Higher Order Gauge Equivariant CNNs on Riemannian Manifolds and Applications

by   Gianfranco Cortes, et al.

With the advent of group equivariant convolutions in deep networks literature, spherical CNNs with ๐–ฒ๐–ฎ(3)-equivariant layers have been developed to cope with data that are samples of signals on the sphere S^2. One can implicitly obtain ๐–ฒ๐–ฎ(3)-equivariant convolutions on S^2 with significant efficiency gains by explicitly requiring gauge equivariance w.r.t. ๐–ฒ๐–ฎ(2). In this paper, we build on this fact by introducing a higher order generalization of the gauge equivariant convolution, whose implementation is dubbed a gauge equivariant Volterra network (GEVNet). This allows us to model spatially extended nonlinear interactions within a given receptive field while still maintaining equivariance to global isometries. We prove theoretical results regarding the equivariance and construction of higher order gauge equivariant convolutions. Then, we empirically demonstrate the parameter efficiency of our model, first on computer vision benchmark data (e.g. spherical MNIST), and then in combination with a convolutional kernel network (CKN) on neuroimaging data. In the neuroimaging data experiments, the resulting two-part architecture (CKN + GEVNet) is used to automatically discriminate between patients with Lewy Body Disease (DLB), Alzheimer's Disease (AD) and Parkinson's Disease (PD) from diffusion magnetic resonance images (dMRI). The GEVNet extracts micro-architectural features within each voxel, while the CKN extracts macro-architectural features across voxels. This compound architecture is uniquely poised to exploit the intra- and inter-voxel information contained in the dMRI data, leading to improved performance over the classification results obtained from either of the individual components.


page 1

page 2

page 3

page 4

โˆ™ 06/08/2023

Scaling Spherical CNNs

Spherical CNNs generalize CNNs to functions on the sphere, by using sphe...
โˆ™ 06/18/2020

Spin-Weighted Spherical CNNs

Learning equivariant representations is a promising way to reduce sample...
โˆ™ 01/28/2022

Mรถbius Convolutions for Spherical CNNs

Mรถbius transformations play an important role in both geometry and spher...
โˆ™ 05/14/2018

H-CNNs: Convolutional Neural Networks for Riemannian Homogeneous Spaces

Convolutional neural networks are ubiquitous in Machine Learning applica...
โˆ™ 12/19/2018

Discriminative analysis of the human cortex using spherical CNNs - a study on Alzheimer's disease diagnosis

In neuroimaging studies, the human cortex is commonly modeled as a spher...
โˆ™ 11/27/2019

Orthogonal Convolutional Neural Networks

The instability and feature redundancy in CNNs hinders further performan...
โˆ™ 07/05/2020

Simplicial Complex based Point Correspondence between Images warped onto Manifolds

Recent increase in the availability of warped images projected onto a ma...

Please sign up or login with your details

Forgot password? Click here to reset