BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Learning and Extrapolation in Graph Neural Networks - Stefanie Jeg
 elka (Massachusetts Institute of Technology)
DTSTART:20211123T120000Z
DTEND:20211123T130000Z
UID:TALK164854@talks.cam.ac.uk
DESCRIPTION:Graph Neural Networks (GNNs) have become a popular tool for le
 arning representations of graph-structured inputs\, with applications in c
 omputational chemistry\, recommendation\, pharmacy\, reasoning\, and many 
 other areas. In this talk\, I will show some recent results on learning wi
 th message-passing GNNs. In particular\, GNNs possess important invariance
 s and inductive biases that affect learning and generalization. Studying t
 he effect of these inductive biases can be challenging\, as they are affec
 ted by the architecture (structure and aggregation functions) and training
  algorithm and interplay with data and learning task. In particular\, we s
 tudy these biases for learning structured tasks\, e.g.\, simulations or al
 gorithms\, and show how for such tasks\, architecture choices affect gener
 alization within and outside the training distribution.\nThis talk is base
 d on joint work with Keyulu Xu\, Jingling Li\, Mozhi Zhang\, Simon S. Du a
 nd Ken-ichi Kawarabayashi.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR