 7351, Université Côte d'Azur
Parc Valrose 06108 NICE CEDEX 2, France
email: samuel.vaiter@cnrs.fr
github: @svaiter,
bluesky: @samuelvaiter.com
one piece of math a day,
news/changelog
Vita
I am a CNRS researcher at the interface of mathematics and computer science. My current host lab is the Laboratoire J. A. Dieudonné of Université Côte d’Azur located in the beautiful city of Nice.
My research interests cover in particular the mathematical foundation of machine learning, especially linked to optimization. In particular, I work on graph machine learning, algorithmic differentiation, (stochastic ∨ bilevel ∨ nonsmooth) optimization. I am currently an Area Chair for NeurIPS, ICML, ICLR, and Action Editor for TMLR. I am also PI of the research projet ANR PRC ANR MAD and 3IA Côte d’Azur chairholder.
Previously, I was a member of Institut de Mathématiques de Bourgogne, located in the city of Dijon, France during 2015-2021 and a postdoc at CMAP, the applied math institute of École Polytechnique with Antonin Chambolle during the academic year 2014-2015. Before that, I did my PhD at CEREMADE (Univ. Paris-Dauphine) under the supervision of Gabriel Peyré. More information are available on my About page.
Five recent representative publications
See here for a full list.
- Jérôme Bolte, Tung Lê, Edouard Pauwels, SV. Geometric and computational hardness of bilevel programming. accepted to Mathematical Programming. 2025
- Yann Traonmilin, Rémi Gribonval, SV. A theory of optimal convex regularization for low-dimensional recovery. Information and Inference: A Journal of the IMA 13(2):66pp. 2024.
- Nicolas Keriven, SV. What functions can Graph Neural Networks compute on random graphs? The role of Positional Encoding. NeurIPS. 2023.
- Edouard Pauwels, SV. The Derivatives of Sinkhorn-Knopp Converge. SIAM Journal on Optimization 33(3):1494–1517. 2023.
- Mathieu Dagréou, Pierre Ablin, SV, Thomas Moreau. A framework for bilevel optimization that enables stochastic and global variance reduction algorithms. NeurIPS. 2022. (Oral paper).