Point Set Similarity

Point set similarity is critical for the classification of complex spatial data, such as image descriptors and semantic document shapes. Fillmore and his collaborators developed a parameter-free similarity distance metric that quantifies the spatial overlap between weighted point sets across any metric space. This approach captures relative density and geometric congruence, demonstrating superior classification accuracy compared to existing methods like the Pyramid Match kernel.