BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Approximate Equivariance SO(3) Needlet Convolution - Kai Yi\, Univ
 ersity of New South Wales (UNSW) in Sydney
DTSTART:20221130T170000Z
DTEND:20221130T180000Z
UID:TALK193115@talks.cam.ac.uk
CONTACT:Pietro Lio
DESCRIPTION:This paper develops a rotation-invariant needlet convolution
  for rotation group SO(3) to distill multiscale information of spherical s
 ignals.\nThe spherical needlet transform is generalized from $\\sS^2$ on
 to the SO(3) group\, which decomposes a spherical signal to approximate an
 d detailed spectral coefficients by a set of tight framelet operators. The
  spherical signal during the decomposition and reconstruction achieves rot
 ation invariance. \nBased on needlet transforms\, we form a Needlet a
 pproximate Equivariance Spherical CNN (NES) with multiple SO(3) needlet 
 convolutional layers. The network establishes a powerful tool to extract g
 eometric-invariant features of spherical signals. \nThe model allows suff
 icient network scalability with multi-resolution representation. A robust 
 signal embedding is learned with wavelet shrinkage activation function\, w
 hich filters out redundant high-pass representation while maintaining appr
 oximate rotation invariance. \nThe NES achieves state-of-the-art performa
 nce for quantum chemistry regression and Cosmic Microwave Background (CMB)
  delensing reconstruction\, which shows great potential for solving scient
 ific challenges with high-resolution and multi-scale spherical signal repr
 esentation.\n\n \n\n*Bio:*\n\nKai Yi is a third-year Ph.D. student majori
 ng in mathematics at the School of Mathematics and Statistics in the Unive
 rsity of New South Wales (UNSW) in Sydney\, Australia. His research intere
 sts lie in geometric deep learning and Bayesian statistics. He has applied
  geometric deep learning methods to estimating gravitational lensing param
 eters in cosmology and used VAE for inpainting CMB maps.
LOCATION:Lecture Theatre 2
END:VEVENT
END:VCALENDAR
