Designed and built with care, filled with creative elements

Top

Maurice Duits: « Determinantal point processes, log-correlated fields, and Jacobi operators »

DMA Salle W

The global fluctuations in models from random matrix theory, random tilings, and non-colliding particle systems are often governed by log-correlated Gaussian fields. In this talk, I will present an operator-theoretic viewpoint based on Jacobi (and CMV) matrices for a broad class of determinantal point processes associated with orthogonal polynomials. Instead of analyzing correlation functions and their asymptotics, this approach captures fluctuations efficiently through the spectral data of the underlying Jacobi operator. The emergence of log-correlated fields can then be traced back to deep results in analysis, such as the Strong […]

Sébastien Martineau

ENS — amphi Galois 45 rue d'Ulm, Paris, France

Séminaires des Mathématiques

Clément Cosco : « Champs log-corrélés et marches branchantes »

DMA Salle W

Il existe des objets mathématiques paraissant tout à fait distincts, comme certaines matrices aléatoires, une dynamique de populations, une équation aux dérivées partielles stochastique ou encore la fonction Zeta de Riemann, qui partagent des propriétés asymptotiques analogues. Ce qui les relie est l’apparition d’un champ log-corrélé décrivant leurs statistiques à grande échelle, ou de manière équivalente, l’existence d’une structure de marche branchante sous-jacente. Dans cet exposé, j’introduirai certains de ces modèles et tenterai de présenter leurs similitudes ainsi qu’une sélection de résultats connus.

Nathan Srebro: Learning to Answer from Correct Demonstrations

Amphi Jean Jaurès 45 rue d'Ulm, PARIS, France

Generative AI is increasingly presented as a potential substitute for humans, including as human research subjects in various disciplines. Yet there is no scientific consensus on how closely these in-silico clones could represent their human counterparts. While some defend the use of these “synthetic users,” others point towards the biases in the responses provided by the LLMs. Through an experiment using survey questionnaires, we demonstrate that these latter critics are right to be wary of using generative AI to emulate respondents, but probably not for the right reason. Our results […]

Olivier Graf

Salle W - ENS PSL 45 rue d'Ulm, Paris, France

Sylvain Ervedoza

Jussieu — salle 15-16-309 4 Place Jussieu, Paris

Thomas Hou

Salle W - ENS PSL 45 rue d'Ulm, Paris, France

Vlad Vicol

Jussieu -- salle 15-16-309 4 Place Jussieu, Paris, France

Hélène Mathis

Salle W - ENS PSL 45 rue d'Ulm, Paris, France