As big data analytics strive to tackle the most pressing modern data challenges, Paromita Dubey, assistant professor of data sciences and operations, has introduced “distance profiles” — the collection of distance distributions around data points — to analyze non-Euclidean data such as images, networks, and curves in a model-free, geometry-aware way. By focusing on the distances between data points in abstract spaces rather than the data itself, Dubey’s innovation addresses the difficulty of analyzing non-Euclidean data that often defy standard mathematical operations. From ranking the data points and detecting changes in complex data generation processes to comparing populations and identifying relationships, distance profiles enable a wide range of statistical tasks where traditional methods fall short.

Published in The Annals of Statistics (Vol. 52, No. 2, 2024), Dubey’s research, “Metric Statistics: Exploration and Inference for Random Objects with Distance Profiles,” is supported by the National Science Foundation (NSF) grant DMS-2311034.

Marshall Minds is a faculty research series offering audiences a glimpse into complex issues made simple.