BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Cambridge Ellis Unit Seminar Series- Dr Neil Houlsby- Learning gen
 eral visual representations: data\, scaling laws\, and fewer convolutions 
 - Speaker to be confirmed
DTSTART:20211029T130000Z
DTEND:20211029T140000Z
UID:TALK163867@talks.cam.ac.uk
CONTACT:Kimberly Cole
DESCRIPTION:Learning general visual representations\, those useful for man
 y tasks\, is a key challenge in Computer Vision. For many years\, Convolut
 ion Neural Networks (CNNs)\, typically trained on the ImageNet dataset\, h
 ave been used as a “backbone”\, or starting point\, for downstream tas
 ks. Perhaps surprisingly\, recent work has demonstrated large CNNs transfe
 r well to small downstream tasks\, even in the few-shot regime. However\, 
 such CNNs need large datasets for pre-training. While CNNs appear to have 
 just the right inductive biases for small or medium-scale training\, are t
 hey still optimal in modern transfer learning regimes? In this talk we wil
 l discuss some recent trends\, and surprising findings\, in architecture d
 esign\, scaling laws\, and visual representation learning.Learning general
  visual representations\, those useful for many tasks\, is a key challenge
  in Computer Vision. For many years\, Convolution Neural Networks (CNNs)\,
  typically trained on the ImageNet dataset\, have been used as a “backbo
 ne”\, or starting point\, for downstream tasks. Perhaps surprisingly\, r
 ecent work has demonstrated large CNNs transfer well to small downstream t
 asks\, even in the few-shot regime. However\, such CNNs need large dataset
 s for pre-training. While CNNs appear to have just the right inductive bia
 ses for small or medium-scale training\, are they still optimal in modern 
 transfer learning regimes? In this talk we will discuss some recent trends
 \, and surprising findings\, in architecture design\, scaling laws\, and v
 isual representation learning.
LOCATION:https://eng-cam.zoom.us/j/86919451784?pwd=N2JLSWdhWUs1U3JEVTZVY0J
 QWmM2QT09 
END:VEVENT
END:VCALENDAR