Learning from the shape of data

To make sense of large data sets, we often look for patterns in how data points are “shaped” in the space of possible measurement outcomes. The emerging field of topological data analysis (TDA) offers a toolkit for formalizing the process of identifying such shapes. This article aims to discover why and how the resulting analysis should be understood as reflecting significant features of the systems that generated the data. I argue that a particular feature of TDA—its functoriality—is what enables TDA to translate visual intuitions about structure in data into precise, computationally tractable descriptions of real-world systems.