Austin J. Stromme
I am an assistant professor in the Statistics Department at ENSAE/CREST. I work at the intersection of optimization, geometry, and probability, and am particularly interested in the effects of symmetries and over-parameterization on learning dynamics and generalization. Previously, I received my PhD from MIT EECS. (CV, Scholar, X)
Recent and selected papers
All papers →On the implicit regularization of Langevin dynamics with projected noise
Govind Menon, Austin J. Stromme, Adrien Vacher
Conference on Learning Theory (COLT) 2026
The ballistic limit of the log-Sobolev constant equals the Polyak-Łojasiewicz constant
Sinho Chewi, Austin J. Stromme
Annales de l'Institut Henri Poincaré (B) Probabilités et Statistiques, 2026+
Minimum intrinsic dimension scaling for entropic optimal transport
Austin J. Stromme
Extended abstract in International Conference on Soft Methods in Probability and Statistics 2024,
Best paper award
On the sample complexity of entropic optimal transport
Philippe Rigollet, Austin J. Stromme
Annals of Statistics 2025
Fast convergence of empirical barycenters in Alexandrov spaces and the Wasserstein space
Thibaut Le Gouic, Quentin Paris, Philippe Rigollet, Austin J. Stromme
Journal of the European Math Society (JEMS) 2022
Writings
All writings →On the industrialization of mathematics
Mostly obvious thoughts on the industrial age of mathematics, largely written prior to the Navier-Stokes announcement.