BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:SCATTERING CONVOLUTION NETWORKS and DUAL-TREE WAVELETS - Prof. Nic
 k Kingsbury\, Signal Processing and Communications Group\, CUED
DTSTART:20140529T130000Z
DTEND:20140529T140000Z
UID:TALK52403@talks.cam.ac.uk
CONTACT:Prof. Ramji Venkataramanan
DESCRIPTION:I will talk about the joint project that I researched with Ant
 on van den Hengel of the University of Adelaide between July and November 
 2013.  The project is based on integrating dual-tree complex wavelets into
  new deep network structures for visual learning tasks\, that have been pr
 oposed by Stephane Mallat of Ecole Polytechnique\, Paris.   We will discus
 s the basic ideas behind Invariant Scattering Networks and show how an eff
 icient implementation can be achieved using dual-tree wavelets.  I will pr
 esent some of our own results from using the algorithm to classify MNIST h
 and-written digits and also digits from the Google Street-View House Numbe
 r dataset.  The ideas behind Scattering Nets are introduced in the followi
 ng paper:\n\nhttp://www.cmap.polytechnique.fr/scattering/pami-final.pdf "I
 nvariant Scattering Convolution Networks"\,\nby Joan Bruna and Stephane Ma
 llat of CMAP\, Ecole Polytechnique\, Palaiseau\, France\, in IEEE Trans. o
 n PAMI\, August 2013.
LOCATION:LR11\, Cambridge University Engineering Department
END:VEVENT
END:VCALENDAR