BEGIN:VCALENDAR
VERSION:2.0
PRODID:researchseminars.org
CALSCALE:GREGORIAN
X-WR-CALNAME:researchseminars.org
BEGIN:VEVENT
SUMMARY:Francesco Di Giovanni (Twitter - UK)
DTSTART:20221209T160000Z
DTEND:20221209T170000Z
DTSTAMP:20260423T022921Z
UID:GEOTOP-A/30
DESCRIPTION:Title: <a href="https://researchseminars.org/talk/GEOTOP-A/30/
 ">Over-squashing and over-smoothing through the lenses of curvature and mu
 lti-particle dynamics</a>\nby Francesco Di Giovanni (Twitter - UK) as part
  of GEOTOP-A seminar\n\n\nAbstract\nI am going to talk about two problems 
 that Message Passing Neural Networks (MPNNs) have been shown to be struggl
 ing from. The first one – known as over-squashing – is unavoidable in 
 the MPNN class and concerns the input graph topology. This relates to how 
 information propagates in a graph. We show that discrete curvature quantit
 ies (old and new) could help us understand where messages are being lost a
 nd we can provably characterize the over-squashing phenomenon in terms of 
 curvature. The second problem consists in analysing GNNs as multi-particle
  dynamics using the lens of gradient flows of an energy. We investigate wh
 at happens when instead of learning the MPNN equations we learn an energy 
 and then let the equations follow the gradient flow of such energy. This a
 llows us to understand further the role of the channel-mixing matrix that 
 is ubiquitous in standard graph convolutional models as a bilinear potenti
 al inducing both attraction and repulsion along edges via its positive and
  negative eigenvalues respectively.\n
LOCATION:https://researchseminars.org/talk/GEOTOP-A/30/
END:VEVENT
END:VCALENDAR