Examples of output obtained by horizontal reflection. From top left to bottom right, we display six global views (blue lines) generated for six planetary candidate KOIs of Kepler Q1-Q17 DR25. Each global view provides a horizontally reflective version (orange line). The horizontal reflection procedure applies to all datasets that are highly unbalanced for the class of NPCs. – Astro-Ph.ep
In identifying new planetary candidates in transport research, employment of deep learning models has proven essential for efficient analysis of the continuous, increasing amount of photometric observation.
To further improve the robustness of these models, we will need to utilize the complementarity of data collected from various transport surveys, including NASA's Kepler, Transport Exoplanet Survey Satellite (TESS), and in the near future vibrations of ESA planetary transit and Plato missions.
In this work, we present a deep learning model named Dart-Vetter to distinguish between planetary candidates (PCs) and false positive signals (NPCs) detected in potential transit surveys. Dart-Vetter is a convolutional neural network that processes only rays folded over periods of relative signals, featuring a simpler and more compact architecture with regard to other triangulation and/or examination models available in the literature.
To prove the validity of the model, we trained and tested the Dart-Vetter on several datasets of published and uniformly labeled Tess and Kepler's light curves. Despite its simplicity, Dart-Vetter achieves highly competitive triangular performance with a recall rate of 91% in the Tess and Kepler data ensemble when compared to Exominer and Astronet-Triage.
With its compact, open source, easy to replicate architecture, DART-Vetter is a particularly useful tool for automating triangulation procedures and assisting human bettors, indicating the separate generalization of TCES with multiple event statistics (MES) >20 and orbital durations less than 50 days.
Stefano Fiscale (1 and 2 and 3), Laura Ino (2 and 3), Alessandra Rotundi (1 and 2), Angelo Cialamella (2), Alessio Ferrone (2), Christian Magliano (3 and 4), Luca Kaciapouoti (5), Beserin Kostoff (6), Eliza Kubana (6), Eliza Kubana (6), Beserin Kostoff (6) and 4 and 8), Maria Teresa Mascaritomajoli (1 and 2), Vito Suggese (4), Lucatnietti (1 and 2 and 3 and 9), Antonio Vanzanella (10), Vincenzo de la Corte (3) ((1) UNESCO Chair “Environment, Resources, Resources, Environmental Development” Science and Technology, University of Parthenop, Naples, Naples, Naples, I-80143, Italy, Italy, (3) INAF, Osservatorio Astronomico di Capodimonte, Salita Moiariello, Salita Moiariello, 16, Naples, I-80131, (4) Department editor of Physic, Naples, Italy, (5) Southern European Observatory, Carl Schwarzchild Stress2 D-85748 Garching Bei Munchen, Germany, (6) NASA Goddard Space Flight Center, 8800 Greenbelt Road, Greenbelt, Md 20771, USA, (8) INFN section of Naples, Cinthia 6, 80126, Naples, Italy, (9) Department of Biology, University of Naples, Naples, Italy, (10) National Centre for Nuclear Research 7, 02-093, Waruso, Poland)
Comments: Number of pages: 24, Number of numbers: 8, Articles published in the Astronomical Journal on 2025-05-30
Subjects: Earth and Planetary Astrophysics (Astro-Ph.ep); Astrophysics Instruments and Methods (Astro-PH.IM); Machine Learning (cs.lg)
Quote: arxiv: 2506.05556 [astro-ph.EP] (Or arxiv: 2506.05556v1 [astro-ph.EP] For this version)
https://doi.org/10.48550/arxiv.2506.05556
Focus to learn more
Related doi:
https://doi.org/10.3847/1538-3881/addf4d
Focus to learn more
Submission history
From: Stefano Fiscale
[v1] Thu, June 5, 2025 20:05:16 UTC (1,511 kb)
https://arxiv.org/abs/2506.05556
Astrobiology,
