English

Cosmological super-resolution of the 21-cm signal

Cosmology and Nongalactic Astrophysics 2025-02-04 v1 Instrumentation and Methods for Astrophysics

Abstract

In this study, we train score-based diffusion models to super-resolve gigaparsec-scale cosmological simulations of the 21-cm signal. We examine the impact of network and training dataset size on model performance, demonstrating that a single simulation is sufficient for a model to learn the super-resolution task regardless of the initial conditions. Our best-performing model achieves pixelwise RMSE0.57 mK\mathrm{RMSE}\sim0.57\ \mathrm{mK} and dimensionless power spectrum residuals ranging from 102101 mK210^{-2}-10^{-1}\ \mathrm{mK^2} for 1283128^3, 2563256^3 and 5123512^3 voxel simulation volumes at redshift 1010. The super-resolution network ultimately allows us to utilize all spatial scales covered by the SKA1-Low instrument, and could in future be employed to help constrain the astrophysics of the early Universe.

Keywords

Cite

@article{arxiv.2502.00852,
  title  = {Cosmological super-resolution of the 21-cm signal},
  author = {Simon Pochinda and Jiten Dhandha and Anastasia Fialkov and Eloy de Lera Acedo},
  journal= {arXiv preprint arXiv:2502.00852},
  year   = {2025}
}

Comments

10 pages, 2 figures, accepted submission for the Machine Learning and the Physical Sciences Workshop at the 38th conference on Neural Information Processing Systems (NeurIPS), OpenReview link: https://openreview.net/forum?id=QGgeqMV8Er