Optimal transport for machine learning

Date(s) : 19/01/2018   iCal
14 h 00 min - 15 h 00 min

First we present a brief introduction to optimal transport and to the Wasserstein distance. Next we will discuss recent applications of OT in Machine Learning. OT can be used to estimate a mapping between distributions for color adaptation between images and domain adaptation. But it is also a very powerful data fitting term for learning with histograms or empirical distributions for classification, audio spectral unmixing and Generative Adversarial networks.


Catégories Pas de Catégories

Retour en haut 

Secured By miniOrange