Collaborative Research: CRCNS Research Proposal: Optimal Transport for Comparative Analysis of Neural Representations

People can recognize the same picture, sound, or sentence even though every brain is somewhat different. This project develops new techniques for measuring how brains are alike and how they differ when they respond to the world. The work focuses on vision, hearing, and language, and it compares patterns in human brain activity with patterns in artificial neural networks. This research leads to a better understanding of healthy brain organization, detect subtle changes related to brain and mental health conditions, and build more useful tests of whether artificial intelligence systems process information in brain-like ways. The project also trains students at the boundary of neuroscience, mathematics, and artificial intelligence, and shares software and benchmarks that other researchers can use. Current methods for comparing neural representations often reduce the similarity between two brains, or between a brain and a model, to one score. That single score hides which response patterns line up, whether cortical topography, meaning the map-like arrangement of response patterns across the brain's surface, is shared across people, whether processing stages align in a similar order, and which stimuli produce shared or different codes across people. This project addresses that problem by developing a family of methods based on Optimal Transport, a mathematical approach for finding interpretable correspondences between complex sets of responses. The research team develops and validates measures that compare the response profiles of individual neurons or small brain locations measured with fMRI. These measures also assess how those response patterns are arranged across cortex, how processing stages align, and which images, sounds, or sentences evoke similar versus different responses across people. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.