Lipschitz functions have \(L_{p}\)-stable persistence (Q967205)

The results in this paper are motivated by the analysis of gene expression data during development. The biological data consists of a series of 13 measurements taken over the development stages of 14 cell types. For each cell type the 13 measurements can be viewed as a samples of a function \(f:[0,1] \rightarrow {\mathbb R}\). The underlying function is assumed to be Lipschitz. Each cross section represents a level surface. For the general results the authors consider a Lipschitz function \(f : {\mathbb R} \rightarrow {\mathbb R}\). Let \({\mathbb X}_{a} = f^{-1}(-\infty,a]\) be the subset defined by a threshold \(a\). Then as \(a\) increases the homology groups of \( {\mathbb X}_{a}\) change. Thus one can observe homology classes appear and disappear. For a given homology class \(\alpha\) one can define a point in \((d(\alpha), b(\alpha)) \in {\mathbb R}^{2}\), where \(b(\alpha)\) is the birth level and \(d(\alpha)\) is the death level. For a given \(\ell\) the set \(Dgm_{\ell}(f)\) of these points corresponding to \(\ell\)-dimensional homology classes is called the \(\ell\)-th persistence diagram of \(f\). The Wasserstein distance between the diagrams for two functions \(f\) and \(g\) gives a measure of the (dis)-similarity between the two functions. The authors show a stability result for two Lipschitz functions on triangulable metric spaces that bounds the Wasserstein distance between them in terms of the \(L_{\infty}\) distance between them.