On the Shift Invariance of Max Pooling Feature Maps in Convolutional Neural Networks - Archive ouverte HAL Access content directly
Preprints, Working Papers, ... Year : 2023

On the Shift Invariance of Max Pooling Feature Maps in Convolutional Neural Networks

Abstract

This paper focuses on improving the mathematical interpretability of convolutional neural networks (CNNs) in the context of image classification. Specifically, we tackle the instability issue arising in their first layer, which tends to learn parameters that closely resemble oriented band-pass filters when trained on datasets like ImageNet. Subsampled convolutions with such Gabor-like filters are prone to aliasing, causing sensitivity to small input shifts. In this context, we establish conditions under which the max pooling operator approximates a complex modulus, which is nearly shift invariant. We then derive a measure of shift invariance for subsampled convolutions followed by max pooling. In particular, we highlight the crucial role played by the filter's frequency and orientation in achieving stability. We experimentally validate our theory by considering a deterministic feature extractor based on the dual-tree complex wavelet packet transform, a particular case of discrete Gabor-like decomposition.
Fichier principal
Vignette du fichier
preprint.pdf (1.71 Mo) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-03779434 , version 1 (16-09-2022)
hal-03779434 , version 2 (24-10-2023)

Licence

Identifiers

  • HAL Id : hal-03779434 , version 2

Cite

Hubert Leterme, Kévin Polisano, Valérie Perrier, Karteek Alahari. On the Shift Invariance of Max Pooling Feature Maps in Convolutional Neural Networks. 2023. ⟨hal-03779434v2⟩
194 View
69 Download

Share

Gmail Mastodon Facebook X LinkedIn More