Applied Mathematics Colloquium

Consider a number of time-varying signals enumerated over a discrete index space, e.g., pixel position in a video image, geneprobe index on an mRNA array sequence, or MAC address on the daily logs of a computer network. The objective of spatio-temporal correlation mining is to detect interesting patterns of dependency that exist between such signals over both space and time. We have developed spatio-temporal correlation mining methods that exploit sparse or low-dimensional dependency structure.