J/AJ/165/95 Identifying Exoplanets with Deep Learning. V. (Tey+, 2023)
Identifying Exoplanets with Deep Learning.
V. Improved Light-curve Classification for TESS Full-frame Image Observations.
Tey E., Moldovan D., Kunimoto M., Huang C.X., Shporer A., Daylan T.,
Muthukrishna D., Vanderburg A., Dattilo A., Ricker G.R., Seager S.
<Astron. J., 165, 95 (2023)>
=2023AJ....165...95T 2023AJ....165...95T
ADC_Keywords: Exoplanets; Optical; Stars, masses; Stars, diameters
Keywords: Neural networks ; Transit photometry ; Exoplanet detection
methods ; Exoplanet catalogs
Abstract:
The TESS mission produces a large amount of time series data, only a
small fraction of which contain detectable exoplanetary transit
signals. Deep-learning techniques such as neural networks have proved
effective at differentiating promising astrophysical eclipsing
candidates from other phenomena such as stellar variability and
systematic instrumental effects in an efficient, unbiased, and
sustainable manner. This paper presents a high-quality data set
containing light curves from the Primary Mission and 1st Extended
Mission full-frame images and periodic signals detected via box
least-squares. The data set was curated using a thorough manual review
process then used to train a neural network called Astronet-Triage-v2.
On our test set, for transiting/eclipsing events, we achieve a 99.6%
recall (true positives over all data with positive labels) at a
precision of 75.7% (true positives over all predicted positives).
Since 90% of our training data is from the Primary Mission, we also
test our ability to generalize on held-out 1st Extended Mission data.
Here, we find an area under the precision-recall curve of 0.965, a 4%
improvement over Astronet-Triage. On the TESS object of interest (TOI)
Catalog through 2022 April, a shortlist of planets and planet
candidates, Astronet-Triage-v2 is able to recover 3577 out of 4140
TOIs, while Astronet-Triage only recovers 3349 targets at an equal
level of precision. In other words, upgrading to Astronet-Triage-v2
helps save at least 200 planet candidates from being lost. The new
model is currently used for planet candidate triage in the Quick-Look
Pipeline.
Description:
During its Prime Mission (2018 July 25-2020 July 4), TESS collected
full-frame images (FFIs) every 30 minutes for 2yr covering 70% of the
entire sky. The FFI cadence was updated to 10 minutes for the 1st
Extended Mission (2020 July 4-2022 September 1). QLP produces light
curves from these images for all observed targets in the TESS Input
Catalog (TIC) with TESS-band magnitude brighter than 13.5.
File Summary:
--------------------------------------------------------------------------------
FileName Lrecl Records Explanations
--------------------------------------------------------------------------------
ReadMe 80 . This file
table2.dat 161 24926 *Threshold-crossing events (TCE) table
--------------------------------------------------------------------------------
Note on table2.dat: The values in this table, particularly in the R*
and Dur columns, do not reflect the actual number of significant
digits. These are the same values that were passed to Astronet during
the training and evaluation.
--------------------------------------------------------------------------------
See also:
IV/38 : TESS Input Catalog - v8.0 (TIC-8) (Stassun+, 2019)
I/352 : Distances to 1.47 billion stars in Gaia EDR3 (Bailer-Jones+, 2021)
J/ApJS/224/12 : Kepler planetary candidates. VII. 48-month (Coughlin+, 2016)
J/AJ/156/102 : TESS Input Catalog and Candidate Target List (Stassun+, 2018)
J/ApJS/235/38 : Kepler planetary cand. VIII. DR25 reliability (Thompson+, 2018)
J/AJ/157/169 : Identifying exoplanets with deep learning K2 (Dattilo+, 2019)
J/AJ/158/25 : Automated triage and vetting of TESS candidates (Yu+, 2019)
J/A+A/633/A53 : TESS planet candidates classification (Osborn+, 2020)
J/ApJS/254/39 : Exoplanet candidates from TESS first 2yr obs (Guerrero+, 2021)
Byte-by-byte Description of file: table2.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 10 I10 --- TIC TESS Input Catalog identifier
12- 22 F11.6 d Epoch [1198/2366] BLS-detected transit center in BTJD
24- 34 F11.7 d Per [0.06/167] BLS-detected period
36- 50 A15 d Dur BLS-detected duration
52- 63 I12 ppm Depth [0/396326348810] BLS-detected depth
65- 72 F8.6 Msun Mass [0.16/3.69]? Stellar mass via TIC 8.2
74- 86 F13.9 Rsun Rad [0.18/151]? Stellar radius via TIC 8.2
88-102 F15.10 Rsun eRad [0.1/1000]? Estimated stellar radius; Section 3.2
104-111 F8.5 mag Tmag [0.57/13.6] TESS-band magnitude
113-113 I1 --- Year [1/3] TESS Cycle signal was detected
115-125 F11.6 d MinT [1325/2362] Minimum BTJD used from the light curve
127-137 F11.6 d MaxT [1541/2390] Maximum BTJD used from the light curve
139-143 A5 --- Split Dataset signal was used for; test, train, or val
145-145 A1 --- CLabel Consensus label (1)
147-147 A1 --- L1 Label assigned by labeller L1
149-149 A1 --- L2 Label assigned by labeller L2
151-151 A1 --- L3 Label assigned by labeller L3
153-153 A1 --- L4 Label assigned by labeller L4
155-155 A1 --- L5 Label assigned by labeller L5
157-157 A1 --- L6 Label assigned by labeller L6
159-159 A1 --- L7 Label assigned by labeller L7
161-161 A1 --- L8 Label assigned by labeller L8
--------------------------------------------------------------------------------
Note (1): Final label used for training by unanimous vote or
discussion. When absent, we used a weighted average of the individual
votes for training.
--------------------------------------------------------------------------------
History:
From electronic version of the journal
References:
Shallue et al. Paper I : 2018AJ....155...94S 2018AJ....155...94S
Dattilo et al. Paper II : 2019AJ....157..169D 2019AJ....157..169D Cat. J/AJ/157/169
Yu et al. Paper III: 2019AJ....158...25Y 2019AJ....158...25Y Cat. J/AJ/158/25
De Beurs et al. Paper IV : 2022AJ....164...49D 2022AJ....164...49D Cat. J/AJ/164/49
(End) Prepared by [AAS], Coralie Fix [CDS], 02-Jun-2023