NASA has released an updated version of its open-source artificial intelligence software, ExoMiner++, designed to identify new transiting exoplanets from archived data collected by the agency's Kepler and Transiting Exoplanet Survey Satellite (TESS) missions. The software has initially identified 7,000 exoplanet candidates from TESS data and is freely available to researchers globally.
Software Development and Capabilities
ExoMiner++ builds upon the original ExoMiner software, which was developed in 2021 by a team at NASA's Ames Research Center. The initial version validated 370 new exoplanets using data from the retired Kepler mission. The updated ExoMiner++ has been trained on datasets from both Kepler and TESS. To date, over 6,000 exoplanets have been discovered, with more than half originating from data provided by these two missions. Both Kepler and TESS datasets are publicly available in NASA archives.
The algorithm, detailed in the Astronomical Journal, processes observations of possible transits. It distinguishes signals caused by exoplanets from those generated by other astronomical events, such as eclipsing binary stars. This deep learning technology is designed to be effective with large datasets that contain numerous signals. Exoplanet candidates are signals that are likely planets but require further confirmation from additional telescopes.
Open-Source Initiative
ExoMiner++ is available for free download on GitHub, allowing the international scientific community to utilize the tool for exoplanet discovery. Kevin Murphy, NASA's Chief Science Data Officer, has stated that open-source software facilitates scientific discovery by enabling the replication of results and deeper data analysis. Jon Jenkins, an exoplanet scientist at NASA Ames, noted that open science initiatives and open-source software contribute to advancements in the exoplanet field. Hamed Valizadegan leads the project, with Miguel Martinho serving as a co-investigator for ExoMiner++.
Data Compatibility and Future Enhancements
The Kepler and TESS missions employ different observing strategies; Kepler observed a specific, smaller area of the sky in depth, while TESS surveys nearly the entire sky for nearby transiting stars. Despite these differences, the missions produce compatible datasets, which allowed ExoMiner++ to train effectively on data from both telescopes.
Future iterations of ExoMiner++ aim to enhance its utility. Plans include enabling the model to identify transit signals directly from raw data, rather than solely flagging candidates from a pre-defined list. The software could also assist in analyzing data from upcoming exoplanet-hunting missions, such as NASA's Nancy Grace Roman Space Telescope, whose data will also be freely available.