BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Are we making progress in unlearning?  - Eleni Triantafillou (Goo
 gle DeepMind)
DTSTART:20241113T110000Z
DTEND:20241113T123000Z
UID:TALK224401@talks.cam.ac.uk
CONTACT:Xianda Sun
DESCRIPTION:Machine unlearning is the problem of removing the influence of
  a subset of training data from machine learning models. This problem is e
 njoying increased attention recently due to excitement around using this t
 echnology to remove outdated\, harmful\, private or no-longer-permissible 
 data from trained models\, in order to increase their accuracy\, safety\, 
 or protect privacy. A straightforward solution to the problem is to remove
  the unwanted data from the training set and retrain a new model from scra
 tch. However\, that solution is inefficient and impractical\, especially i
 n the era of increasingly large models that are increasingly expensive to 
 train. Can we instead cause models to "forget" a subset of their training 
 data after the fact? While this problem has close ties to many research ar
 eas\, including continual learning\, transfer learning and privacy\, machi
 ne unlearning is still at its infancy\, with many open questions remaining
 \, both in how to evaluate success as well as how to improve upon existing
  methods. In this talk\, I will discuss recent progress and challenges rem
 aining\, highlighting open questions and important directions for the comm
 unity.
LOCATION:Cambridge University Engineering Department\, CBL Seminar room BE
 4-38.
END:VEVENT
END:VCALENDAR