Next was a fantastic talk by @zdeborova.bsky.social on investigating the foundations of generalization in attention-based models at the @kempnerinstitute.bsky.social. Highly recommend www.youtube.com/watch?v=hECI... (3/7)
Get to know @zdeborova.bsky.social, Assoc Prof at EPFL🇨🇭. She’s an ELLIS Fellow, in ELLIS Unit Lausanne (@epfl-ai-center.bsky.social) & Program Quantum & Physics Based ML.
She works at the intersection of statistical physics, machine learning, and high-dimensional statistics.
#WomenInELLIS
#GRETSI2025, #EUSIPCO2025 sont terminés, la soumission à #ICASSP révolue, il fait gris. Abonner-vous à la chaîne YouTube #GRETSI @gretsi6095, et visionnez par exemple Lenka Zdeborova (Professeur EPFL), "Apprentissage (profond) et Physique Statistique"
youtu.be/yLFQUuVcC_0?...
Would you like to come to my lab for a postdoctoral position? Apply for the EPFL AI Centre fellowship: epfl.ch/research/fun.... FYI, I have funding (ERC & Simon's collaboration) also for 1-2 top candidates who will not secure the fellowship. Same for
@krzakalaf.bsky.social lab.
Scaling laws don’t just show up in test error — they leave fingerprints in the weight spectrum.
In the feature learning regime, we map this connection: phase diagrams of scaling exponents <-> spectral signatures of trained weights. The paper is: arxiv.org/abs/2509.24882
See the speakers for 2025-26 #KempnerInstitute Seminar Series!
bit.ly/46xphKq
@moiradillon.bsky.social @lisik.bsky.social @profsanjeevarora.bsky.social @kyunghyuncho.bsky.social @scychan.bsky.social @stellaathena.bsky.social @judithfan.bsky.social @engeltatiana.bsky.social @zdeborova.bsky.social
#Communiqué 🗞️ La médaille d'or 2025 du CNRS est décernée à Stéphane Mallat, mondialement reconnu pour ses travaux autour des mathématiques appliquées au traitement du signal et à l’intelligence artificielle. 👏
👉 cnrs.fr/fr/presse/en...
#TalentsCNRS 🏅
Led by @stanford.edu's @suryaganguli.bsky.social, our Simons Collaboration on the Physics of Learning & Neural Computation will study AI's fundamental scientific principles: www.simonsfoundation.org/simons-foundation-launches-collaboration-on-the-physics-of-learning-and-neural-computation/ #science
A no-miss opportunity to join EPFL. We are a vibrant (had to use it 😉) community working in all branches of basic sciences and engineering, and interdisciplinary studies are strongly encouraged.
If you want more information, contact the committee heads, or feel free to DM me.
tinyurl.com/bd27ky7p
Today's presentation was given by Andrew Saxe, University College London "Models of representation learning dynamics " during the conference organized by @krzakalaf.bsky.social & @zdeborova.bsky.social
EPFL Université de Corse Université Côte d'Azur @cnrs-dr12.bsky.social
This morning, Praya Sur from @harvard.edu gave her presentation "Two problems at the intersection of statistics and machine learning: Calibration and Data Integration"
during the conference organized by @krzakalaf.bsky.social & @zdeborova.bsky.social
Brice Huang, from Stanford University gave a talk this morning at @iescargese.bsky.social about "Spins Glass, Algorithms, and Inference" during the conference organized by@krzakalaf.bsky.social & @zdeborova.bsky.social
@icepfl.bsky.social@cnrs-dr12.bsky.social
#statisticalPhysic #MachineLearning
Saturday, during the conference organized by @krzakalaf.bsky.social & @zdeborova.bsky.social, M. Yue Lu @harvard.edu gave a talk about " Gaussian Equivalence for Nonlinear random matrices:Why it works, and where it fails"
@icepfl.bsky.social Université Côte d’Azur Université de Corse
Peter Bartlett UCBerkeley gave a talk yesterday at @iescargese.bsky.social about "Optimization in neural nets, large steps size and implicit bias" during the conference organized by @krzakalaf.bsky.social & @zdeborova.bsky.social @icepfl.bsky.social
#statisticalPhysic #MachineLearning
Presentation on Thursday, August 7, given by Bruno Loureiro, ENS Paris, on “An overview on Gaussian Multi-Index Model" during the conference organized at IES Cargese by @krzakalaf.bsky.social & @zdeborova.bsky.social
EPFL Université Côte d’Azur Université de Corse @cnrs-dr12.bsky.social
From August 05 to August 15, l'IES Cargese will be hosting "Statisticial Physic and Machine Learning : Moving Forward" organized by Florent Krzakala & Lenka Zdeborova @krzakalaf.bsky.social @zdeborova.bsky.social
Université de Corse Université Côte d'Azur @cnrs-dr12.bsky.social @icepfl.bsky.social
First talk today, given by @giuliobiroli.bsky.social, @lpens-phys.bsky.social on “Statistical physics and diffusion models” during the conference “Statistical physics and machine learning: Moving forward” organized at IES Cargese by @krzakalaf.bsky.social & @zdeborova.bsky.social
Today at the IES Cargese during conference “Statistical Physics & Machine Learning: Moving Forward”, Michael M. Bronstein, @mmbronstein.bsky.social @ox.ac.uk #AITHYRA, gave a presentation on “Geometric Deep Learning”.
@krzakalaf.bsky.social & @zdeborova.bsky.social
Little by little all of them … all intellectuals are uncomfortable for such a government. How is that not clear?
🎉 Congratulations to Anastasia Ailamaki, Mathias Payer, Lenka Zdeborová & Pascal Fua for receiving Advanced Grants for research in data systems, security, neural networks & computer-assisted engineering, a record 4 awards for EPFL’s IC School! 🏆
🔗 Learn more: news.epfl.ch/news/ic-rese...
While I contributed to this line of work and to this piece; I’m quite sure that modern AI would exist even without Hopfield and Hinton. There is a huge community behind each influential concept, and while prizes are given to individuals, it is the community that creates the science.
Towards StatPhys2025: interview with @zdeborova.bsky.social (EPFL) by Franco Bagnoli (Dept. Physics and Astronomy, University of Florence & Caffè-Scienza).
youtu.be/hplPGzFSRyo
Happy to share the recording of my plenary talk at Cosyne 2025 two days ago. You will learn about the statistical physics approach, phase transitions in learning, transformers, sequence models, attention etc. www.youtube.com/watch?v=PurZ...
Excited about our progress in characterizing The Computational Advantage of Depth in Learning with Neural Networks. Check out the number of samples that can be saved when GD runs on a multi-layer rather than on a two-layer neural network. arxiv.org/pdf/2502.13961
The series of three Simons Lectures I gave in MIT Math last April is now on YouTube. If you want to hear about *Computation through the lens of spin glasses*, have a look: www.youtube.com/embed/94XXGG...
Sure, feel free!
EPFL's new AI Center has a Call for applications for postdoc fellowships in all AI-related areas. Come join if you're interested in working with me and fantastic AI colleagues!
Extra Perk: We actually do have lots of GPUs !
Deadline: November 29th
More info at:
www.epfl.ch/research/fun...
I am doing just that since about a decade! No Elsevier no Nature/Springer I do not review for them, do not publish there. I refused a couple of editorial boards with them, numerous special issue invitations. If a young colleague writes in their cv this is their moral code I would deeply respect it.
Our advance in statistical physics analysis of simple graph convolutional neural networks arxiv.org/pdf/2402.03818 will appear at the Learning on Graphs conference: logconference.org. My student Odilon has now extended this to the multi-layer case, so more is coming soon.
🙋♀️
