Persistent Homology

Persistent homology is the central algorithm of topological data analysis, tracking how the topological features of a data cloud — its connected components, loops, voids, and higher-dimensional holes — appear and disappear as a distance threshold grows from zero to infinity. A feature that persists across many scales is interpreted as genuine structure; one that vanishes quickly is interpreted as noise. The method produces a barcode or persistence diagram — a visual record of which features live and which die.

The mathematical foundation lies in simplicial complexes and algebraic topology: as the threshold increases, simplices are added, and the homology groups of the resulting complex are computed at each stage. The persistence of a feature is measured by the difference between its birth and death thresholds. Persistent homology has been used to discover new cancer subtypes, to detect anomalies in complex networks, and to characterize the geometry of random structures. It is proof that shape can be measured even when the shape has no equation.