Artificial Intelligence for the Search of Rare Astrophysical Events

Rare-event searches enable discovery of new physics inaccessible through conventional methods by leveraging specialized radiation detectors. Machine learning provides powerful tools to maximize information extraction from detector data. However, the inherent scarcity of signal events demands building specialized algorithms that integrate raw detector data with domain knowledge and physics constraints. This talk focuses on two key rare event search directions: majorana neutrino and dark matter searches. We provide a comprehensive overview of these rare astrophysical events, their corresponding detecting mechanisms, and specialized core machine learning approaches to address key challenges in detecting them. These algorithms draw from multiple machine learning domains—including geometric deep learning, uncertainty-aware surrogate models, time series analysis, and hardware-AI codesign—and are specifically tailored to the unique challenges posed by rare-event physics.