|
Abstract: Monitoring high-dimensional data streams has become
increasingly important for real-time detection of abnormal
activities in many data-rich applications. In this talk, I will give
a brief discussion on recent works from the following aspects: i)
develop an efficient global monitoring procedure when we do not know
which subset of data streams is affected by an occurring event; ii)
suggest a procedure which is able to control the conditional false
discovery rate at each time point when our focus is detecting
changes in each individual data stream; iii) propose a
distribution-free detection scheme in the sense that its in-control
run-length distribution is free of the underlying distribution.
|
