Learning from the Shape of Data

This paper examines the epistemic value of using topological methods to study the "shape" of data sets. It is argued that the category theoretic notion of "functoriality" aids in translating visual intuitions about structure in data into precise, computable descriptions of real-world systems