A Mathematical Theory of Deep Convolutional Neural Networks for Feature Extraction

TL;DR

This paper develops a comprehensive mathematical framework for deep convolutional neural networks, analyzing their feature extraction capabilities, translation invariance, and deformation stability across various transforms, nonlinearities, and pooling methods.

Contribution

It extends Mallat's scattering network analysis to general convolutional transforms, nonlinearities, and pooling, providing new theoretical insights into deep CNNs' invariance and stability properties.

Findings

Proves translation invariance increases with network depth.

Establishes deformation sensitivity bounds for various signal classes.

Applies to a wide range of convolutional transforms and nonlinearities.

Abstract

Deep convolutional neural networks have led to breakthrough results in numerous practical machine learning tasks such as classification of images in the ImageNet data set, control-policy-learning to play Atari games or the board game Go, and image captioning. Many of these applications first perform feature extraction and then feed the results thereof into a trainable classifier. The mathematical analysis of deep convolutional neural networks for feature extraction was initiated by Mallat, 2012. Specifically, Mallat considered so-called scattering networks based on a wavelet transform followed by the modulus non-linearity in each network layer, and proved translation invariance (asymptotically in the wavelet scale parameter) and deformation stability of the corresponding feature extractor. This paper complements Mallat's results by developing a theory that encompasses general convolutional transforms, or in more technical parlance, general semi-discrete frames (including Weyl-Heisenberg filters, curvelets, shearlets, ridgelets, wavelets, and learned filters), general Lipschitz-continuous non-linearities (e.g., rectified linear units, shifted logistic sigmoids, hyperbolic tangents, and modulus functions), and general Lipschitz-continuous pooling operators emulating, e.g., sub-sampling and averaging. In addition, all of these elements can be different in different network layers. For the resulting feature extractor we prove a translation invariance result of vertical nature in the sense of the features becoming progressively more translation-invariant with increasing network depth, and we establish deformation sensitivity bounds that apply to signal classes such as, e.g., band-limited functions, cartoon functions, and Lipschitz functions.