This paper addresses the problem of quickest detection of a change in the maximal coherence between columns of a random matrix based on a sequence of matrix observations having a single unknown change point. The random matrix is assumed to have identically distributed rows and the maximal coherence is defined as the largest of the correlation coefficients associated with any row. Likewise, the nearest neighbor (kNN) coherence is defined as the -th largest of these correlation coefficients. The forms of the pre- and post-change distributions of the observed matrices are assumed to belong to the family of elliptically contoured densities with sparse dispersion matrices but are otherwise unknown. A non-parametric stopping rule is proposed that is based on the maximal k-nearest neighbor sample coherence between columns of each observed random matrix. This is a summary statistic that is related to a test of the existence of a hub vertex in a sample correlation graph having a degree at least . Performance bounds on the delay and false alarm performance of the proposed stopping rule are obtained in the purely high dimensional regime where and is fixed. When the pre-change dispersion matrix is diagonal it is shown that, among all functions of the proposed summary statistic, the proposed stopping rule is asymptotically optimal under a minimax quickest change detection (QCD) model as the stopping threshold approaches infinity. The theory developed also applies to sequential hypothesis testing and fixed sample size tests.
View on arXiv