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:920@i2m.univ-amu.fr
DTSTART;TZID=Europe/Paris:20151106T140000
DTEND;TZID=Europe/Paris:20151106T150000
DTSTAMP:20240524T072505Z
URL:https://www.i2m.univ-amu.fr/evenements/remi-flamary-lagrange-universit
 e-de-nice-sophia-antipolis-optimal-transport-for-domain-adaptation-2/
SUMMARY: (...): Rémi Flamary (Lagrange\, Université de Nice-Sophia Antipo
 lis): Optimal transport for domain adaptation
DESCRIPTION:: Title: Optimal transport for domain adaptationnnAbstract: Dom
 ain adaptation is one of the most challenging tasks of modern data analyti
 cs. If the adaptation is done correctly\, models built on a specific data 
 representations become more robust when confronted to data depicting the s
 ame semantic concepts (the classes)\, but observed by another observation 
 system with its own specificities. Among the many strategies proposed to a
 dapt a domain to another\, finding domain-invariant representations has sh
 own excellent properties\, as a single classifier can use labelled samples
  from the source domain under this representation to predict the unlabelle
 d samples of the target domain. In this paper\, we propose a regularized u
 nsupervised optimal transportation model to perform the alignment of the r
 epresentations in the source and target domains. We learn a transportation
  plan matching both PDFs\, which constrains labelled samples in the source
  domain to remain close during transport. This way\, we exploit at the sam
 e time the few labeled information in the source and distributions of the 
 input/observation variables observed in both domains. Experiments in toy a
 nd challenging real visual adaptation examples show the interest of the me
 thod\, that consistently outperforms state of the art approaches. 
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/Paris
X-LIC-LOCATION:Europe/Paris
BEGIN:STANDARD
DTSTART:20151025T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
END:VTIMEZONE
END:VCALENDAR