Deep Haar scattering networks

DOI10.1093/IMAIAI/IAW007MaRDI QIDQ4603720zbMATH OpenOpenAlexFDO

Authors Xiuyuan Cheng, Xu Chen, Stéphane Mallat

Publication date 19 February 2018

Published in Information and Inference: A Journal of the IMA (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1509.09187

zbMATH Keywords

deep learninggraphsneural networkHaar waveletscattering transformclassification images

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05)

Abstract: An orthogonal Haar scattering transform is a deep network, computed with a hierarchy of additions, subtractions and absolute values, over pairs of coefficients. It provides a simple mathematical model for unsupervised deep network learning. It implements non-linear contractions, which are optimized for classification, with an unsupervised pair matching algorithm, of polynomial complexity. A structured Haar scattering over graph data computes permutation invariant representations of groups of connected points in the graph. If the graph connectivity is unknown, unsupervised Haar pair learning can provide a consistent estimation of connected dyadic groups of points. Classification results are given on image data bases, defined on regular grids or graphs, with a connectivity which may be known or unknown.

Recommendations

• Graph convolutional neural networks via scattering
• The Scattering Transform
• Multi-link wavelets on hierarchical graphs
• scientific article; zbMATH DE number 7234683
• Group invariant scattering

Cites work

• scientific article; zbMATH DE number 3980111 (Why is no real title available?)
• scientific article; zbMATH DE number 1304326 (Why is no real title available?)
• A Fast Learning Algorithm for Deep Belief Nets
• An Efficient Implementation of Edmonds' Algorithm for Maximum Matching on Graphs
• Diffusion wavelets
• Group invariant scattering
• Harmonic analysis of digital data bases
• Paths, Trees, and Flowers
• Reducing the Dimensionality of Data with Neural Networks
• Some deviation inequalities

Cited in

(16)

• Fast Haar transforms for graph neural networks
• Graph transform learning
• Phase harmonic correlations and convolutional neural networks
• On the convergence of formally diverging neural net-based classifiers
• Stability for the training of deep neural networks and other classifiers
• Asymptotic analysis of higher-order scattering transform of Gaussian processes
• The unreasonable effectiveness of Haar frames
• Graph Fourier transform based on \(\ell_1\) norm variation minimization
• Regularity properties of Haar frames
• Understanding Graph Neural Networks with Generalized Geometric Scattering Transforms
• scientific article; zbMATH DE number 7234683 (Why is no real title available?)
• Rotationally invariant time-frequency scattering transforms
• Analysis of time-frequency scattering transforms
• GSNs: generative stochastic networks
• Continuous Generative Neural Networks: A Wavelet-Based Architecture in Function Spaces
• Central and Noncentral Limit Theorems Arising from the Scattering Transform and Its Neural Activation Generalization

Describes a project that uses

Uses Software

• DeepFace
• darch

This page was built for publication: Deep Haar scattering networks

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4603720)