Diffusion Geometry

TL;DR

Diffusion geometry introduces a Riemannian-based framework for geometric data analysis, offering robustness to noise, computational efficiency, and richer topological insights, with applications in biomedical data analysis.

Contribution

It develops a new diffusion geometry framework using Bakry-Emery calculus, enabling statistical estimation of geometric objects on probability spaces, surpassing existing methods in robustness and richness.

Findings

Outperforms multiparameter persistent homology in tumor histology analysis

Robustly detects singularities in manifold-like data

Provides a faster, noise-robust alternative to existing geometric methods

Abstract

We introduce diffusion geometry as a new framework for geometric and topological data analysis. Diffusion geometry uses the Bakry-Emery $\Gamma$-calculus of Markov diffusion operators to define objects from Riemannian geometry on a wide range of probability spaces. We construct statistical estimators for these objects from a sample of data, and so introduce a whole family of new methods for geometric data analysis and computational geometry. This includes vector fields and differential forms on the data, and many of the important operators in exterior calculus. Unlike existing methods like persistent homology and local principal component analysis, diffusion geometry is explicitly related to Riemannian geometry, and is significantly more robust to noise, significantly faster to compute, provides a richer topological description (like the cup product on cohomology), and is naturally vectorised for statistics and machine learning. We find that diffusion geometry outperforms multiparameter persistent homology as a biomarker for real and simulated tumour histology data and can robustly measure the manifold hypothesis by detecting singularities in manifold-like data.