BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Theoretical Foundations of Graph Neural Networks - Dr Petar Velič
 ković - DeepMind
DTSTART:20210217T150000Z
DTEND:20210217T160000Z
UID:TALK155341@talks.cam.ac.uk
CONTACT:Ben Karniely
DESCRIPTION:Recent years have seen a surge in research on graph representa
 tion learning\, including techniques for deep graph embeddings\, generaliz
 ations of CNNs to graph-structured data\, and neural message-passing appro
 aches. These advances in graph neural networks (GNNs) and related techniqu
 es have led to new state-of-the-art results in numerous domains: chemical 
 synthesis\, vehicle routing\, 3D-vision\, recommender systems\, question a
 nswering\, continuous control\, self-driving and social network analysis. 
 Accordingly\, GNNs regularly top the charts on fastest-growing trends and 
 workshops at virtually all top machine learning conferences. \n\nBut\, wha
 t even is a GNN? Quick online searching reveals many different definitions
 . These definitions may drastically differ (or even use entirely different
  terminology) depending on the background that the writer is assuming. And
  this is no coincidence: the concepts that we now attribute to graph neura
 l networks have independently emerged over the past decade(s) from a varie
 ty of machine learning directions.\n\nIn this talk\, I will attempt to pro
 vide a "bird's eye" view on GNNs. Following a quick motivation on the util
 ity of graph representation learning\, I will derive GNNs from first princ
 iples of permutation invariance and equivariance. Through this lens\, I wi
 ll then describe how researchers from various fields (graph embeddings\, g
 raph signal processing\, probabilistic graphical models\, and graph isomor
 phism testing) arrived---independently---at essentially the same concept o
 f a GNN.\n\nThe talk will be geared towards a generic computer science aud
 ience\, though some basic knowledge of machine learning with neural networ
 ks will be useful. I also hope that seasoned GNN practitioners may benefit
  from the categorisation I will present.\n\nThe content is inspired by the
  work of Will Hamilton\, as well as my ongoing work on the categorisation 
 of geometric deep learning\, alongside Joan Bruna\, Michael Bronstein and 
 Taco Cohen.\n\n\nLink to join: https://cl-cam-ac-uk.zoom.us/j/91253900399?
 pwd=SU5TNnpYdDlQbzQ4SEVPVWVWa0Nldz09\n\nA recording of this talk is availa
 ble at the following link: https://www.cl.cam.ac.uk/seminars/wednesday/vid
 eo/
LOCATION:Online
END:VEVENT
END:VCALENDAR