From CNNs to Shift-Invariant Twin Models Based on Complex Wavelets

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Leterme, Hubert, Polisano, Kévin, Perrier, Valérie, Alahari, Karteek
Format: Preprint
Published: 2022
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916267010555904
author Leterme, Hubert
Polisano, Kévin
Perrier, Valérie
Alahari, Karteek
author_facet Leterme, Hubert
Polisano, Kévin
Perrier, Valérie
Alahari, Karteek
contents We propose a novel method to increase shift invariance and prediction accuracy in convolutional neural networks. Specifically, we replace the first-layer combination "real-valued convolutions + max pooling" (RMax) by "complex-valued convolutions + modulus" (CMod), which is stable to translations, or shifts. To justify our approach, we claim that CMod and RMax produce comparable outputs when the convolution kernel is band-pass and oriented (Gabor-like filter). In this context, CMod can therefore be considered as a stable alternative to RMax. To enforce this property, we constrain the convolution kernels to adopt such a Gabor-like structure. The corresponding architecture is called mathematical twin, because it employs a well-defined mathematical operator to mimic the behavior of the original, freely-trained model. Our approach achieves superior accuracy on ImageNet and CIFAR-10 classification tasks, compared to prior methods based on low-pass filtering. Arguably, our approach's emphasis on retaining high-frequency details contributes to a better balance between shift invariance and information preservation, resulting in improved performance. Furthermore, it has a lower computational cost and memory footprint than concurrent work, making it a promising solution for practical implementation.
format Preprint
id arxiv_https___arxiv_org_abs_2212_00394
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle From CNNs to Shift-Invariant Twin Models Based on Complex Wavelets
Leterme, Hubert
Polisano, Kévin
Perrier, Valérie
Alahari, Karteek
Computer Vision and Pattern Recognition
Artificial Intelligence
Image and Video Processing
Machine Learning
We propose a novel method to increase shift invariance and prediction accuracy in convolutional neural networks. Specifically, we replace the first-layer combination "real-valued convolutions + max pooling" (RMax) by "complex-valued convolutions + modulus" (CMod), which is stable to translations, or shifts. To justify our approach, we claim that CMod and RMax produce comparable outputs when the convolution kernel is band-pass and oriented (Gabor-like filter). In this context, CMod can therefore be considered as a stable alternative to RMax. To enforce this property, we constrain the convolution kernels to adopt such a Gabor-like structure. The corresponding architecture is called mathematical twin, because it employs a well-defined mathematical operator to mimic the behavior of the original, freely-trained model. Our approach achieves superior accuracy on ImageNet and CIFAR-10 classification tasks, compared to prior methods based on low-pass filtering. Arguably, our approach's emphasis on retaining high-frequency details contributes to a better balance between shift invariance and information preservation, resulting in improved performance. Furthermore, it has a lower computational cost and memory footprint than concurrent work, making it a promising solution for practical implementation.
title From CNNs to Shift-Invariant Twin Models Based on Complex Wavelets
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Image and Video Processing
Machine Learning
url https://arxiv.org/abs/2212.00394