Generative Models and Optimal Transport
Marco Cuturi (ENSAE, Université Paris-Saclay)
When |
Feb 12, 2018
from 01:30 to 02:30 |
---|---|
Where | ENS de Lyon, Site Monod, salle à préciser |
Add event to calendar |
vCal iCal |
Abstract: A recent wave of contributions in machine learning center on the concept of generative models for extremely complex data such as natural images. These approaches provide principled ways to use deep network architectures, large datasets and automatic differentiation to come up with algorithms that are able to synthesize realistic images. We will present in this talk how optimal transport is gradually establishing itself as a valuable tool to carry out this estimation procedure.