BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Abstract Diagrammatic Reasoning with Multiplex Graph Networks - Du
 o Wang
DTSTART:20200121T130000Z
DTEND:20200121T140000Z
UID:TALK132424@talks.cam.ac.uk
CONTACT:Mateja Jamnik
DESCRIPTION:Abstract reasoning\, particularly in the visual domain\, is a 
 complex human ability\, but it remains a challenging problem for artificia
 l neural learning systems. In this work we propose MXGNet\,\n	a multilayer
  graph neural network for multi-panel diagrammatic reasoning tasks. MXGNet
  combines three powerful concepts\, namely\, object-level representation\,
  graph neural networks and multiplex graphs\, for solving visual reasoning
  tasks. MXGNet first extracts\n	object-level representations for each elem
 ent in all panels of the diagrams\, and then forms a multi-layer multiplex
  graph capturing multiple relations between objects across different diagr
 am panels. MXGNet summarises the multiple graphs extracted from the diagra
 ms of the task\, and uses this summarisation to pick the most probable ans
 wer from the given candidates. We have tested MXGNet on two types of diagr
 ammatic reasoning tasks\, namely Diagram Syllogisms and Raven Progressive 
 Matrices (RPM). For an Euler Diagram Syllogism task MXGNet achieves state-
 of-the-art accuracy of 99.8%.\nFor PGM and RAVEN\, two comprehensive datas
 ets for RPM reasoning\, MXGNet outperforms the state-of-the-art models by 
 a considerable margin.\n\nThis work will be presented at ICLR.
LOCATION:SS03\, Computer Laboratory\, William Gates Building
END:VEVENT
END:VCALENDAR