TY - GEN
T1 - A Hybrid Scattering Transform for Signals with Isolated Singularities
AU - Perlmutter, Michael
AU - He, Jieqian
AU - Iwen, Mark
AU - Hirn, Matthew
N1 - Publisher Copyright:
© 2021 IEEE.
PY - 2021
Y1 - 2021
N2 - The scattering transform is a wavelet-based model of Convolutional Neural Networks originally introduced by S. Mallat. Mallat's analysis shows that this network has desirable stability and invariance guarantees and therefore helps explain the observation that the filters learned by early layers of a Convolutional Neural Network typically resemble wavelets. Our aim is to understand what sort of filters should be used in the later layers of the network. Towards this end, we propose a two-layer hybrid scattering transform. In our first layer, we convolve the input signal with a wavelet filter transform to promote sparsity, and, in the second layer, we convolve with a Gabor filter to leverage the sparsity created by the first layer. We show that these measurements characterize information about signals with isolated singularities. We also show that the Gabor measurements used in the second layer can be used to synthesize sparse signals such as those produced by the first layer.
AB - The scattering transform is a wavelet-based model of Convolutional Neural Networks originally introduced by S. Mallat. Mallat's analysis shows that this network has desirable stability and invariance guarantees and therefore helps explain the observation that the filters learned by early layers of a Convolutional Neural Network typically resemble wavelets. Our aim is to understand what sort of filters should be used in the later layers of the network. Towards this end, we propose a two-layer hybrid scattering transform. In our first layer, we convolve the input signal with a wavelet filter transform to promote sparsity, and, in the second layer, we convolve with a Gabor filter to leverage the sparsity created by the first layer. We show that these measurements characterize information about signals with isolated singularities. We also show that the Gabor measurements used in the second layer can be used to synthesize sparse signals such as those produced by the first layer.
KW - deep learning
KW - scattering transforms
KW - sparsity
KW - time-frequency analysis
KW - wavelets
UR - http://www.scopus.com/inward/record.url?scp=85127020577&partnerID=8YFLogxK
U2 - 10.1109/IEEECONF53345.2021.9723364
DO - 10.1109/IEEECONF53345.2021.9723364
M3 - Conference contribution
AN - SCOPUS:85127020577
T3 - Conference Record - Asilomar Conference on Signals, Systems and Computers
SP - 1322
EP - 1329
BT - 55th Asilomar Conference on Signals, Systems and Computers, ACSSC 2021
A2 - Matthews, Michael B.
PB - IEEE Computer Society
T2 - 55th Asilomar Conference on Signals, Systems and Computers, ACSSC 2021
Y2 - 31 October 2021 through 3 November 2021
ER -