Spherical convergence maps dataset

Raphaël Sgier, T. Kacprzak, Nathanaël Perraudin, Michaël DefferrardZenodo 2018 · DOI · OpenAlex

Abstract

This dataset is composed of 60 spherical convergence maps (30 per class), separated into a training set of 40 maps and a testing set of 20 maps. As a preprocessing, we recommend to remove the mean of each map and to smooth them with a Gaussian symmetric beam of 3 arcmins (sphtfunc.smoothing in Healpy). This dataset was created by R. Sgier and T. Kacprzak based on the work in paper by Sgier R. et al. 2018, "Fast Generation of Covariance Matrices for Weak Lensing", arxiv.org/abs/1801.05745. This dataset is used in "DeepSphere: Efficient spherical Convolutional Neural Network with HEALPix sampling for cosmological applications" arxiv.org/abs/1810.12186.
BibTeX
@inproceedings{sgier2018spherical,
  author = {Raphaël Sgier and T. Kacprzak and Nathanaël Perraudin and Michaël Defferrard},
  title = {{Spherical convergence maps dataset}},
  booktitle = {Zenodo},
  year = {2018},
  doi = {10.5281/zenodo.1303272},
}

Loading related content...

Figure