BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Haar Graph Pooling  - Dr Yu Guang Wang - University of New South W
 ales
DTSTART:20201028T150000Z
DTEND:20201028T160000Z
UID:TALK153532@talks.cam.ac.uk
CONTACT:jo de bono
DESCRIPTION:Deep Graph Neural Networks (GNNs) are useful models for graph 
 classification and graph-based regression tasks. In these tasks\, graph po
 oling is a critical ingredient by which GNNs adapt to input graphs of vary
 ing size and structure. We propose a new graph pooling operation based on 
 compressive Haar transforms --- HaarPooling.\nHaarPooling implements a cas
 cade of pooling operations\; it is computed by following a sequence of clu
 sterings of the input graph. A HaarPooling layer transforms a given input 
 graph to an output graph with a smaller node number and the same feature d
 imension\; the compressive Haar transform filters out fine detail informat
 ion in the Haar wavelet domain. In this way\, all the HaarPooling layers t
 ogether synthesise the features of any given input graph into a feature ve
 ctor of uniform size. Such transforms provide a sparse characterisation of
  the data and preserve the structure information of the input graph. GNNs 
 implemented with standard graph convolution layers\, and HaarPooling layer
 s achieve state-of-the-art performance on diverse graph classification and
  regression problems. \n\nThis talk is based on the joint works with Yanan
  Fan (UNSW)\, Junbin Gao (U Sydney)\, Ming Li (ZJNU)\, Pietro Lio (Cambrid
 ge)\, Zheng Ma (Princeton)\, Guido Montufar (UCLA)\, Xuebin Zheng (U Sydne
 y)\, Bingxin Zhou (U Sydney)\, Xiaosheng Zhuang (CityU HK).\n\n\nThe recor
 ding of the seminar can be found at: \n\nhttps://www.cl.cam.ac.uk/seminars
 /wednesday/video/lt2-201028-wed-1500-153532.html\n\n\nTalk slides availabl
 e to download at\nhttps://web.maths.unsw.edu.au/~yuguangwang/talks/Haar_gr
 aph_pool_Yuguang.pdf\n\n
LOCATION:Online
END:VEVENT
END:VCALENDAR