Ph.D. student in at @Inria_Saclay working on Optimization and Machine Learning @matdag.bsky.social
Paris, France
Joined July 2019
- Tweets389
- Following544
- Followers522
- Likes1.2K
Pinned Tweet
📣📣 Preprint alert 📣📣
« A Lower Bound and a Near-Optimal Algorithm for Bilevel Empirical Risk Minimization »
w. @tomamoral, @vaiter & @PierreAblin
arxiv.org/abs/2302.08766
1/3
Mathieu Dagréou retweeted
#AISTATS 2027 Call for Paper is out, and submissions are open!
This year's edition is taking place in Montreal on May 3-6th, 2027 🇨🇦
virtual.aistats.org/Conferen…
New this year: AI review, and more!
Abstract deadline: September 29, 2026
Paper deadline: October 6, 2026
Mathieu Dagréou retweeted
Submissions will open soon for #AISTATS 2027
More info (soon) at: virtual.aistats.org/Conferen…
Submission opens Sept 15th 2027 at 🌍 🕜
- Beijing: 7:30 PM
- Paris: 1:30 PM
- Montreal/NY: 7:30 AM
- San Francisco: 4:30 AM
- AOE: 11:30PM (Sept. 14th)
Mathieu Dagréou retweeted
#AISTATS 2027 Call for Paper is out!
This year's edition is taking place in Montreal on May 3-6th, 2027 🇨🇦
Aymeric Dieuleveut and I are chairing
virtual.aistats.org/Conferen…
New this year: AI review, and more!
Abstract deadline: September 29, 2026
Paper deadline: October 6, 2026
Mathieu Dagréou retweeted
We released a major update to our book, "The Elements of Differentiable Programming", including a comprehensive, Gemini-assisted proofread.
arxiv.org/abs/2403.14606
Mathieu Dagréou retweeted
Now, to appear in Mathematical Programming :)
A new paper out with @TaeHo_Y00N and @ErnestRyu:
We looked at the design of optimal fixed-point algorithms.
That is, seeking to approximately solve T(y)=y using as few evaluations of the operator T() as possible. Maximally efficient methods are "minimax optimal" 1/
Mathieu Dagréou retweeted
Back in 2019 Yura Malitsky and I wrote a paper about adaptive estimation of gradient Lipschitzness in gradient descent.
In 2023 we tightened the constant factor in the estimate from 1/2 to 1/√2. We couldn't figure out if it could be further tightened to exactly 1.
Finally gpt 5.6 pro gave us the answer: even with a constant factor of 0.98 the method would not converge. It created a PEP-style counterexample, which I couldn't do when I tried using PEP back in 2019. It's not anyone's big conjecture but I think it gives a better example of how LLMs are useful for doing math.
Mathieu Dagréou retweeted
It was just announced at the International Congress of Mathematicians that Yurii Nesterov has been awarded the 2026 Gauss Prize, given once every 4 years. Hard to imagine a more deserving recipient.
Mathieu Dagréou retweeted
New preprint alert (my final PhD paper ♥️)
Twisted Schrödinger Bridge Matching (TSBM)
We “twist” classic diffusion losses to tackle trajectory inference under physical constraints, encoded via a differentiable state cost
Paper: arxiv.org/abs/2607.16987
with Github link
(1/5)
Mathieu Dagréou retweeted
Slides from my ICML tutorial "Is numerical optimization theory irrelevant to machine learning practice in 2026?":
cs.ubc.ca/~schmidtm/Document…
(Updated to fix some typos, incorporate feedback, and add some things I did not have time for. Will link to the video here when posted.)
Mathieu Dagréou retweeted
"It's easier to tune the LR for method A than for B."
We tried to formalize this for model-based stochastic optimization methods.
We find a key quantity, called stability index, that describes how stable a (weakly) convex bound is as a function of LR.
📚arxiv.org/abs/2602.09842
Mathieu Dagréou retweeted
What do JEPA-style self-distillation dynamics actually learn — and why do they sometimes avoid collapse?
In our new work with @BasileTerv987 and Jean Ponce, we tackle this question.
What surprised us: These dynamics provably recover representations with nonlinear-CCA structure.
Mathieu Dagréou retweeted
Our work "Busemann Functions in the Wasserstein Space" was accepted at #AISTATS2026
This is a joint work with Elsa Cazelles, Lucas Drumetz and @nicolas_courty.
I will be presenting it tomorrow at the poster 96, see you there!
Link: openreview.net/forum?id=Xpt0…
Mathieu Dagréou retweeted
Nesterov dropped a new paper last week on what functions can be optimized with gradient descent.
The idea is simple: we know GD can optimize both nonsmooth (bounded grads) and smooth (Lipschitz grads) functions, but smooth+nonsmooth satisfies neither property, so what can we do?
Mathieu Dagréou retweeted
🚟 New blog post:
On "infinite" learning-rate schedules and how to construct them from one checkpoint to the next
fabian-sp.github.io/posts/20…
Mathieu Dagréou retweeted
For evolving unknown PDEs, ML models are trained on next-state prediction. But do they actually learn the time dynamics: the "physics"?
Check out our poster (W-107) at #ICML2025 this Wed, Jul 16. Our "DISCO" model learns the physics while staying SOTA on next states prediction!
Mathieu Dagréou retweeted
Back from MLSS Senegal 🇸🇳, where I had the honor of giving lectures on differentiable programming. Really grateful for all the amazing people I got to meet 🙏 My slides are here github.com/diffprog/slides/b…
Mathieu Dagréou retweeted
❓ How long does SGD take to reach the global minimum on non-convex functions?
With @FranckIutzeler, J. Malick, P. Mertikopoulos, we tackle this fundamental question in our new ICML 2025 paper: "The Global Convergence Time of Stochastic Gradient Descent in Non-Convex Landscapes"
Mathieu Dagréou retweeted
I want to address one very common misconception about optimization. I often hear that (approximately) preconditioning with the Hessian diagonal is always a good thing. It's not. In fact, finding a good preconditioner is an open problem, which I think deserves more attention.
1/4
