Optimal Transport as a Metric for Measuring Segregation

Daniel Peng, James M. Murphy

Session: General

Keywords: Graph signal processing, optimal transport, spatial geography, segregation

TL;DR: We propose and analyze a method for measuring segregation on graphs via optimal transport.

Abstract: In this paper, we propose the OTSeg score, a new quantitative metric that utilizes optimal transport to measure the degree of segregation on graphs. We characterize maximizers and minimizers for the metric and consider its relationship to classical measures of segregation including those based on spectral graph theory. We validate the effectiveness of OTSeg through experiments on synthetic and real datasets derived from American political geography, highlighting its potential for applications in computational social science.

Submission Number: 15