Zusammenfassung
Emulation of the Global (sky-averaged) 21-cm signal from the Cosmic Dawn and
Epoch of Reionization with neural networks has been shown to be an essential
tool for physical signal modelling. In this paper we present globalemu, a
Global 21-cm signal emulator that uses redshift as a character defining
variable along side a set of astrophysical parameters to estimate the
brightness temperature of the 21-cm signal. Combined with a physically
motivated pre-processing of the data this makes for a reliable and fast
emulator that is relatively insensitive to the neural network design. A single
high resolution signal can be emulated in 1.3 ms when using globalemu in
comparison to 133 ms, a factor of 102 improvement, when using the existing
public state of the art emulator 21cmGEM evaluated with the same computing
power. We illustrate, with the same training and test data used for 21cmGEM,
that globalemu is almost twice as accurate as 21cmGEM and for 95% of models in
a test set of $\approx1,700$ we can achieve a RMSE of $5.37$ mK and a mean
RMSE of 2.52 mK across the band z = 7 -28 (approximately 10% the expected noise
of 25 mK for the Radio Experiment for the Analysis of Cosmic Hydrogen (REACH)).
Further, globalemu provides a flexible framework in which the neutral fraction
history and Global signal models with updated astrophysics can be emulated
easily. The emulator is pip installable and available at:
https://github.com/htjb/globalemu. globalemu will be used by the REACH
collaboration to perform physical signal modelling inside a Bayesian nested
sampling loop.
Nutzer