BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//labri.fr//NONSGML kigkonsult.se iCalcreator 2.41.92//
CALSCALE:GREGORIAN
METHOD:PUBLISH
UID:a167f034-72ae-40fa-9991-4e137e4b1730
X-WR-CALNAME:GT AlgoDist\, «(Online) Continual Learning»\, Akka Zemari
X-WR-TIMEZONE:Europe/Paris
BEGIN:VTIMEZONE
TZID:Europe/Paris
TZUNTIL:20260329T010000Z
BEGIN:STANDARD
TZNAME:CET
DTSTART:20231029T030000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
RDATE:20241027T030000
RDATE:20251026T030000
END:STANDARD
BEGIN:DAYLIGHT
TZNAME:CEST
DTSTART:20240331T020000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
RDATE:20250330T020000
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:a167f034-72ae-40fa-9991-4e137e4b1730
DTSTAMP:20260510T233745Z
CLASS:PUBLIC
DESCRIPTION:Akka Zemari (LaBRI)\n\nTitle: (Online) Continual Learning\n\nAb
 stract:\n\nIn real-world supervised learning\, training data is often unav
 ailable simultaneously\, requiring models to adapt to incoming information
 . Sequential or naive training of pre-trained models on new tasks can lead
  to «forgetting» of prior knowledge. Incremental learning methods aim to a
 dapt models to new data while retaining past knowledge. Focusing on the st
 reaming scenario\, where data arrives one sample at a time\, our «Move-to-
 Data» method selectively adjusts network weights without systematic gradie
 nt descent. Compared to the state-of-the-art methods\, our approach outper
 forms and learns significantly faster\, presenting a promising solution fo
 r efficient and effective continual learning. \n\nN.B. : No prerequisites 
 are required to attend the presentation\; we will provide the necessary re
 minders to make the presentation self-contained.\n\n\n\nhttps://algodist.l
 abri.fr/index.php/Main/GT
DTSTART;TZID=Europe/Paris:20240515T110000
DTEND;TZID=Europe/Paris:20240515T120000
LOCATION:LaBRI salle 178 - lien visio https://webconf.u-bordeaux.fr/b/arn-4
 tr-7gp
SEQUENCE:0
SUMMARY:GT AlgoDist\, «(Online) Continual Learning»\, Akka Zemari
TRANSP:OPAQUE
END:VEVENT
END:VCALENDAR
