BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:ACMP: Allen-Cahn Message Passing with Attractive and Repulsive For
 ces for Graph Neural Networks - Yu Guang Wang
DTSTART:20221025T120000Z
DTEND:20221025T130000Z
UID:TALK176777@talks.cam.ac.uk
CONTACT:Pietro Lio
DESCRIPTION:Neural message passing is a basic feature extraction unit for 
 graph-structured data considering neighboring node features in network pro
 pagation from one layer to the next. We model such a process by an interac
 ting particle system with attractive and repulsive forces and the Allen-Ca
 hn force arising in the modeling of phase transition. The dynamics of the 
 system is a reaction-diffusion process which can separate particles withou
 t blowing up. This induces an Allen-Cahn message passing (ACMP) for graph 
 neural networks where the numerical iteration for the particle system solu
 tion constitutes the message passing propagation. ACMP which has a simple 
 implementation with a neural ODE solver can propel the network depth up to
  one hundred of layers with theoretically proven strictly positive lower b
 ound of the Dirichlet energy. It thus provides a deep model of GNNs circum
 venting the common GNN problem of oversmoothing. GNNs with ACMP achieve st
 ate of the art performance for real-world node classification tasks on bot
 h homophilic and heterophilic datasets.\n\nJoint with Yuelin Wang (SJTU)\,
  Kai Yi (UNSW)\, Xinliang Liu (KAUST) and Shi Jin (SJTU).
LOCATION:Lecture Theatre 2
END:VEVENT
END:VCALENDAR