Octobre 2019


Retour à la vue des calendrier
Jeudi 3 Octobre
Heure: 12:15 - 13:30
Lieu: Salle B107, bâtiment B, Université de Villetaneuse
Résumé: Optimal Transport for Machine Learning
Description: Bernard Kamsu-Foguem Optimal transport defines a set of geometric tools with interesting properties (comparison and morphology of probability measurements) that make it particularly suitable for solving large scale artificial learning problems.
Since probability distributions are omnipresent in Machine Learning (ML), whether theoretical or empirical, optimal transport can be useful in order to measure their separation or, even better, to be able to transform them at a lower cost.
We will investigate techniques based on optimal transport in certain practical and theoretical contexts (text classification, multi-label classification and domain adaptation), to contribute to solving the associated Machine Learning (ML) problems.