iiinova — where your digital self keeps evolving

Tam Le

@ntamle.bsky.social

34 followers 32 following 9 posts

Assistant professor at Université Paris Cité - LPSM. Working on optimization and machine learning.

Posts Replies Media

Tam Le @ntamle.bsky.social · 21/10/2025

I guess sometimes assumptions are not just about being realistic but rather help to gain insights and to understand

110

Tam Le @ntamle.bsky.social · 19/07/2025

I will be at ICCOPT (USC Los Angeles) next week to present this work. This will be on Wednesday July 23, alongside other nice talks on first-order methods: heavy ball ODE, optimal smoothing and nonsmoothness.

010

Tam Le @ntamle.bsky.social · 05/06/2025

We studied both continuous time and discretized dynamics. The paper also contains other results, on complexity in the convex case, on limit of limit points for discretized set-valued dynamics ...

010

Tam Le @ntamle.bsky.social · 05/06/2025

For instance, if a critical point is flat, it may be more sensible to errors, since the vanishing gradient cannot compensate the perturbations. We thus obtain an estimate (rho) of the fluctuations around the critical set, depending on the coefficients theta and beta.

120

Tam Le @ntamle.bsky.social · 05/06/2025

The idea of the analysis was to quantify how much critical points are flat or sharp. So we relied on KL inequality and a metric subregularity condition. They are satisfied for a large class of functions called "definable" or semialgebraic ones (say, piecewise polynomial).

110

Tam Le @ntamle.bsky.social · 05/06/2025

🎉🎉🎉Our paper "Inexact subgradient methods for semialgebraic functions" is accepted at Mathematical Programming !! This is a joint work with Jerome Bolte, Eric Moulines and Edouard Pauwels where we study a subgradient method with errors for nonconvex nonsmooth functions. arxiv.org/pdf/2404.19517

arxiv.org

383

Tam Le @ntamle.bsky.social · 06/05/2025

If it went down, then it must be a definable function, and I know you used a conservative gradient.

010

Tam Le @ntamle.bsky.social · 29/03/2025

I find Coste definitely more accessible to learn the topic, but when it comes to find/cite a specific property I prefer Van den dries!

020

Tam Le @ntamle.bsky.social · 22/01/2025

Our paper "Universal generalization guarantees for Wasserstein distributionally robust models" with Jérôme Malick is accepted at ICLR 2025!!!! I'm so happy about this one, we really improved the presentation since the first submission. arxiv.org/abs/2402.11981

arxiv.org

Universal generalization guarantees for Wasserstein distributionally robust models

Distributionally robust optimization has emerged as an attractive way to train robust machine learning models, capturing data uncertainty and distribution shifts. Recent statistical analyses have prov...

150