J/A+A/678/A103 Lensed quasar identification in multiband images (Andika+, 2023)
Streamlined lensed quasar identification in multiband images via ensemble
networks.
Andika I.T., Suyu S.H., Canameras R., Melo A., Schuldt S., Shu Y.,
Eilers A.-C., Jaelani A.T., Yue M.
<Astron. Astrophys., 678, A103 (2023)>
=2023A&A...678A.103A 2023A&A...678A.103A (SIMBAD/NED BibCode)
ADC_Keywords: QSOs ; Gravitational lensing ; Photometry, SDSS ;
Photometry, infrared
Keywords: galaxies: active - quasars: general -
quasars: supermassive black holes - gravitational lensing: strong -
methods: data analysis
Abstract:
Quasars experiencing strong lensing offer unique viewpoints on
subjects related to the cosmic expansion rate, the dark matter profile
within the foreground deflectors, and the quasar host galaxies.
Unfortunately, identifying them in astronomical images is challenging
since they are overwhelmed by the abundance of non-lenses. To address
this, we have developed a novel approach by ensembling cutting-edge
convolutional networks (CNNs) - for instance, ResNet, Inception,
NASNet, MobileNet, EfficientNet, and RegNet - along with vision
transformers (ViTs) trained on realistic galaxy-quasar lens
simulations based on the Hyper Suprime-Cam (HSC) multiband images.
While the individual model exhibits remarkable performance when
evaluated against the test dataset, achieving an area under the
receiver operating characteristic curve of >97.3% and a median false
positive rate of 3.6%, it struggles to generalize in real data,
indicated by numerous spurious sources picked by each classifier. A
significant improvement is achieved by averaging these CNNs and ViTs,
resulting in the impurities being downsized by factors up to 50.
Subsequently, combining the HSC images with the UKIRT, VISTA, and
unWISE data, we retrieve approximately 60 million sources as parent
samples and reduce this to 892 609 after employing a photometry
preselection to discover z>1.5 lensed quasars with Einstein radii of
θE<5". Afterward, the ensemble classifier indicates 3080
sources with a high probability of being lenses, for which we visually
inspect, yielding 210 prevailing candidates awaiting spectroscopic
confirmation. These outcomes suggest that automated deep learning
pipelines hold great potential in effectively detecting strong lenses
in vast datasets with minimal manual visual inspection involved.
Description:
We conduct a systematic hunt for lensed quasars at 1.5≤z≤7.2 by
exploiting the HSC, UKIRT, VISTA, unWISE, and Gaia data.
Complete list of our lens candidates.
File Summary:
--------------------------------------------------------------------------------
FileName Lrecl Records Explanations
--------------------------------------------------------------------------------
ReadMe 80 . This file
tableb1.dat 116 242 List of rediscovered strong lenses and
newly found lensed quasar candidates
--------------------------------------------------------------------------------
Byte-by-byte Description of file: tableb1.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 6 I6 --- ID Identification number
8- 26 A19 --- Name Name of the source (JDDD.ddddd+DD.ddddd, where
DDD.ddddd and +DD.ddddd are, respectively,
the RA and DE in decimal degrees (J2000))
28- 32 F5.2 mag gmag HSC g band magnitude (AB) (1)
34- 37 F4.2 mag e_gmag ? HSC g band magnitude 1σ uncertainty
39- 43 F5.2 mag rmag HSC r band magnitude (AB) (1)
45- 48 F4.2 mag e_rmag ? HSC r band magnitude 1σ uncertainty
50- 54 F5.2 mag imag HSC i band magnitude (AB) (1)
56- 59 F4.2 mag e_imag HSC i band magnitude 1σ uncertainty
61- 65 F5.2 mag zmag HSC z band magnitude (AB) (1)
67- 70 F4.2 mag e_zmag HSC z band magnitude 1σ uncertainty
72- 76 F5.2 mag ymag HSC y band magnitude (AB) (1)
78- 81 F4.2 mag e_ymag HSC y band magnitude 1σ uncertainty
83- 87 F5.2 mag Jmag ?=- NIR J-band magnitude (AB) (1)
89- 92 F4.2 mag e_Jmag ? NIR J-band magnitude error
94- 98 F5.2 mag W1mag ?=- unWISE W1-band magnitude (AB) (1)
100-103 F4.2 mag e_W1mag ? unWISE W1-band magnitude error
105-108 F4.2 --- Plens Lens probability based on the ensemble
network classification
110-111 A2 --- Grade Grade after visual inspection (2)
113-114 I2 --- Ref ? Reference for the list of lenses or
candidates published by earlier works (3)
116 A1 ---- GradeRef Grade reported in the reference
--------------------------------------------------------------------------------
Note (1): magnitudes corrected for Galactic extinction based on the dust map of
Schlafly et al. (2019ApJS..240...30S 2019ApJS..240...30S, Cat. II/363) and considering the
Fitzpatrick (1999PASP..111...63F 1999PASP..111...63F) reddening equation.
Note (2): Based on the visual inspection grades, best and good lens candidates
with Plens>0.3 are marked with A and B, respectively. On the other hand,
sources with Plens≤0.3 but show lensing features and might be missed by our
classifier are marked with grade A* or B*, depending on their quality.
Note (3): References as follows:
1 = Canameras et al. (2021A&A...653L...6C 2021A&A...653L...6C, Cat. J/A+A/653/L6)
2 = Canameras et al. (2021A&A...653L...6C 2021A&A...653L...6C, Cat. J/A+A/653/L6)
3 = Shu et al. (2022A&A...662A...4S 2022A&A...662A...4S, Cat. J/A+A/662/A4)
4 = Jacobs et al. (2019ApJS..243...17J 2019ApJS..243...17J, Cat. J/ApJS/243/17)
5 = Storfer et al. (2022, ArXiv e-prints [arXiv: 2206.02764])
6 = Chan et al. (2023, ArXiv e-prints [arXiv: 2304.05425] )
7 = Wong et al. (2022PASJ...74.1209W 2022PASJ...74.1209W, Cat. J/PASJ/74/1209)
8 = Li et al. (2020ApJ...899...30L 2020ApJ...899...30L)
9 = Petrillo et al. (2019MNRAS.484.3879P 2019MNRAS.484.3879P, Cat. J/MNRAS/484/3879)
10 = Sonnenfeld et al. (2020A&A...642A.148S 2020A&A...642A.148S, Cat. J/A+A/642/A148)
11 = Jaelani et al. (2020MNRAS.495.1291J 2020MNRAS.495.1291J, Cat. J/MNRAS/495/1291)
12 = Sonnenfeld et al. (2018PASJ...70S..29S 2018PASJ...70S..29S)
13 = Stein et al. (2022ApJ...932..107S 2022ApJ...932..107S)
14 = Huang et al. (2020ApJ...894...78H 2020ApJ...894...78H, Cat. J/ApJ/894/78)
--------------------------------------------------------------------------------
History:
From electronic version of the journal
(End) Patricia Vannier [CDS] 07-Dec-2023