[ICMNS] Séminaire Neuromathématiques Lorenzo Rosasco

[Alessandro Sarti]

Séminaire 

Neuromathématiques 




Organisé par 

Giovanna Citti (University of Bologna) 

JP Nadal ( CAMS - EHESS) 

Jean Petitot (CAMS - EHESS) 

Jerome Ribot (Collège de France) 

Alessandro Sarti (CAMS - EHESS/CNRS) (reférent) 




mardi 10 mars 14h30-16h30 




Salle D2.2, Collège de France , 11, Place Marcelin-Berthelot, 75005, Paris 

On se retrouve à l'accueille à 14h15 et on y va à la salle tous ensemble 




Lien Zoom: https://cnrs.zoom.us/j/95769026148?pwd=QZgsojSV83DeALyG1pGbvZtmr0H332.1 




______________ 


Lorenzo Rosasco 

Université de Genova 



Learning theory of neural networks through the lens of reproducing kernel Banach spaces 




We present a functional framework for shallow neural networks based on reproducing kernel Banach spaces. This approach enables a nonparametric treatment of neural networks, in direct analogy with kernel methods. A representer theorem shows that finite networks are optimal for empirical risk minimization. Estimation and approximation error bounds can then be derived in linear function spaces. As a byproduct, universality results and approximation bounds can be proved, showing that neural networks can adapt to latent structure in the problem. Further, we derive complexity estimates based on the Rademacher complexities of RKBS balls, independent of network size. Time permitting we will discuss extension to deep networks. 



Programme 2026 




mardi 10 février 

Marcelo Bertalmio 

Consejo Superior de Investigaciones Científicas, Madrid 




mardi 10 mars 

Lorenzo Rosasco 

Université de Genova 




mardi 14 Apr 

Cyril Monier 

Université Paris-Saclay 




12 Mai 

Remco Duits 
Eindhoven University 



[ https://enseignements.ehess.fr/2025-2026/ue/339 | https://enseignements.ehess.fr/2025-2026/ue/339 ] 

Inscription au lien [ https://participations.ehess.fr/ | https://participations.ehess.fr/ ] 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://listserver.vu.nl/pipermail/icmns-beta/attachments/20260304/065fcdd0/attachment-0001.html>