A0641
Title: Optimal spatial anomaly detection: Theory and applications
Authors: Baiyu Wang - University of Southampton (United Kingdom) [presenting]
Chao Zheng - University of Southampton (United Kingdom)
Abstract: There has been a growing interest in multiple changepoints/anomaly detection problems recently, whilst their focus is mostly on changes taking place on the time index. The anomaly-in-mean model is investigated on the multidimensional spatial lattice, that is, to detect the number and locations of anomaly spatial regions from the baseline. In addition to the usual minimization over cost function with a penalization related to the number of anomalies, a new penalty is introduced on the area of the minimum convex hull that covers the anomaly regions. It is shown that the estimation of the number and locations of anomalies is consistent, and it is proven that the method achieves optimal localization error under the minimax framework. A dynamic programming algorithm is also proposed to solve the penalized cost minimization approximately and carry out large-scale Monte Carlo simulations to examine its performance. The method has a wide range of applications in climate problem. As an example, it is applied to detect the marine heatwaves using the sea surface temperature data from European Space Agency.
