BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:DLN: Toward Developmentally Reasonably Self-Supervised Learning\, 
 Prof Dan Yamins (Stanford)
DTSTART:20210511T170000
DTEND:20210511T180000
DTSTAMP:20260407T024602Z
UID:eb57d879b0487489312f0707760c95cf1d61596aacba7fb0f9f704c2
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof Dan Yamins\nAbstract: Neural networks have proven effect
 ive learning machines for a variety of challenging AI tasks\, as well as s
 urprisingly good models of brain areas that underly real human intelligenc
 e.  However\, most successful neural networks are totally unrealistic as
  developmental models\, because they are trained in a supervised fashion o
 n large labelled datasets.  Unsupervised approaches to learning in neural
  networks are thus of substantial interest for furthering artificial intel
 ligence\, both because they would enable the training of networks without 
 the need for annotation\, and because they would be better models of the k
 ind of general-purpose learning deployed by humans.  In this talk\, I wil
 l describe a spectrum of recent approaches to unsupervised learning\, base
 d on ideas from cognitive science and neuroscience.  First\, I will discu
 ss breakthroughs in neurally-inspired unsupervised learning of deep visual
  embeddings that achieve that achieve performance levels on challenging vi
 sual categorization tasks that are competitive with those of direct superv
 ision of modern convnets.  Second\, I'll discuss our work building percep
 tion systems that make accurate long-range predictions of physical futures
  in realistic environments\, and show how these support richer self-superv
 ised visual learning. I'll also talk about the use of intrinsic motivation
  and curiosity to create interactive agents that self-curricularize\, pr
 oducing novel visual behaviors and learning powerful sensory representatio
 ns.   Finally\, I'll suggest ways in which these models are a better sta
 rting point for models of actual human visual development.\n\nA short bio 
 can be found here.
LOCATION:https://epfl.zoom.us/j/68683276041?pwd=WkFIQ1NjekhYQ3M3Y2pCNTBaUk
 V3Zz09
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR