Amphithéâtre Marguerite de Navarre, Site Marcelin Berthelot
En libre accès, dans la limite des places disponibles
-

Résumé

Denoising diffusion models have led to impressive generative models in many domains. These algorithms learn from data a probability distribution in high dimension. They invert a stochastic differential which maps the probability distribution to a Gaussian. This talk will present the princple of these algorithms and provide a different perspective based on recent progress, with a focus on formulations that do not involve stochastic differential equations.

Intervenant(s)

Francis Bach

Directeur de recherche à l'Inria, professeur à l'ENS