Metrics and barycenters for point pattern data

摘要: We introduce the transport–transform and relative metrics between finite point patterns on a general space, which provide unified framework for earlier pattern metrics, in particular generalized spike time normalized unnormalized optimal subpattern assignment metrics. Our main focus is barycenters, i.e., minimizers of q-th-order Frechet functional with respect to these present heuristic algorithm that terminates local minimum shown be fast reliable simulation study. The serves as plug-in method can applied any state space where an appropriate solving location problem individual points available. applications geocoded data crimes Euclidean street network, illustrating barycenters serve informative summary statistics. work first step toward statistical inference covariate-based models repeated observations.