@article{5851b1d3bcd343e4a8342f724446a4a4,
title = "Accurate Machine-learning Atmospheric Retrieval via a Neural-network Surrogate Model for Radiative Transfer",
abstract = "Atmospheric retrieval determines the properties of an atmosphere based on its measured spectrum. The low signal-to-noise ratios of exoplanet observations require a Bayesian approach to determine posterior probability distributions of each model parameter, given observed spectra. This inference is computationally expensive, as it requires many executions of a costly radiative transfer (RT) simulation for each set of sampled model parameters. Machine learning (ML) has recently been shown to provide a significant reduction in runtime for retrievals, mainly by training inverse ML models that predict parameter distributions, given observed spectra, albeit with reduced posterior accuracy. Here we present a novel approach to retrieval by training a forward ML surrogate model that predicts spectra given model parameters, providing a fast approximate RT simulation that can be used in a conventional Bayesian retrieval framework without significant loss of accuracy. We demonstrate our method on the emission spectrum of HD 189733 b and find good agreement with a traditional retrieval from the Bayesian Atmospheric Radiative Transfer (BART) code (Bhattacharyya coefficients of 0.9843–0.9972, with a mean of 0.9925, between 1D marginalized posteriors). This accuracy comes while still offering significant speed enhancements over traditional RT, albeit not as much as ML methods with lower posterior accuracy. Our method is ∼9× faster per parallel chain than BART when run on an AMD EPYC 7402P central processing unit (CPU). Neural-network computation using an NVIDIA Titan Xp graphics processing unit is 90×–180× faster per chain than BART on that CPU.",
author = "Himes, {Michael D.} and Joseph Harrington and Cobb, {Adam D.} and Baydin, {Atılım G{\"u}ne{\c s}} and Frank Soboczenski and O{\textquoteright}Beirne, {Molly D.} and Simone Zorzan and Wright, {David C.} and Zacchaeus Scheffer and Domagal-Goldman, {Shawn D.} and Arney, {Giada N.}",
note = "Funding Information: We gratefully acknowledge Jon Malkin, Lee Rhodes, and Edo Liberty for valuable contributions to the streaming quantiles method used in this work. We thank James Mang and Nicholas Susemiehl for useful feedback on the software developed for this work. We thank Michael Lund and the NASA Exoplanet Archive for preparing and hosting the online RRC. We thank Jennifer Adams for helpful discussions on radiative transfer emulation in Earth science. We also thank contributors to the Datasketches library, NumPy, SciPy, Matplotlib, Tensorflow, Keras, the Python Programming Language, the free and open-source community, and the NASA Astrophysics Data System for software and services. We gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan Xp GPU used for this research. This research was supported by the NASA Fellowship Activity under NASA Grant 80NSSC20K0682 and NASA Exoplanets Research Program grant NNX17AB62G. We thank FDL (http://www.frontierdevelopmentlab.org/) and SETI (https://www.seti.org) for making this collaboration possible. Funding Information: We gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan Xp GPU used for this research. This research was supported by the NASA Fellowship Activity under NASA Grant 80NSSC20K0682 and NASA Exoplanets Research Program grant NNX17AB62G. We thank FDL (http://www.frontierdevelopmentlab.org/) and SETI (https://www.seti.org) for making this collaboration possible. Publisher Copyright: {\textcopyright} 2022. The Author(s). Published by the American Astronomical Society.",
year = "2022",
month = apr,
day = "1",
doi = "10.3847/PSJ/abe3fd",
language = "English",
volume = "3",
journal = "Planetary Science Journal",
issn = "2632-3338",
publisher = "IOP Publishing Ltd.",
number = "4",
}