Wasserstein Metric

Fillmore and his collaborators have investigated the use of the Wasserstein metric to quantify spatial overlap between weighted point sets in diverse metric spaces. By developing a parameter-free similarity distance measure, this work demonstrates that capturing relative density and geometric congruence enhances the classification of complex data, including image descriptors and semantic document shapes. The resulting framework improves classification accuracy over traditional kernel-based methods when analyzing multivariate statistical distributions.