Abstract:

Root-cause identification can be considered as a quickest change detection/isolation problem, which is of importance for a variety of applications. Efficient statistical decision tools that have sufficient power while controlling the false alarm rate are needed for detecting and isolating abrupt changes in the properties of stochastic signals and dynamical systems, ranging from on-line fault diagnosis in complex technical systems to the target detection/classification in radar, infrared, and sonar signal processing. Noting that there are usually some natural grouping in those system, this talk introduces some new approaches that utilize sparsity and group structure to handle multiple testing problem and demonstrate their advantages over existing methods.