BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
X-WR-CALNAME:Probability and Stochastic Processes Seminar: Rethinking proje
 ction networks: a functional approach to self-supervised learning
X-WR-TIMEZONE:Eastern Time (US & Canada)
BEGIN:VEVENT
DTSTAMP:20260513T224711Z
UID:tag:localist.com\,2008:EventInstance_50790567099451
DTSTART:20250922T180000Z
DTEND:20250922T190000Z
DESCRIPTION:Speaker: Wyatt Mackey\, ARL\n\nTitle: Rethinking projection net
 works: a functional approach to self-supervised learning.\n\nAbstract. Sel
 f-supervised learning (SSL)\, or learning without any human-provided label
 s\, has achieved a surge of interest in the machine learning community as 
 it begins to significantly outpace traditional training techniques in dive
 rse tasks from image classification and automatic segmentation to high lev
 el semantic scene decomposition. However\, SSL remains susceptible to cata
 strophic forgetting in large tasks like video understanding\, and depends 
 on a plethora of highly tuned parameters and auxiliary trained networks to
  converge to strong local minima. We investigate and reinterpret projectio
 n networks\, an auxiliary network that helps prevent dimension collapse in
  SSL. By re-examining the ultimate goals of SSL and transfer learning from
  the lens of linear dualities\, we introduce multiple new methods to strea
 mline and improve gradient conditioning in all SSL frameworks. We demonstr
 ate strong results on key benchmarks like image classification with small 
 models using these innovations\, suggesting a new and promising path forwa
 rd to scaling large SSL systems.
GEO:39.681267;-75.754997
LOCATION:Ewing Hall\, 336
SUMMARY:Probability and Stochastic Processes Seminar: Rethinking projection
  networks: a functional approach to self-supervised learning
URL;VALUE=URI:https://events.udel.edu/event/probability-and-stochastic-proc
 esses-seminar-rethinking-projection-networks-a-functional-approach-to-self
 -supervised-learning
CATEGORIES:Academics
CATEGORIES:College of Arts and Sciences
END:VEVENT
END:VCALENDAR
