Circular-symmetric correlation layer

DOI10.1007/S10994-022-06288-4MaRDI QIDQ6097162zbMATH OpenFDO

Authors Bahar Azari, Deniz Erdogmus

Publication date 12 June 2023

Published in Machine Learning (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2107.12480

zbMATH Keywords

fast Fourier transform (FFT)equivariant neural networksLiDAR segmentationpanoramic image processing

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05)

Abstract: Despite the vast success of standard planar convolutional neural networks, they are not the most efficient choice for analyzing signals that lie on an arbitrarily curved manifold, such as a cylinder. The problem arises when one performs a planar projection of these signals and inevitably causes them to be distorted or broken where there is valuable information. We propose a Circular-symmetric Correlation Layer (CCL) based on the formalism of roto-translation equivariant correlation on the continuous group $S^{1}imesmathbbR$, and implement it efficiently using the well-known Fast Fourier Transform (FFT) algorithm. We showcase the performance analysis of a general network equipped with CCL on various recognition and classification tasks and datasets. The PyTorch package implementation of CCL is provided online.

Recommendations

• Homogeneous vector bundles and \(G\)-equivariant convolutional neural networks
• \(\mathrm{SU}(1,1)\) equivariant neural networks and application to robust Toeplitz Hermitian positive definite matrix classification
• Parallel Transport Convolution: Deformable Convolutional Networks on Manifold-Structured Data
• Representation learning on unit ball with 3D roto-translational equivariance
• Equivariant neural networks for inverse problems

Cites work

• A fast cosine transform in one and two dimensions

Cited in

(1)

• SepNet: a neural network for directionally correlated data

This page was built for publication: Circular-symmetric correlation layer

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6097162)