Analysis of (sub-)Riemannian PDE-G-CNNs

DOI10.1007/S10851-023-01147-WMaRDI QIDQ6186783zbMATH OpenOpenAlexFDO

Authors Gijs Bellaard, Daan L. J. Bon, Gautam Pai, Bart M. N. Smets, Remco Duits

Publication date 10 January 2024

Published in Journal of Mathematical Imaging and Vision (Search for Journal in Brave)

Copyright license Creative Commons Attribution 4.0 International

Full work available at URL https://arxiv.org/abs/2210.00935

zbMATH Keywords

Riemannian geometrysub-Riemannian geometryconvolutional neural networksPDEsgeometric deep learningmorphological convolutionsscale space theory

Mathematics Subject Classification ID

Artificial neural networks and deep learning (68T07) PDEs in connection with computer science (35Q68)

Abstract: Group equivariant convolutional neural networks (G-CNNs) have been successfully applied in geometric deep learning. Typically, G-CNNs have the advantage over CNNs that they do not waste network capacity on training symmetries that should have been hard-coded in the network. The recently introduced framework of PDE-based G-CNNs (PDE-G-CNNs) generalises G-CNNs. PDE-G-CNNs have the core advantages that they simultaneously 1) reduce network complexity, 2) increase classification performance, and 3) provide geometric interpretability. Their implementations primarily consist of linear and morphological convolutions with kernels. In this paper we show that the previously suggested approximative morphological kernels do not always accurately approximate the exact kernels accurately. More specifically, depending on the spatial anisotropy of the Riemannian metric, we argue that one must resort to sub-Riemannian approximations. We solve this problem by providing a new approximative kernel that works regardless of the anisotropy. We provide new theorems with better error estimates of the approximative kernels, and prove that they all carry the same reflectional symmetries as the exact ones. We test the effectiveness of multiple approximative kernels within the PDE-G-CNN framework on two datasets, and observe an improvement with the new approximative kernels. We report that the PDE-G-CNNs again allow for a considerable reduction of network complexity while having comparable or better performance than G-CNNs and CNNs on the two datasets. Moreover, PDE-G-CNNs have the advantage of better geometric interpretability over G-CNNs, as the morphological kernels are related to association fields from neurogeometry.

Recommendations

• Equivariant deep learning via morphological and linear scale space PDEs on the space of positions and orientations
• Homogeneous vector bundles and \(G\)-equivariant convolutional neural networks
• Functional properties of PDE-based group equivariant convolutional neural networks
• Analysis of deep convolutional networks from group theory viewpoint

Cites work

• scientific article; zbMATH DE number 1739749 (Why is no real title available?)
• scientific article; zbMATH DE number 5681750 (Why is no real title available?)
• A PDE approach to data-driven sub-Riemannian geodesics in \(\mathrm{SE}(2)\)
• A cortical based model of perceptual completion in the roto-translation space
• A geometric model of multi-scale orientation preference maps via Gabor functions
• Association fields via cuspless sub-Riemannian geodesics in SE(2)
• Cut locus and optimal synthesis in the sub-Riemannian problem on the group of motions of a plane
• Elements of neurogeometry. Functional architectures of vision. Translated from the French by the author
• Equivariant deep learning via morphological and linear scale space PDEs on the space of positions and orientations
• Extension of mathematical morphology in Riemannian spaces
• Geometrical optical illusion via sub-Riemannian geodesics in the roto-translation group
• Left-invariant parabolic evolutions on \(SE(2)\) and contour enhancement via invertible orientation scores. I: Linear left-invariant diffusion equations on \(SE(2)\)
• Maxwell strata in sub-Riemannian problem on the group of motions of a plane
• Morphological and linear scale spaces for fiber enhancement in DW-MRI
• Morphological counterparts of linear shift-invariant scale-spaces
• Nilpotent approximations of sub-Riemannian distances for fast perceptual grouping of blood vessels in 2D and 3D
• Nonsmooth analysis and Hamilton--Jacobi equations on Riemannian manifolds
• Optimal paths for variants of the 2D and 3D Reeds-Shepp car with applications in image analysis
• PDE-based group equivariant convolutional neural networks
• Weak KAM theorem on non compact manifolds
• Weighted subcoercive operators on Lie groups

Cited in

(1)

• Functional properties of PDE-based group equivariant convolutional neural networks

This page was built for publication: Analysis of (sub-)Riemannian PDE-G-CNNs

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6186783)