English

Kan Approximations of the Persistent Homology Transform

Algebraic Topology 2025-07-31 v1 Computational Geometry Category Theory

Abstract

The persistent homology transform (PHT) of a subset MRdM \subset \mathbb{R}^d is a map PHT(M):Sd1Dgm\text{PHT}(M):\mathbb{S}^{d-1} \to \mathbf{Dgm} from the unit sphere to the space of persistence diagrams. This map assigns to each direction vSd1v\in \mathbb{S}^{d-1} the persistent homology of the filtration of MM in direction vv. In practice, one can only sample the map PHT(M)\text{PHT}(M) at a finite set of directions ASd1A \subset \mathbb{S}^{d-1}. This suggests two natural questions: (1) Can we interpolate the PHT from this finite sample of directions to the entire sphere? If so, (2) can we prove that the resulting interpolation is close to the true PHT? In this paper we show that if we can sample the PHT at the module level, where we have information about how homology from each direction interacts, a ready-made interpolation theory due to Bubenik, de Silva, and Nanda using Kan extensions can answer both of these questions in the affirmative. A close inspection of those techniques shows that we can infer the PHT from a finite sample of heights from each direction as well. Our paper presents the first known results for approximating the PHT from finite directional and scalar data.

Keywords

Cite

@article{arxiv.2507.22816,
  title  = {Kan Approximations of the Persistent Homology Transform},
  author = {Shreya Arya and Justin Curry},
  journal= {arXiv preprint arXiv:2507.22816},
  year   = {2025}
}

Comments

22 pages, 4 figures. Dedicated to our beloved graduate and post-doc mentor, Sayan Mukherjee (1971--2025)