Generalized Sequential Probability Ratio Test for Families of Hypotheses [pdf]
Summary
This article introduces Generalized SPRT for families of hypotheses, extending sequential testing to handle multiple competing hypotheses with error control. It covers the construction of likelihood ratio statistics, stopping boundaries, and theoretical properties, highlighting practical guidelines for implementation in real data streams. The work is valuable for rapidly making reliable decisions across many potential signals or models, with potential applications in AI, anomaly detection, and IT/security contexts.