Invariant scattering convolution networks

Joan Bruna Estrach, Stephane Mallat

Research output: Contribution to journalArticle

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 nonlinear modulus and averaging operators. The first network layer outputs SIFT-type descriptors, whereas the next layers provide complementary invariant information that improves classification. The mathematical analysis of wavelet scattering networks explains important properties of deep convolution networks for classification. A scattering representation of stationary processes incorporates higher order moments and can thus discriminate textures having the 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.

Original languageEnglish (US)
Article number6522407
Pages (from-to)1872-1886
Number of pages15
JournalIEEE Transactions on Pattern Analysis and Machine Intelligence
Volume35
Issue number8
DOIs
StatePublished - 2013

Fingerprint

Convolution
Scattering
Invariant
Texture
Wavelets
Textures
Higher Order Moments
Fourier Spectrum
Averaging Operators
Gaussian Kernel
Image Representation
Network layers
Scale Invariant Feature Transform
Stationary Process
Power spectrum
Mathematical Analysis
Digit
Power Spectrum
Wavelet transforms
Wavelet Transform

Keywords

  • Classification
  • convolution networks
  • deformations
  • invariants
  • wavelets

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Vision and Pattern Recognition
  • Software
  • Computational Theory and Mathematics
  • Applied Mathematics

Cite this

Invariant scattering convolution networks. / Bruna Estrach, Joan; Mallat, Stephane.

In: IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 35, No. 8, 6522407, 2013, p. 1872-1886.

Research output: Contribution to journalArticle

Bruna Estrach, Joan ; Mallat, Stephane. / Invariant scattering convolution networks. In: IEEE Transactions on Pattern Analysis and Machine Intelligence. 2013 ; Vol. 35, No. 8. pp. 1872-1886.
@article{5dcb9f488dd74f42a256dee282958a2d,
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 nonlinear modulus and averaging operators. The first network layer outputs SIFT-type descriptors, whereas the next layers provide complementary invariant information that improves classification. The mathematical analysis of wavelet scattering networks explains important properties of deep convolution networks for classification. A scattering representation of stationary processes incorporates higher order moments and can thus discriminate textures having the 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.",
keywords = "Classification, convolution networks, deformations, invariants, wavelets",
author = "{Bruna Estrach}, Joan and Stephane Mallat",
year = "2013",
doi = "10.1109/TPAMI.2012.230",
language = "English (US)",
volume = "35",
pages = "1872--1886",
journal = "IEEE Transactions on Pattern Analysis and Machine Intelligence",
issn = "0162-8828",
publisher = "IEEE Computer Society",
number = "8",

}

TY - JOUR

T1 - Invariant scattering convolution networks

AU - Bruna Estrach, Joan

AU - Mallat, Stephane

PY - 2013

Y1 - 2013

N2 - 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 nonlinear modulus and averaging operators. The first network layer outputs SIFT-type descriptors, whereas the next layers provide complementary invariant information that improves classification. The mathematical analysis of wavelet scattering networks explains important properties of deep convolution networks for classification. A scattering representation of stationary processes incorporates higher order moments and can thus discriminate textures having the 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.

AB - 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 nonlinear modulus and averaging operators. The first network layer outputs SIFT-type descriptors, whereas the next layers provide complementary invariant information that improves classification. The mathematical analysis of wavelet scattering networks explains important properties of deep convolution networks for classification. A scattering representation of stationary processes incorporates higher order moments and can thus discriminate textures having the 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.

KW - Classification

KW - convolution networks

KW - deformations

KW - invariants

KW - wavelets

UR - http://www.scopus.com/inward/record.url?scp=84879877798&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84879877798&partnerID=8YFLogxK

U2 - 10.1109/TPAMI.2012.230

DO - 10.1109/TPAMI.2012.230

M3 - Article

C2 - 23787341

AN - SCOPUS:84879877798

VL - 35

SP - 1872

EP - 1886

JO - IEEE Transactions on Pattern Analysis and Machine Intelligence

JF - IEEE Transactions on Pattern Analysis and Machine Intelligence

SN - 0162-8828

IS - 8

M1 - 6522407

ER -