Tropical Sufficient Statistics for Persistent Homology

Abstract

We show that an embedding in Euclidean space based on tropical geometry generates stable sufficient statistics for barcodes. In topological data analysis, barcodes are multiscale summaries of algebraic topological characteristics that capture the `shape' of data; however, in practice, they have complex structures that make them difficult to use in statistical settings. The sufficiency result presented in this work allows for classical probability distributions to be assumed on the tropical geometric representation of barcodes. This makes a variety of parametric statistical inference methods amenable to barcodes, all while maintaining their initial interpretations. More specifically, we show that exponential family distributions may be assumed, and that likelihood functions for persistent homology may be constructed. We conceptually demonstrate sufficiency and illustrate its utility in persistent homology dimensions 0 and 1 with concrete parametric applications to human immunodeficiency virus and avian influenza data.

31 pages, 5 figures

References in corpus (8)

• A roadmap for the computation of persistent homology
• Statistical topological data analysis using persistence landscapes
• Clique topology reveals intrinsic geometric structure in neural correlations
• Confidence sets for persistence diagrams
• Deep Learning with Topological Signatures
• Evasion Paths in Mobile Sensor Networks
• Modeling and replicating statistical topology, and evidence for CMB non-homogeneity
• Symmetric and r-Symmetric Tropical Polynomials and Rational Functions

Cited by in corpus (2)

• Geometric and Probabilistic Limit Theorems in Topological Data Analysis
• From Persistent Homology to Reinforcement Learning with Applications for Retail Banking