BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Sliced Wasserstein on Manifolds: Spherical and Hyperbolical cases
DTSTART:20230119T140000
DTEND:20230119T150000
DTSTAMP:20260513T181953Z
UID:f786c4ad0a044f105b7fc0caa0ddb6b137189df13c47e41fcc02b78d
CATEGORIES:Conferences - Seminars
DESCRIPTION:Nicolas Courty is Full Professor at University Bretagne Sud 
 since 2018. He obtained his PhD degree in 2002 from INSA Rennes and his 'H
 abilitation à diriger des recherches' in. 2013\, on the topic of computer
  graphics and animation (avatars\, crowds)\, with a specialization in data
 -driven methods. He now leads the Obelix team in IRISA\, dedicated to ma
 chine learning and its applications to Earth Observation. He is an experie
 nced researcher in the domain of machine learning and AI. Among others\, h
 e has published several papers in top tier machine learning conferences (N
 eurIPS\, ICLR\, ICML\, AISTATS\, etc.)\, computer vision (IEEE TPAMI\, ECC
 V\, ACCV) and remote sensing (IEEE TGRS\, ISPRS journal). From 2014\, he h
 as developed an expertise in the domain of optimal transport and related a
 pplications to machine learning. From 2020\, he pilots an ANR Chair progra
 m on AI (OTTOPIA)\, on the topic of applied optimal transport for Remote S
 ensing.\n\n\nOptimal transport has received a lot of attention into the ma
 chine learning and computational geometry communities recently. Many varia
 nts of the associated Wasserstein distance have been introduced to reduce 
 its original computational burden. In particular the Sliced-Wasserstein di
 stance (SW)\, which leverages one-dimensional projections for which a clos
 ed-form solution of the Wasserstein distance is available\, has received a
  lot of interest. Yet\, it is restricted to data living in Euclidean space
 s\, while the Wasserstein distance has been studied and used recently on m
 anifolds. In this talk I will discuss novel methodologies to transpose SW 
 to the Riemannian manifold case. By appropriately choosing a proper Radon 
 transform\, we show how fast and differentiable algorithms can be designed
  in two cases: Spherical and Hyperbolic manifolds. After discussing some o
 f the theoretical properties of those novel discrepancies\, I will showcas
 e applications in machine learning problems\, where data naturally live on
  those spaces. \n\n
LOCATION:ELD 020 https://plan.epfl.ch/?room==ELD%20020
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR