talks

Conference presentations, seminars, and workshop talks on optimization, machine learning, and LLM research.

PhD Defense

Université Grenoble Alpes (June 2026) Stochastic Optimization Algorithms in Machine Learning: Dynamics, Convergence, Generalization — Grenoble, France Materials: slides

Invited Seminars

MaLGa, Università di Genova (May 2026) The Long-Run Behavior of SGD in Non-Convex Landscapes: A Large Deviation Analysis — Genoa, Italy Materials: slides

ENS Lyon MLSP Seminar (April 2026) The Long-Run Behavior of SGD in Non-Convex Landscapes: A Large Deviation Analysis — Lyon, France Materials: seminar page

Morgan Stanley ML Research (December 2025) What is the lon-run behaviour of SGD — New York, USA Materials: slides

Inria Argo Team Seminar (December 2025) What is the lon-run behaviour of SGD — Grenoble, France Materials: slides

Université de Nice (December 2024) What is the long-run distribution of SGD? — Nice, France Materials: seminar page • slides

LPSM, Université de Paris (October 2024) What is the long-run distribution of SGD? — Paris, France Materials: seminar page • slides

Thoth Team Seminar, Inria (October 2024) What is the long-run distribution of SGD? — Grenoble, France Materials: seminar page • slides

SMAI MODE Conference (April 2024) Mirror methods: deterministic analysis — Lyon, France Materials: slides

Thoth Team Seminar, Inria (April 2021) Expressive power of invariant and equivariant graph neural networks — Grenoble, France Materials: slides

Conference and Workshop Presentations

SMAI MODE 2026 — Nice, France — Prix Dodu (best student talk) What is the long-run distribution of stochastic gradient descent? A large deviations analysis Materials: slides • paper • announcement

ICML 2025 — Upcoming The global convergence time of stochastic gradient descent in non-convex landscapes Materials: poster • paper

ICML 2024 — Montreal, Canada What is the long-run distribution of stochastic gradient descent? A large deviations analysis Materials: poster • paper

NeurIPS@Paris 2023 — Paris, France Exact generalization guarantees for (regularized) Wasserstein distributionally robust models Materials: workshop page • slides • paper

FOCM 2023 — Paris, France Exact generalization guarantees for (regularized) Wasserstein distributionally robust models Materials: conference page • poster • paper

ICCOPT 2022 — Bethlehem, USA Mirror methods in stochastic settings Materials: slides

Workshop in Erice 2022 — Sicily, Italy Regularized WDRO and generalization guarantees Materials: workshop page • slides

COLT 2021 — Boulder, USA The last-iterate convergence rate of optimistic mirror descent in stochastic variational inequalities Materials: slides • poster • paper

MIPT-UGA Workshop 2021 — Online Expressive power of invariant and equivariant graph neural networks Materials: workshop page • slides • paper

AISTATS 2020 — Online A tight and unified analysis of gradient-based methods for a whole spectrum of differentiable games Materials: slides • paper