Scattering Networks on the Sphere for Scalable and Rotationally Equivariant Spherical CNNs
arXiv:2102.02828
Abstract
Convolutional neural networks (CNNs) constructed natively on the sphere have been developed recently and shown to be highly effective for the analysis of spherical data. While an efficient framework has been formulated, spherical CNNs are nevertheless highly computationally demanding; typically they cannot scale beyond spherical signals of thousands of pixels. We develop scattering networks constructed natively on the sphere that provide a powerful representational space for spherical data. Spherical scattering networks are computationally scalable and exhibit rotational equivariance, while their representational space is invariant to isometries and provides efficient and stable signal representations. By integrating scattering networks as an additional type of layer in the generalized spherical CNN framework, we show how they can be leveraged to scale spherical CNNs to the high-resolution data typical of many practical applications, with spherical signals of many tens of megapixels and beyond.
arXiv comment18 pages, 6 figures, accepted by ICLR, code at https://www.kagenova.com/products/fourpiAI/
References in corpus (30)
- Understanding Deep Convolutional Networks
- Spherical Needlets for CMB Data Analysis
- DeepSphere: Efficient spherical Convolutional Neural Network with HEALPix sampling for cosmological applications
- A novel sampling theorem on the sphere
- Wavelets, ridgelets and curvelets on the sphere
- Spherical CNNs
- Exact reconstruction with directional wavelets on the sphere
- Joint Time-Frequency Scattering
- Fast directional continuous spherical wavelet transform algorithms
- Spherical CNNs on Unstructured Grids
- Exact Wavelets on the Ball
- S2LET: A code to perform fast wavelet analysis on the sphere
- Convolutional Networks for Spherical Signals
- Localisation of directional scale-discretised wavelets on the sphere
- Simulating full-sky interferometric observations
- Stability of Graph Scattering Transforms
- Diffusion Scattering Transforms on Graphs
- On the computation of directional scale-discretized wavelet transforms on the sphere
- Directional spin wavelets on the sphere
- Spherical Bayesian mass-mapping with uncertainties: full sky observations on the celestial sphere
- A novel sampling theorem on the rotation group
- A directional continuous wavelet transform on the sphere
- SILC: a new Planck Internal Linear Combination CMB temperature map using directional wavelets
- Wavelet reconstruction of E and B modes for CMB polarisation and cosmic shear analyses
- Polarized wavelets and curvelets on the sphere
- A Simple Proposal for Radial 3D Needlets
- Spin-SILC: CMB polarisation component separation with spin wavelets
- 3D weak lensing with spin wavelets on the ball
- Fourier-Laguerre transform, convolution and wavelets on the ball
- Spin Needlets for Cosmic Microwave Background Polarization Data Analysis