Project Grant 2334898

This $298,988 federal Project Grant awarded by the National Science Foundation (NSF) under its Computer and Information Science and Engineering (CISE) program (CFDA 47.070) will support research to develop a novel score-based approach for quickly detecting abrupt changes in the statistical characteristics of online data streams. The project aims to leverage deep neural networks to learn the score (gradient of the log probability density) of data, which can enable change detection without requiring prior knowledge of probability distributions. The research will establish the theoretical foundations, develop robust methods, and create distributed algorithms for this score-based quickest change detection approach. The proposed techniques have broad applications in areas such as anomaly detection, power grid monitoring, pandemic onset tracking, and cyber-attack identification. This project, led by researchers at Duke University, will also incorporate broader impact efforts to disseminate the developed algorithms and provide research opportunities for underrepresented students.