Chance favors the prepared astrophysicists: forcing serendipitous discoveries in the era of big surveys and AI

University Of Delaware, Newark DE

Investigators

Abstract

The Vera C. Rubin Observatory will take high-cadence observations of the entire southern hemisphere sky over a 10-year period as part of its Legacy Survey of Space and Time (LSST), providing an unprecedented decade-long “movie” of the Universe. One of the most exciting promises of LSST is the discovery of entirely new astrophysical phenomena that have not been previously observed or predicted. The PI, from the University of Delaware, will develop software using machine learning and artificial intelligence (AI) methods to detect anomalous phenomena in the LSST data. The project will include new methodologies specifically designed for LSST data to unveil new objects and phenomena, a software package, and a catalog that includes full characterization of each detected source and its time evolution, as well as tools to aid the interpretation. The products of this research will be publicly available on the Rubin Science Platform. The project will include research and training opportunities a graduate student researcher. Production of the anomalous phenomena detection software package involves three key milestones. The first is feature extraction; methods for light curve characterization will be developed using artificial neural networks and generative AI. The second is development of the open software package for anomaly detection in LSST; an ensemble method for the detection of anomalous light curves mostly using existing methods grounded in the literature and of demonstrated effectiveness will be assembled. Third is application of distance-based methods to enhance interpretability of the anomalous light curves; an existing light curve classification package will be modified to detect anomalous light curves and aid the interpretability of anomalous detections. The project will deliver novel methods for the encoding of LSST data and anomaly detection, along with open-source software for the identification of anomalous light curves. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.