21cmlstm: A Fast Memory-based Emulator of the Global 21 cm Signal with Unprecedented Accuracy

Neural network (NN) emulators of the global 21 cm signal need an emulation error much less than the observational noise in order to be used to perform unbiased Bayesian parameter inference. To this end, we introduce 21cmLSTM —a long short-term memory (LSTM) NN emulator of the global 21 cm signal tha...

Full description

Saved in:
Bibliographic Details
Main Authors: J. Dorigo Jones, S. M. Bahauddin, D. Rapetti, J. Mirocha, J. O. Burns
Format: Article
Language:English
Published: IOP Publishing 2024-01-01
Series:The Astrophysical Journal
Online Access:https://doi.org/10.3847/1538-4357/ad8b20
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850266448828563456
author J. Dorigo Jones
S. M. Bahauddin
D. Rapetti
J. Mirocha
J. O. Burns
author_facet J. Dorigo Jones
S. M. Bahauddin
D. Rapetti
J. Mirocha
J. O. Burns
author_sort J. Dorigo Jones
collection DOAJ
description Neural network (NN) emulators of the global 21 cm signal need an emulation error much less than the observational noise in order to be used to perform unbiased Bayesian parameter inference. To this end, we introduce 21cmLSTM —a long short-term memory (LSTM) NN emulator of the global 21 cm signal that leverages the intrinsic correlation between frequency channels to achieve exceptional accuracy compared to previous emulators, which are all feedforward, fully connected NNs. LSTM NNs are a type of recurrent NN designed to capture long-term dependencies in sequential data. When trained and tested on the same simulated set of global 21 cm signals as the best previous emulators, 21cmLSTM has an average relative rms error of 0.22%—equivalently 0.39 mK—and comparably fast evaluation time. We perform seven-dimensional Bayesian parameter estimation analyses using 21cmLSTM to fit global 21 cm signal mock data with different adopted observational noise levels, σ _21 . The posterior 1 σ rms error is ≈three times less than σ _21 for each fit and consistently decreases for tighter noise levels, showing that 21cmLSTM can sufficiently exploit even very optimistic measurements of the global 21 cm signal. We have made the emulator, code, and data sets publicly available so that 21cmLSTM can be independently tested and used to retrain and constrain other 21 cm models.
format Article
id doaj-art-6fc048fc7cf9432c8faf6c3b7ed3831f
institution OA Journals
issn 1538-4357
language English
publishDate 2024-01-01
publisher IOP Publishing
record_format Article
series The Astrophysical Journal
spelling doaj-art-6fc048fc7cf9432c8faf6c3b7ed3831f2025-08-20T01:54:11ZengIOP PublishingThe Astrophysical Journal1538-43572024-01-0197711910.3847/1538-4357/ad8b2021cmlstm: A Fast Memory-based Emulator of the Global 21 cm Signal with Unprecedented AccuracyJ. Dorigo Jones0https://orcid.org/0000-0002-3292-9784S. M. Bahauddin1https://orcid.org/0000-0003-0016-5377D. Rapetti2https://orcid.org/0000-0003-2196-6675J. Mirocha3https://orcid.org/0000-0002-8802-5581J. O. Burns4https://orcid.org/0000-0002-4468-2117Center for Astrophysics and Space Astronomy, Department of Astrophysical and Planetary Sciences, University of Colorado Boulder , CO 80309, USA ; johnny.dorigojones@colorado.eduLaboratory for Atmospheric and Space Physics, University of Colorado , Boulder, CO 80303, USACenter for Astrophysics and Space Astronomy, Department of Astrophysical and Planetary Sciences, University of Colorado Boulder , CO 80309, USA ; johnny.dorigojones@colorado.edu; NASA Ames Research Center , Moffett Field, CA 94035, USA; Research Institute for Advanced Computer Science, Universities Space Research Association , Washington, DC 20024, USAJet Propulsion Laboratory, California Institute of Technology , 4800 Oak Grove Drive, Pasadena, CA 91109, USA; California Institute of Technology , 1200 E. California Boulevard, Pasadena, CA 91125, USACenter for Astrophysics and Space Astronomy, Department of Astrophysical and Planetary Sciences, University of Colorado Boulder , CO 80309, USA ; johnny.dorigojones@colorado.eduNeural network (NN) emulators of the global 21 cm signal need an emulation error much less than the observational noise in order to be used to perform unbiased Bayesian parameter inference. To this end, we introduce 21cmLSTM —a long short-term memory (LSTM) NN emulator of the global 21 cm signal that leverages the intrinsic correlation between frequency channels to achieve exceptional accuracy compared to previous emulators, which are all feedforward, fully connected NNs. LSTM NNs are a type of recurrent NN designed to capture long-term dependencies in sequential data. When trained and tested on the same simulated set of global 21 cm signals as the best previous emulators, 21cmLSTM has an average relative rms error of 0.22%—equivalently 0.39 mK—and comparably fast evaluation time. We perform seven-dimensional Bayesian parameter estimation analyses using 21cmLSTM to fit global 21 cm signal mock data with different adopted observational noise levels, σ _21 . The posterior 1 σ rms error is ≈three times less than σ _21 for each fit and consistently decreases for tighter noise levels, showing that 21cmLSTM can sufficiently exploit even very optimistic measurements of the global 21 cm signal. We have made the emulator, code, and data sets publicly available so that 21cmLSTM can be independently tested and used to retrain and constrain other 21 cm models.https://doi.org/10.3847/1538-4357/ad8b20
spellingShingle J. Dorigo Jones
S. M. Bahauddin
D. Rapetti
J. Mirocha
J. O. Burns
21cmlstm: A Fast Memory-based Emulator of the Global 21 cm Signal with Unprecedented Accuracy
The Astrophysical Journal
title 21cmlstm: A Fast Memory-based Emulator of the Global 21 cm Signal with Unprecedented Accuracy
title_full 21cmlstm: A Fast Memory-based Emulator of the Global 21 cm Signal with Unprecedented Accuracy
title_fullStr 21cmlstm: A Fast Memory-based Emulator of the Global 21 cm Signal with Unprecedented Accuracy
title_full_unstemmed 21cmlstm: A Fast Memory-based Emulator of the Global 21 cm Signal with Unprecedented Accuracy
title_short 21cmlstm: A Fast Memory-based Emulator of the Global 21 cm Signal with Unprecedented Accuracy
title_sort 21cmlstm a fast memory based emulator of the global 21 cm signal with unprecedented accuracy
url https://doi.org/10.3847/1538-4357/ad8b20
work_keys_str_mv AT jdorigojones 21cmlstmafastmemorybasedemulatoroftheglobal21cmsignalwithunprecedentedaccuracy
AT smbahauddin 21cmlstmafastmemorybasedemulatoroftheglobal21cmsignalwithunprecedentedaccuracy
AT drapetti 21cmlstmafastmemorybasedemulatoroftheglobal21cmsignalwithunprecedentedaccuracy
AT jmirocha 21cmlstmafastmemorybasedemulatoroftheglobal21cmsignalwithunprecedentedaccuracy
AT joburns 21cmlstmafastmemorybasedemulatoroftheglobal21cmsignalwithunprecedentedaccuracy