Multistream quickest change detection: Asymptotic optimality under a sparse signal

摘要: In multichannel sequential change detection, multiple sensors monitor a system in which an abrupt occurs at some unknown time and is perceived by subset of sensors. The goal to detect this quickly, while controlling the rate false alarms. traditional asymptotic analysis problem, alarm goes 0 all other parameters remain fixed. We argue that framework not very informative, as corresponding optimality property cannot differentiate between universal parsimonious rules. propose number also infinity, we show context rules may fail be asymptotically optimal when streams small. On hand, are shown under reasonable sparsity conditions.