Stability of Extended Functional Persistence in Dimensions Zero and One

TL;DR

This paper corrects a previous stability claim for extended persistence diagrams in dimensions zero and one, showing that diagrams of a complex and its one-skeleton may not be close, especially in the 1-dimensional case.

Contribution

It identifies and clarifies the incorrectness of the previous stability result for extended persistence diagrams in low dimensions.

Findings

Extended persistence diagrams of a complex and its one-skeleton can differ significantly.

Counterexample with a disk and distance-to-boundary function illustrates the instability.

Previous stability claims in the literature are invalid in certain cases.

Abstract

The stability result, as stated, is incorrect. In particular, the 1-dimensional extended persistence diagrams of a finely-triangulated simplicial complex X equipped with a continuous real-valued function f, and its one-skeleton graph G (also equipped with f), need not be close. To take an example, let X be a finely-triangulated disc of radius r and let f be the distance-to -the-boundary function, which increases radially from the boundary circle of X and is maximized at the center. The extended persistence diagram of (X,f) contains a point in 1-dimensional persistence of the form (0,r), whereas the 1-dimensional extended persistence diagram of its one-skeleton only has points near the diagonal.