BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Pre-training Molecular Graph Representation with 3D Geometry - Han
 chen Wang\, Cambridge
DTSTART:20220627T130000Z
DTEND:20220627T133000Z
UID:TALK175970@talks.cam.ac.uk
CONTACT:Dr Christoph Schran
DESCRIPTION:Molecular graph representation learning is a fundamental probl
 em in modern drug and material discovery. Molecular graphs are typically m
 odeled by their 2D topological structures\, but it has been recently disco
 vered that 3D geometric information plays a more vital role in predicting 
 molecular functionalities. However\, the lack of 3D information in real-wo
 rld scenarios has significantly impeded the learning of geometric graph re
 presentation. To cope with this challenge\, we propose the Graph Multi-Vie
 w Pre-training (GraphMVP) framework where self-supervised learning (SSL) i
 s performed by leveraging the correspondence and consistency between 2D to
 pological structures and 3D geometric views. GraphMVP effectively learns a
  2D molecular graph encoder that is enhanced by richer and more discrimina
 tive 3D geometry. We further provide theoretical insights to justify the e
 ffectiveness of GraphMVP. Finally\, comprehensive experiments show that Gr
 aphMVP can consistently outperform existing graph SSL methods.
LOCATION:https://zoom.us/j/92447982065?pwd=RkhaYkM5VTZPZ3pYSHptUXlRSkppQT0
 9
END:VEVENT
END:VCALENDAR