Interferometric measurements of the 21cm signal are a prime example of the data-driven era in astrophysics we are entering with current and upcoming experiments. We showcase the use of deep networks that are tailored for the structure of 3D tomographic 21cm light-cones to firstly detect and characterise HI sources and to secondly directly infer global astrophysical and cosmological model parameters. We compare different architectures and highlight how 3D CNN architectures that mirror the data structure are the best-performing model.
Cite
@article{arxiv.2311.17553,
title = {Deep Learning 21cm Lightcones in 3D},
author = {Caroline Heneka},
journal= {arXiv preprint arXiv:2311.17553},
year = {2023}
}
Comments
5 pages, 2 figures. This is a preprint of the following chapter: Heneka, C., Deep Learning 21 cm Lightcones in 3D, published in Machine Learning for Astrophysics, ML4Astro 2022, edited by Bufano, F., Riggi, S., Sciacca, E., Schilliro, F., 2023, Springer, Cham. The final authenticated version is available online at: https://doi.org/10.1007/978-3-031-34167-0_34