This paper presents Astronet-Triage-v2, an improved deep learning model for classifying light curves from the TESS mission to identify exoplanets. Using a large, manually reviewed dataset of light curves, Astronet-Triage-v2 achieves 99.6% recall and 75.7% precision in identifying transiting/eclipsing events on the test set. It shows a 4% improvement in AUC-PR over its predecessor, Astronet-Triage, and recovers more planet candidates from the TESS Object of Interest (TOI) catalog. The model is currently used in the TESS Quick-Look Pipeline.
Publisher
arXiv
Published On
Jan 05, 2023
Authors
Evan Tey, Dan Moldovan, Michelle Kunimoto, Chelsea X Huang, Avi Shporer, Tansu Daylan, Daniel Muthukrishna, Andrew Vanderburg, Anne Dattilo, George R Ricker, S Seager
Tags
exoplanets
TESS
deep learning
light curves
transiting events
Astronet-Triage-v2
classification
Related Publications
Explore these studies to deepen your understanding of the subject.