Invariant Scattering Convolution Networks

Nov 08, 2019 - 3:00 pm to 4:00 pm

SLAC, Kavli 3rd Floor Conf. Room

Sean McLaughlin (Stanford)

Join us Friday Nov 8th at 3pm in the SLAC 3rd Floor Conference Room (and on zoom) for the Stats & ML Journal Club. 

This week, I, Sean McLaughlin will lead a discussion on Scattering Wavelets, a unique type of transform that appears to naturally represent the complex underlying signals in images, and may be a more natural basis for comparing many types of scientific images. See the paper and abstract below! 




Title: Invariant Scattering Convolution Networks

Abstract: A wavelet scattering network computes a translation invariant image representation, which is stable to deformations and preserves

high frequency information for classification. It cascades wavelet transform convolutions with non-linear modulus and averaging operators. The
first network layer outputs SIFT-type descriptors whereas the next layers
provide complementary invariant information which improves classification. The mathematical analysis of wavelet scattering networks explain
important properties of deep convolution networks for classification.
A scattering representation of stationary processes incorporates
higher order moments and can thus discriminate textures having same
Fourier power spectrum. State of the art classification results are obtained for handwritten digits and texture discrimination, with a Gaussian
kernel SVM and a generative PCA classifier.