J/ApJ/935/90 Superflare candidates in ∼72000 G-stars (Tu+, 2022)
Convolutional Neural Networks for Searching Superflares from Pixel-level Data of
the Transiting Exoplanet Survey Satellite.
Tu Z.-L., Wu Q., Wang W., Zhang G.Q., Liu Z.-K., Wang F.Y.
<Astrophys. J., 935, 90 (2022)>
=2022ApJ...935...90T 2022ApJ...935...90T
ADC_Keywords: Stars, flare ; Stars, G-type; Stars, diameters;
Effective temperatures; Optical
Keywords: Stellar flares ; G stars ; Convolutional neural networks
Abstract:
In this work, six convolutional neural networks (CNNs) have been
trained based on 15638 superflare candidates on solar-type stars,
which are collected from the three years of Transiting Exoplanet
Survey Satellite (TESS) observations. These networks are used to
replace the manually visual inspection, which was a direct way of
searching for superflares, and exclude false-positive events in recent
years. Unlike other methods, which only used stellar light curves to
search for superflare signals, we try to identify superflares through
TESS pixel-level data with lower risk of mixing false-positive events
and give more reliable identification results for statistical
analysis. The evaluated accuracy of each network is around 95.57%.
After applying ensemble learning to these networks, the stacking
method promotes accuracy to 97.62% with a 100% classification rate,
and the voting method promotes accuracy to 99.42% with a relatively
lower classification rate at 92.19%. We find that superflare
candidates with short duration and low peak amplitude have lower
identification precision, as their superflare features are hard to be
identified. The database includes 71732 solar-type stars and 15,638
superflare candidates from TESS with corresponding feature images and
arrays, and the trained CNNs in this work are public available.
Description:
This work focuses on developing convolutional neural networks (CNNs)
for superflare automatic searching on pixel-level data. We use 15638
superflare candidates.
File Summary:
--------------------------------------------------------------------------------
FileName Lrecl Records Explanations
--------------------------------------------------------------------------------
ReadMe 80 . This file
table1.dat 72 71732 Properties of solar-type stars
table2.dat 340 15638 Information of superflare candidates
--------------------------------------------------------------------------------
See also:
I/311 : Hipparcos, the New Reduction (van Leeuwen, 2007)
I/345 : Gaia DR2 (Gaia Collaboration, 2018)
IV/38 : TESS Input Catalog - v8.0 (TIC-8) (Stassun+, 2019)
I/350 : Gaia EDR3 (Gaia Collaboration, 2020)
J/AJ/141/50 : White-light flares cool stars from Kepler (Walkowicz+, 2011)
J/other/Nat/485.478 : Superflares on solar-type stars (Maehara+, 2012)
J/AJ/143/93 : Rotational velocities in early-M stars (Reiners+, 2012)
J/ApJS/209/5 : Superflares of Kepler stars. I. (Shibayama+, 2013)
J/ApJS/211/24 : Rotation periods of Kepler MS stars (McQuillan+, 2014)
J/ApJ/829/23 : Stellar flares from Q0-Q17 Kepler LCs (Davenport, 2016)
J/MNRAS/489/437 : Origin stellar flares on M dwarfs using TESS (Doyle+, 2019)
J/ApJS/241/29 : Flare catalog through LC data of Kepler DR25 (Yang+, 2019)
J/ApJ/890/46 : Superflares on solar-type stars TESS first year (Tu+, 2020)
J/MNRAS/502/2033 : Flares in the Kepler short cadence data (Jackman+, 2021)
J/ApJ/906/72 : 2344 superflares on 266 solar-type stars (Okamoto+, 2021)
J/MNRAS/503/5263 : Sorting of 4XMM-DR9 sources machine learning (Zhang+, 2021)
Byte-by-byte Description of file: table1.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 10 I10 --- ID [185/2055866439] TESS identifier
12- 16 F5.2 mag Tmag [2.75/19.5] TESS apparent magnitude (1)
18- 24 F7.2 K Teff [5100/6000] Effective stellar surface
temperature (1)
26- 29 F4.2 [cm/s2] logg [4.0/5.0] log stellar surface gravity (1)
31- 34 F4.2 solRad Rad [0.49/1.74] Stellar radius (1)
36- 41 F6.2 d Per [0.1/346] Orbital period
43- 51 F9.6 --- Rvar [0.0002/20] Stellar photometric variability;
Equ. 1 (Rvar=F95%-F5%) (2)
53- 72 A20 --- Flag Cross-matching flag (3)
--------------------------------------------------------------------------------
Note (1): From TESS input catalog (TIC v8 ; Stassun, 2019, IV/38).
Note (2): The photometric variability (Rvar) is derived by
Rvar=F95%-F5%
where F95% and F5% stand for the upper 95% and lower 5% of the ranked
normalized flux of stellar light curves (Basri+ 2010ApJ...713L.155B 2010ApJ...713L.155B)
See Section 2.1.
Note (3): Cross-matching with catalogs from Hipparocos-2 and Gaia-EDR3. Here,
HB stands for stars may be binary systems from Hipparocos-2 catalog.
GM and GB stand for the stars may contain M-type stars or other
brighter stars nearby, which are collected by Gaia-EDR3. 21 and 42
stand for the possible contamination, which are located within 21"
or within 21" to 42" distance from the main targets.
--------------------------------------------------------------------------------
Byte-by-byte Description of file: table2.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 10 I10 --- ID [64267/1980393697] TESS identifier
12- 13 I2 --- Sector [1/39] Superflare candidate sector number
15- 19 A5 --- Source Source identifier (1)
21- 26 A6 --- PeakNum Flare candidate serial number (2)
28 I1 --- Label [0/2] Label identifier (3)
30- 38 F9.4 d PDate [1325.3014/2389.7044] TESS timestamp os superflare
peak
40- 47 E8.2 10-7W Lum ? Flare peak luminosity; erg/s (4)
49- 56 E8.2 10-7J Energy ? Superflare total energy from Equ. 6; erg
58- 65 F8.2 s Duration [239/18480]? Superflare duration (5)
67-340 A274 --- File Feature candidate image & array file names (6)
--------------------------------------------------------------------------------
Note (1): 'long' and 'short' stand for n=5 and n=2 in Equation 2.
Note (2): Captured from the star in a specific sector. For example, the first
row stands for the candidate, which is the eleventh one detected
from the light curve of TIC382575967 in sector seven, and n=5 is
set in Equation 2.
Note (3): Labels of Gold-class, Silver-class and None-class, which is
represented by 0, 1, and 2, respectively. All labels are set by
visual inspection of candidates. Here, we have 1268 candidates in
Gold-class, 3792 candidates in Silver-class. And other 10578
candidates are in None-class.
Note (4): Deduced by stellar luminosity (L* in Equation 3) multiplying
by Fflare(t) at the peak time as defined in Equation 4.
Note (5): Calculated by tend - tstart. Start and end times of superflare
is derived through Equation 4.
Note (6): These files are used for training CNNs. Their names are structured by
types of data, the number of sector, source, and the serial number of
the candidate. Note that, data of TD-normal. and TD-unnormal. are not
saved in any picture formats but in array format of Numpy (a Python
package) with filename extension .npy.
--------------------------------------------------------------------------------
History:
From electronic version of the journal
(End) Prepared by [AAS], Coralie Fix [CDS], 06-Jun-2024