Development of a neural network based on pyTorch for the identification of planetary candidates among TESS mission targets

Zichen Huang

doi:10.62051/8bvemw44

Authors

Zichen Huang

DOI:

https://doi.org/10.62051/8bvemw44

Keywords:

Exoplanet detection, Transit, Neural network, Deep learning, Data analysis.

Abstract

This paper Since NASA launched the Transiting Exoplanet Survey Satellite (TESS) in 2018, many exoplanets beyond the solar system have been discovered. The TESS mission has collected a magnificent amount of photometric data for scientists to analyze and its instrument observes more than 200,000 target stars in its mission. Scientists identify and analyze the patterns of light curve data for each target and determine if it is a planetary candidate or false positive due to an eclipsing binary star system or instrumental noise. We present a tool to automatically analyze the data of target stars through machine learning and a neural network built with pyTorch. When the Tess satellite finds stars that may have transits, the observation data can be downloaded through computer programs or archived. This will be a simple and time-saving tool that allows people to distinguish real transit planets from eclipsing binaries and observation equipment noise among the targets. The neural network will use the timing data of planet luminosity as input to output the probability that the galaxy really contains exoplanets. By training and testing the neural network, we find that the recall rate and accuracy rate of the network are a and B respectively. Moreover, in the case of X percent, the probability of recognition by this neural network is greater than that of non-planets.

Downloads

Download data is not yet available.

References

[1] Ansdell M, Ioannou Y, Osborn H P, et al. Scientific domain knowledge improves exoplanet transit classification with deep learning[J]. The Astrophysical journal letters, 2018, 869(1): L7.

[2] Batalha N M, Borucki W J, Bryson S T, et al. Kepler’s first rocky planet: Kepler- 10b[J]. The Astrophysical Journal, 2011, 729(1): 27.

[3] William J Borucki 2016 Rep. Prog. Phys. 79 036901

[4] Burke C J, Christiansen J L, Mullally F, et al. Terrestrial planet occurrence rates for the Kepler GK dwarf sample[J]. The Astrophysical Journal, 2015, 809(1): 8.

[5] Karen A. Collins et al 2018 AJ 156 234

[6] Dattilo A, Vanderburg A, Shallue C J, et al. Identifying exoplanets with deep learning. ii. two new super-earths uncovered by a neural network in k2 data[J]. The Astronomical Journal, 2019, 157(5): 169.

[7] Everett M E, Barclay T, Ciardi D R, et al. High-resolution multi-band imaging for validation and characterization of small Kepler planets[J]. The Astronomical Journal, 2015, 149(2): 55.

[8] Kaltenegger L, Selsis F, Fridlund M, et al. Deciphering spectral fingerprints of habitable exoplanets[J]. Astrobiology, 2010, 10(1): 89-102.

[9] Livingston J H, Endl M, Dai F, et al. 44 Validated Planets from K2 Campaign 10[J]. The Astronomical Journal, 2018, 156(2): 78.

[10] Rowe J F, Bryson S T, Marcy G W, et al. Validation of Kepler’s multiple planet candidates. III. Light curve analysis and announcement of hundreds of new multi-planet systems[J]. The Astrophysical Journal, 2014, 784(1): 45.

[11] Shallue C J, Vanderburg A. Identifying exoplanets with deep learning: A five- planet resonant chain around kepler-80 and an eighth planet around kepler- 90[J]. The Astronomical Journal, 2018, 155(2): 94.

[12] Vigan A, Patience J, Marois C, et al. The International Deep Planet Survey-I. The frequency of wide-orbit massive planets around A-stars[J]. Astronomy Astrophysics, 2012, 544: A9.

[13] Wright J T, Gaudi B S. Exoplanet detection methods[J]. arXiv preprint arXiv:1210.2471, 2012.