| Name: Anoop Gavankar |
| Affiliation: TATA INSTITUTE OF FUNDAMENTAL RESEARCH, MUMBAI |
| Conference ID: ASI2026_313 |
| Title: Vision Transformers as a Robust Alternative for Identifying Planetary Candidates in Solar EPRV Data |
| Abstract Type: Poster |
| Abstract Category: Sun, Solar System, Exoplanets, and Astrobiology |
| Author(s) and Co-Author(s) with Affiliation: Anoop Gavankar(Tata Institute of Fundamental Research, Mumbai), Tanish Mittal(Birla Institute of Technology, Pilani), Joe Ninan(Tata Institute of Fundamental Research, Mumbai), Shravan Hanasoge(Tata Institute of Fundamental Research, Mumbai) |
| Abstract: Extreme precision radial velocity (EPRV) surveys usually require extensive observational baselines to confirm planetary candidates, making them resource-intensive. Traditionally, periodograms are used to identify promising signals before further observational investment, but their effectiveness is often limited for low-amplitude signals due to stellar activity. We introduce a machine-learning (ML) framework that extracts planetary signals from spectroscopic time-series data. Injection-recovery tests on randomly selected 100-observation subsets from NEID solar data (2020–2022 period) show that for low-amplitude systems (<1 m/s), our model improves planetary candidate identification by a factor of two compared to the traditional Lomb-Scargle periodogram. This highlights the potential of ML as a robust alternative for identifying planetary candidate signals in EPRV surveys.
Our ML model is based on Vision Transformer (ViT) architecture that intakes a reduced representation of the solar spectrum observations and predicts the period and semi-amplitude of a planetary signal candidate. In this talk, I will present our model framework, data preprocessing, and the final results in comparison to the traditional Lomb-Scargle periodogram. |
