BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.2.3.1//EN
TZID:Europe/Paris
X-WR-TIMEZONE:Europe/Paris
BEGIN:VEVENT
UID:6351@i2m.univ-amu.fr
DTSTART;TZID=Europe/Paris:20210702T143000
DTEND;TZID=Europe/Paris:20210702T153000
DTSTAMP:20241120T201410Z
URL:https://www.i2m.univ-amu.fr/evenements/differentially-private-sliced-w
 asserstein-distance/
SUMMARY:Liva Ralaivola (Criteo AI Lab): Differentially-private sliced Wasse
 rstein distance
DESCRIPTION:Liva Ralaivola: Developing machine learning methods that are pr
 ivacy preserving is today a central topic of research\, with huge practica
 l impacts. Among the numerous ways to address privacy-preserving learning\
 , we here take the perspective of computing the divergences between distri
 butions under the Differential Privacy (DP) framework --- being able to co
 mpute divergences between distributions is pivotal for many machine learni
 ng problems\, such as learning generative models or domain adaptation prob
 lems. Instead of resorting to the popular gradient-based sanitiziation met
 hod for DP\, we tackle the problem at its roots by focusing on the Sliced 
 Wasserstein Distance and seamlessly making it differentially private. Our 
 main contribution is as follows: we analyze the property of adding a Gauss
 ian perturbation to the intrinsic randomized mechanism of the Sliced Wasse
 rstein Distance\, and we establish the sensitivity of the resulting differ
 entially private mechanism. One of our important finding is that this DP m
 echanism transforms the Sliced Wasserstein distance into another distance\
 , that we call the Smoothed Sliced Wasserstein Distance. This new differen
 tially-private distribution distance can be plugged into generative models
  and domain adaptation algorithms in a tranparent way\, and we empirically
  show that it yields highly competitive performance compared with gradient
 -based DP approaches from the literature\, with almost no loss in accuracy
  for the domain adaptation problems that we consider.\nhttps://icml.cc/Con
 ferences/2021/Schedule?q=Differentially-Private+Sliced+Wasserstein+Distanc
 e\nThis seminar will be online\, please contact the organisers if you wish
  to attend (https://listes.math.cnrs.fr/wws/info/sem-signal-apprentissage)
 .
ATTACH;FMTTYPE=image/jpeg:https://www.i2m.univ-amu.fr/wp-content/uploads/2
 021/05/Liva_Ralaivola.png
CATEGORIES:Séminaire,Signal et Apprentissage,Virtual event
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/Paris
X-LIC-LOCATION:Europe/Paris
BEGIN:DAYLIGHT
DTSTART:20210328T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
END:VTIMEZONE
END:VCALENDAR