Francis Bach

Together with the great Frederik Kunstner (fkunstner.github.io)

Not all scaling laws are nice power laws. This month’s blog post: Zipf’s law in next-token prediction and why Adam (ok, sign descent) scales better to large vocab sizes than gradient descent: francisbach.com/scaling-laws...

Avec en prime une photo du lac de Roselend.

Jamais deux sans trois. Le département d’informatique de l’ENS toujours en forme. 131 km, 4500m de dénivelé, cette fois-ci avec du Beaufort aux ravitaillements!
Bravo Olivier Cappé! #LEtapeduTour

Tired of lengthy computations to derive scaling laws? This post is made for you: discover the sharpness of the z-transform!
francisbach.com/z-transform/

Tired of lengthy computations to derive scaling laws? This post is made for you: discover the sharpness of the z-transform!
francisbach.com/z-transform/

A Collectivist, Economic Perspective on AI Information technology is in the midst of a revolution in which omnipresent data collection and machine learning are impacting the human world as never before. The word "intelligence" is being used as...

What if AI isn’t about building solo geniuses, but designing social systems?
Michael Jordan advocates blending ML, economics, and uncertainty management to prioritize social welfare over mere prediction.
A must-read rethink.
arxiv.org/abs/2507.062...

Big thanks to the COLT 2025 organizers for an awesome event in Lyon! Here are the slides from my keynote this morning in case you’re curious about the references I mentioned: www.di.ens.fr/~fbach/fbach...

Register at PAISS 1-5 Sept 2025 @inria_grenoble with very talented speakers this year 🙂
paiss.inria.fr
cc @mvladimirova.bsky.social

The PAISS summer school is back with an incredible line of speakers (and more to come). Spread the word !

Épisode 5 de notre série "Les nouveaux visages de l’Académie des sciences" : La statistique, une science en son temps.

Retrouvez la vidéo sur Youtube : www.youtube.com/watch?v=2AhT...

Announcing : The 2nd International Summer School on Mathematical Aspects of Data Science
mathsdata2025.github.io
EPFL, Sept 1–5, 2025

Speakers:
Bach @bachfrancis.bsky.social
Bandeira
Mallat
Montanari
Peyré @gabrielpeyre.bsky.social

For PhD students & early-career researchers
Apply before May 15!

[NOUVELLE SERIE "LES NOUVEAUX VISAGES DE L'ACADEMIE DES SCIENCES]
Episode n°1 : Anne Canteaut : une architecte de la cryptographie moderne
www.youtube.com/watch?v=xyGC...

Futur best seller!

Characterizing finely the decay of eigenvalues of kernel matrices: many people need it, but explicit references are hard to find. This blog post reviews amazing asymptotic results from Harold Widom (1963!) and proposes new non-asymptotic bounds.
francisbach.com/spectrum-ker...

The must-read introduction to PAC-Bayes!

🔬✨ Journée des Femmes et des filles de science✨🔬

À travers leurs parcours inspirants et leurs engagements, les académiciennes, et toutes les femmes et filles de science façonnent la recherche d’aujourd’hui et de demain.

Pour aller plus loin : urls.fr/PK5xg9

Building Bridges between Regression, Clustering, and Classification Regression, the task of predicting a continuous scalar target y based on some features x is one of the most fundamental tasks in machine learning and statistics. It has been observed and...

Check out our paper, with Lawrence Stewart and @bachfrancis.bsky.social

Link: arxiv.org/abs/2502.02996

1/8

Behind the Term Sheet: Bioptimus’ $41M Series A The GPT of Biology Raising the Bar of AI-Driven Scientific Research

If you're curious of what is "behind a term sheet", don't miss this account by the Cathay innovation team who led our recent series A at Bioptimus

medium.com/cathay-innov...

An inspirational talk by Michael Jordan: a refreshing, deep, and forward-looking vision for AI beyond LLMs.
www.youtube.com/live/W0QLq4q...

The Surprising Agreement Between Convex Optimization Theory and Learning-Rate Scheduling for Large Model Training We show that learning-rate schedules for large model training behave surprisingly similar to a performance bound from non-smooth convex optimization theory. We provide a bound for the constant schedul...

Learning rate schedules seem mysterious? Why is the loss going down so fast during cooldown?
Turns out that this behaviour can be described with a bound from *convex, nonsmooth* optimization.

A short thread on our latest paper 🚞

arxiv.org/abs/2501.18965

Early stopping on validation loss? This leads to suboptimal calibration and refinement errors—but you can do better!
With @dholzmueller.bsky.social, Michael I. Jordan, and @bachfrancis.bsky.social, we propose a method that integrates with any model and boosts classification performance across tasks.

I've been eagerly awaiting this book for years! At last, a standalone and meticulous exposition of the current mathematical principles of machine learning, adorned with beautiful proofs. Well done and thank you,
@bachfrancis.bsky.social . Discover more here: francisbach.com/my-book-is-o...

A happy author discovering the first hard copies

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...

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...

Breaking news : L'Académie des sciences accueille 18 nouveaux membres dès 2025, avec une majorité féminine pour la 1ère fois depuis 1666 👩‍🔬 : un symbole fort pour la parité en science ! 💥

🔗 En savoir plus sur les nouveaux membres : urlr.me/mntDHX

New opening! Post-doctoral position on relaxation methods for large-scale optimization and the management of electrical systems, in collaboration between EDF and Inria Saclay and Paris. See more details here: laurentpfeiffer.github.io/postdoc/