TY - JOUR
T1 - Deblending Galaxies with Generative Adversarial Networks
AU - Hemmati, Shoubaneh
AU - Huff, Eric
AU - Nayyeri, Hooshang
AU - Ferté, Agnès
AU - Melchior, Peter
AU - Mobasher, Bahram
AU - Rhodes, Jason
AU - Shahidi, Abtin
AU - Teplitz, Harry
N1 - Funding Information:
We wish to thank the referee for constructive comments that greatly improved the content and presentation of this paper. S.H. is thankful to A. Cooray group at UCI who let us use their GPU servers. We are thankful to NVIDIA for the GPU granted as their academic grant program. This work used SOMPY , a Python package for self-organizing maps (main contributors: Vahid Moosavi @sevamoo, Sebastian Packmann @sebastiandev, Iván Vallás @ivallesp). Parts of this research were carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration, and at the California Institute of Technology IPAC, sponsored by the National Aeronautics and Space Administration. This work was in part funded by the HST cycle 29 archival program, 16615.
Funding Information:
We wish to thank the referee for constructive comments that greatly improved the content and presentation of this paper. S.H. is thankful to A. Cooray group at UCI who let us use their GPU servers. We are thankful to NVIDIA for the GPU granted as their academic grant program. This work used SOMPY, a Python package for self-organizing maps (main contributors: Vahid Moosavi @sevamoo, Sebastian Packmann @sebastiandev, Iván Vallás @ivallesp). Parts of this research were carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration, and at the California Institute of Technology IPAC, sponsored by the National Aeronautics and Space Administration. This work was in part funded by the HST cycle 29 archival program, 16615.
Publisher Copyright:
© 2022. The Author(s). Published by the American Astronomical Society.
PY - 2022/12/1
Y1 - 2022/12/1
N2 - Deep generative models including generative adversarial networks (GANs) are powerful unsupervised tools in learning the distributions of data sets. Building a simple GAN architecture in PyTorch and training on the CANDELS data set, we generate galaxy images with the Hubble Space Telescope (HST) resolution starting from a noise vector. We proceed by modifying the GAN architecture to improve Subaru Hyper Suprime-Cam (HSC) ground-based images by increasing their resolution to the HST resolution. We use the super-resolution GAN on a large sample of blended galaxies, which we create using CANDELS cutouts. In our simulated blend sample, ∼20% would unrecognizably be blended even in the HST-resolution cutouts. In the HSC-like cutouts this fraction rises to ∼90%. With our modified GAN we can lower this value to ∼50%. We quantify the blending fraction in the high, low, and GAN resolutions over the whole manifold of angular separation, flux ratios, sizes, and redshift difference between the two blended objects. The two peaks found by the GAN deblender result in improvement by a factor of 10 in the photometry measurement of the blended objects. Modifying the architecture of the GAN, we also train a multiwavelength GAN with HST cutouts in seven optical + near-infrared bands. This multiwavelength GAN improves the fraction of detected blends by another ∼10% compared to the single-band GAN. This is most beneficial to the current and future precision cosmology experiments (e.g., LSST, SPHEREx, Euclid, Roman), specifically those relying on weak gravitational lensing, where blending is a major source of systematic error.
AB - Deep generative models including generative adversarial networks (GANs) are powerful unsupervised tools in learning the distributions of data sets. Building a simple GAN architecture in PyTorch and training on the CANDELS data set, we generate galaxy images with the Hubble Space Telescope (HST) resolution starting from a noise vector. We proceed by modifying the GAN architecture to improve Subaru Hyper Suprime-Cam (HSC) ground-based images by increasing their resolution to the HST resolution. We use the super-resolution GAN on a large sample of blended galaxies, which we create using CANDELS cutouts. In our simulated blend sample, ∼20% would unrecognizably be blended even in the HST-resolution cutouts. In the HSC-like cutouts this fraction rises to ∼90%. With our modified GAN we can lower this value to ∼50%. We quantify the blending fraction in the high, low, and GAN resolutions over the whole manifold of angular separation, flux ratios, sizes, and redshift difference between the two blended objects. The two peaks found by the GAN deblender result in improvement by a factor of 10 in the photometry measurement of the blended objects. Modifying the architecture of the GAN, we also train a multiwavelength GAN with HST cutouts in seven optical + near-infrared bands. This multiwavelength GAN improves the fraction of detected blends by another ∼10% compared to the single-band GAN. This is most beneficial to the current and future precision cosmology experiments (e.g., LSST, SPHEREx, Euclid, Roman), specifically those relying on weak gravitational lensing, where blending is a major source of systematic error.
UR - http://www.scopus.com/inward/record.url?scp=85145314730&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85145314730&partnerID=8YFLogxK
U2 - 10.3847/1538-4357/aca1b8
DO - 10.3847/1538-4357/aca1b8
M3 - Article
AN - SCOPUS:85145314730
SN - 0004-637X
VL - 941
JO - Astrophysical Journal
JF - Astrophysical Journal
IS - 2
M1 - 141
ER -