NASA's ExoMiner++ AI model expands from Kepler to TESS data

NASA's updated open-source AI, ExoMiner++, was trained on Kepler and TESS observations and flagged about 7,000 TESS signals as exoplanet candidates; the software is available on GitHub.

NASA's ExoMiner++ is an updated open-source AI tool for finding transiting exoplanets in mission data. The model was developed by a team at NASA's Ames Research Center and builds on ExoMiner, which validated 370 exoplanets from Kepler data in 2021. ExoMiner++ was trained on both Kepler and TESS observations and on an initial run identified about 7,000 TESS signals as exoplanet candidates. The software is freely available on GitHub and the team is working on versions that can identify signals directly from raw data. Key facts: - More than 6,000 exoplanets have been discovered overall, with many found using data from NASA's Kepler and TESS missions. - ExoMiner, introduced in 2021, used AI to validate 370 planets from Kepler data. - ExoMiner++ combines Kepler and TESS training and flagged roughly 7,000 TESS targets as exoplanet candidates in an initial pass. - An exoplanet candidate is reported as a signal likely caused by a planet but typically requires follow-up observations to confirm. - The tool is open-source and available on GitHub, and forthcoming missions such as the Nancy Grace Roman Space Telescope will provide additional compatible data. Summary: ExoMiner++ demonstrates how shared AI tools can process large volumes of telescope data to highlight signals that warrant further study. The team is refining the model to detect transit signals from raw data, and work with TESS and future Roman data is expected to continue shaping detection efforts.