The doi above is for this specific version of this dataset, which is currently the latest. Newer versions may be published in the future. For a link that will always point to the latest version, please use
doi: 10.4121/96acc880-ac70-455e-af1e-2343567aa8bf

Attila Lengyel; Edixhoven, Tom; van Gemert, Jan (2023): Code underlying the publication: Using and Abusing Equivariance. Version 1. 4TU.ResearchData. software. https://doi.org/10.4121/96acc880-ac70-455e-af1e-2343567aa8bf.v1

Software

• Artificial Intelligence and Image Processing
• Information and Computing Sciences

computer vision, data efficient deep learning, equivariance, equivariant convolutions

Apache-2.0

RefWorks, BibTeX, Reference Manager, Endnote, DataCite, NLM, DC, CFF

by Attila Lengyel , Tom Edixhoven, Jan van Gemert

Code corresponding to ICCVw 2023 conference workshop paper "Using and Abusing Equivariance".

Abstract

In this paper we show how Group Equivariant Convolutional Neural Networks use subsampling to learn to break equivariance to their symmetries. We focus on 2D rotations and reflections and investigate the impact of broken equivariance on network performance. We show that a change in the input dimension of a network as small as a single pixel can be enough for commonly used architectures to become approximately equivariant, rather than exactly. We investigate the impact of networks not being exactly equivariant and find that approximately equivariant networks generalise significantly worse to unseen symmetries compared to their exactly equivariant counterparts. However, when the symmetries in the training data are not identical to the symmetries of the network, we find that approximately equivariant networks are able to relax their own equivariant constraints, causing them to match or outperform exactly equivariant networks on common benchmark datasets.

• 2023-11-29 first online, published, posted

4TU.ResearchData

• https://research.tudelft.nl/en/publications/using-and-abusing-equivariance

• Tabula Inscripta: Prior knowledge for deep learning (grant code VI.Vidi.192.100) [more info...] Dutch Research Council

TU Delft, TU Delft, Faculty of Electrical Engineering, Mathematics and Computer Science, Department of Intelligent Systems, Computer Vision Lab