DeepSphere: Graph-based spherical convolutional neural network for cosmology

Discussion topics for individual codes
Post Reply
Ada Coda
ASCL Robot
Posts: 1615
Joined: Thu May 08, 2014 5:37 am

DeepSphere: Graph-based spherical convolutional neural network for cosmology

Post by Ada Coda » Mon Jun 29, 2020 9:47 pm

DeepSphere: Graph-based spherical convolutional neural network for cosmology

Abstract: DeepSphere implements a generalization of Convolutional Neural Networks (CNNs) to the sphere. It models the discretized sphere as a graph of connected pixels. The resulting convolution is more efficient (especially when data doesn't span the whole sphere) and mostly equivariant to rotation (small distortions are due to the non-existence of a regular sampling of the sphere). The pooling strategy exploits a hierarchical pixelization of the sphere (HEALPix) to analyze the data at multiple scales. The graph neural network model is based on ChebNet and its TensorFlow implementation.

Credit: Perraudin, Nathanaël; Defferrard, Michaël; Kacprzak, Tomasz; Sgier, Raphael

Site: https://github.com/deepsphere/DeepSphere
https://ui.adsabs.harvard.edu/abs/2019arXiv190405146D

Bibcode: 2020ascl.soft06008P

Preferred citation method: https://ui.adsabs.harvard.edu/abs/2019A%26C....27..130P and https://ui.adsabs.harvard.edu/abs/2019arXiv190405146D

ID: ascl:2006.008

Post Reply