A1039
Title: A general methodology for fast online changepoint detection
Authors: Per August Moen - University of Oslo (Norway) [presenting]
Abstract: A general methodology for online changepoint detection is presented. By using a sparse and dynamically updating grid of probe points, the methodology allows the user to apply a wide range of offline changepoint tests on sequentially observed data. The methodology is designed to have low update and storage costs, and for a certain class of test statistics, the methodology is guaranteed to have update and storage costs scaling logarithmically with the sample size. In particular, the methodology can be used to detect a change in the mean vector or covariance matrix with near-optimal non-asymptotic mathematical guarantees, which can be shown to be near-optimal from a minimax perspective. The effectiveness of the methodology is demonstrated via a simulation study, and its applicability is illustrated by considering real-time detection of covariance changes in currency exchange rates.
