English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Imaging the cosmic 21 cm signal will map out the first billion years of ourUniverse. The resulting 3D lightcone (LC) will encode the properties of theunseen first galaxies and physical cosmology. Here, we build on previous workusing neural networks (NNs) to infer astrophysical parameters directly from 21cm LC images. We introduce recurrent neural networks (RNNs), capable ofefficiently characterizing the evolution along the redshift axis of 21 cm LCimages. Using a large database of simulated cosmic 21 cm LCs, we compare therelative performance in parameter estimation of different networkarchitectures. These including two types of RNNs, which differ in theircomplexity, as well as a more traditional convolutional neural network (CNN).For the ideal case of no instrumental effects, our simplest and easiest totrain RNN performs the best, with a mean squared parameter estimation error(MSE) that is lower by a factor of $\ge 2$ compared with the otherarchitectures studied here, and a factor of $\ge 8$ lower than thepreviously-studied CNN. We also corrupt the cosmic signal by adding noiseexpected from a 1000 h integration with the Square Kilometre Array, as well asexcising a foreground-contaminated 'horizon wedge'. Parameter prediction errorsincrease when the NNs are trained on these contaminated LC images, thoughrecovery is still good even in the most pessimistic case (with $R^2 \ge0.5-0.95$). However, we find no notable differences in performance betweennetwork architectures on the contaminated images. We argue this is due to thesize of our data set, highlighting the need for larger data sets and/or betterdata augmentation in order to maximize the potential of NNs in 21 cm parameterestimation.

Machine learning astrophysics from 21 cm lightcones: impact of network architectures and signal contamination