MFAI26

Mathematical Foundations of Artificial Intelligence

Mexico City — National Autonomous University of Mexico (UNAM) · April 20–23, 2026

Organizers

Ramsés H. Mena (UNAM) Juan Carlos Pardo Millán (CIMAT) Philippe Rigollet (MIT)

The workshop will focus on mathematical structures underlying contemporary AI models. Topics include high-dimensional probability, stochastic analysis, PDE and mean-field limits, statistical mechanics, and optimization. The goal is to promote cross-disciplinary interaction and to deepen theoretical understanding of modern AI systems.

View program

Program

Time	Monday	Tuesday	Wednesday	Thursday
10:00-10:30	Peter Bartlett	Eric Vanden-Eijnden	Surbhi Goel	Arturo Jaramillo
10:30-11:00	Nicolas Flammarion	Michael Albergo	Andrej Risteski	Marco Avella
11:00-11:30	Joan Bruna	Nick Boffi	Enric Boix	Oscar Madrid Padilla
11:30-12:00	☕ Coffee break	☕ Coffee break	☕ Coffee break	☕ Coffee break
12:00-12:30	Damek Davis	Greg Pavliotis	Vardan Papyan	-
12:30-1:00	Alex Damian	Hugo Lavenant	Yusu Wang	-
1:00-1:30	Alexandre d'Aspremont	James Melbourne	Jaouad Mourtada	-
1:30-3:30	🌮 Lunch break	🌮 Lunch break	🌮 Lunch break	-
3:30-4:00	Sasha Rakhlin	Borjan Geshkovski	Lorenzo Rosasco	-
4:00-4:30	Joaquin Fontbona	Andrea Agazzi	Krishna Balasubramanian	-
4:30-5:00	☕ Coffee break	☕ Coffee break	☕ Coffee break	-
5:00-5:30	Murat Erdogdu	Jorge Gonzalez Cazares	Jose Blanchet	-
5:30-6:00	Nicolas Garcia-Trillos	Bernardo Flores	Matthew Zhang	-

Theme & goals

We aim to assemble a compact but diverse program where rigorous theory meets pressing questions from practice.

AI topics

Transformers & attention
Optimization landscapes, implicit regularization
Generalization & inductive bias in deep models
Graph neural networks
Foundations of generative models and diffusion-based learning

Mathematical topics

High-dimensional probability & concentration
Partial differential equations
Statistical mechanics & spin glass models
Stochastic analysis
Interacting particle systems

Confirmed participants

Andrea AgazziBern Michael AlbergoHarvard Marco AvellaColumbia Peter BartlettBerkeley/Google Krishna BalasubramanianUC Davis Jose BlanchetStanford Nick BoffiCarnegie Mellon Enric BoixUPenn Joan BrunaNYU Alexandre d'AspremontENS Paris Alex DamianHarvard Damek DavisUPenn Murat ErdogduU of Toronto Nicolas FlammarionEPFL Joaquin FontbonaU Chile Bernardo FloresU of Texas Nicolas Garcia-TrillosU Wisconsin Borjan GeshkovskiINRIA Surbhi GoelUPenn Jorge Gonzalez CazaresIIMAS Arturo JaramilloCIMAT Hugo LavenantBocconi University Oscar Madrid PadillaUCLA James MelbourneCIMAT Jaouad MourtadaENSAE Vardan PapyanU of Toronto Greg PavliotisImperial College Sasha RakhlinMIT Andrej RisteskiCarnegie Mellon Lorenzo RosascoGenova Eric Vanden-EijndenNYU Yusu WangUCSD Matthew ZhangU of Toronto

Venue

The conference will be hosted at Universidad Nacional Autónoma de México — Ciudad Universitaria, a UNESCO World Heritage site known for its iconic murals, museums and green campus.

Accommodations

We recommend that participants stay in the Coyoacán neighborhood. Coyoacán offers a wide range of guest houses and boutique hotels, and participants are invited to book their own accommodation independently.

A daily shuttle to the workshop will be arranged from the center of Coyoacán.

Coyoacán is a historic and bohemian area of Mexico City, known for its colonial architecture, cobblestone streets, and vibrant plazas such as Plaza Hidalgo. With its colorful markets, lively cafés, and strong artistic heritage—including landmarks like the Frida Kahlo Blue House Museum and the former home of Leon Trotsky—it offers a culturally rich and relatively tranquil setting with deep roots as an ancient village and early Spanish settlement.