Ph.D. student in at @Inria_Saclay working on Optimization and Machine Learning @matdag.bsky.social
Joined July 2019
- Tweets 381
- Following 546
- Followers 516
- Likes 1,186
10 Photos and videos
Pinned Tweet
19 Apr 2023
📣📣 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
2
14
43
11,350
Mathieu Dagréou retweeted
May 27
"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
3
9
67
7,245
Mathieu Dagréou retweeted
May 26
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.
1
11
93
34,604
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…
1
7
34
2,032
Mathieu Dagréou retweeted
6 Oct 2025
This is the way.
1 Oct 2025
This is the third, last, and best paper from my PhD. By some metrics, an ML PhD student who writes just three conference papers is "unproductive." But I wouldn't have had it any other way 😉 !
1
3
72
10,371
Mathieu Dagréou retweeted
3 Oct 2025
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?
11
54
465
30,755
Mathieu Dagréou retweeted
1 Sep 2025
🚟 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…
1
12
81
4,964
Mathieu Dagréou retweeted
14 Jul 2025
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!
5
50
301
21,138
Mathieu Dagréou retweeted
1 Jul 2025
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…
4
21
67
5,498
Mathieu Dagréou retweeted
17 Jun 2025
❓ 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"
8
65
486
34,589
Mathieu Dagréou retweeted
18 Jun 2025
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
5
15
204
20,267
Mathieu Dagréou retweeted
7 Jun 2025
🧵 I'll be at CVPR next week presenting our FiRe work 🔥
TL;DR: We go beyond denoising models in PnP with more general restoration (e.g. deblurring) models!
A starting point observation is that images are not fixed-points of restoration models:
1
5
17
2,201
Mathieu Dagréou retweeted
3 Jun 2025
📣 New preprint 📣
**Differentiable Generalized Sliced Wasserstein Plans**
w/
L. Chapel
@rtavenar
We propose a Generalized Sliced Wasserstein method that provides an approximated transport plan and which admits a differentiable approximation.
arxiv.org/abs/2505.22049 1/5
1
4
30
2,688
Mathieu Dagréou retweeted
24 Apr 2025
It was received quite enthusiastically here so time to share it again!!!
Our #ICLR2025 blog post on Flow M atching was published yesterday : iclr-blogposts.github.io/202…
My PhD student Anne Gagneux will present it tomorrow in ICLR, 👉poster session 4, 3 pm, #549 in Hall 3/2B 👈
1
5
11
848
Mathieu Dagréou retweeted
19 Feb 2025
Optimization algorithms come with many flavors depending on the structure of the problem. Smooth vs non-smooth, convex vs non-convex, stochastic vs deterministic, etc. en.wikipedia.org/wiki/Mathem…
4
108
511
21,396
Mathieu Dagréou retweeted
14 Feb 2025
A really fun project to work on. Looking at these plots side-by-side still amazes me! How well can **convex optimization theory** match actual LLM runs?
My favorite points of our paper on the agreement for LR schedules in theory and practice: 1/n
5 Feb 2025
Learning rate schedules seem mysterious?
Turns out that their behaviour can be described with a bound from *convex, nonsmooth* optimization.
Short thread on our latest paper 🚇
arxiv.org/abs/2501.18965
1
5
43
4,875
Mathieu Dagréou retweeted
5 Feb 2025
Learning rate schedules seem mysterious?
Turns out that their behaviour can be described with a bound from *convex, nonsmooth* optimization.
Short thread on our latest paper 🚇
arxiv.org/abs/2501.18965
3 Feb 2025
The sudden loss drop when annealing the learning rate at the end of a WSD (warmup-stable-decay) schedule can be explained without relying on non-convexity or even smoothness, a new paper shows that it can be precisely predicted by theory in the convex, non-smooth setting!
1/2
5
27
141
31,661
Mathieu Dagréou retweeted
3 Feb 2025
Learning rate schedulers used to be a big mistery. Now you can just take a guarantee for *convex non-smooth* problems (from arxiv.org/abs/2310.07831), and they give you *precisely* what you see in training large models.
See this empirical study:
arxiv.org/abs/2501.18965
1/3
5
71
430
28,690
Mathieu Dagréou retweeted
22 Jan 2025
Our work on geometric disentangled representation learning has been accepted to ICLR 2025! 🎊See you in Singapore if you want to understand this gif better :)
14 Dec 2024
Curious about the potential of optimal transport (OT) in representation learning? Join @CuturiMarco's talk at the UniReps workshop today at 2:30 PM! Marco will notably discuss our latest paper on using OT to learn disentangled representations. Details below ⬇️
19
152
14,285
Mathieu Dagréou retweeted
22 Jan 2025
The Mathematics of Artificial Intelligence: In this introductory and highly subjective survey, aimed at a general mathematical audience, I showcase some key theoretical concepts underlying recent advancements in machine learning. arxiv.org/abs/2501.10465
27
406
1,941
180,902
Mathieu Dagréou retweeted
21 Dec 2024
My book is (at last) out, just in time for Christmas!
A blog post to celebrate and present it: francisbach.com/my-book-is-o…
33
310
1,934
235,486