Stability and Extension of Steady and Ranging Persistence
Yann-Situ Gazull · 2025
Read Stability and Extension of Steady and Ranging Persistence by Yann-Situ Gazull (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-07911.txt.
Introduction
Yann-Situ Gazull extends two flavors of persistent homology, steady and ranging, from graphs to general objects, and checks when the summaries stay stable. Persistent homology tracks how features of data appear and vanish across scales. The steady and ranging variants were built to follow a chosen feature along graph filtrations. Writing in twenty twenty-five, Gazull rebuilds them with category theory so they apply far beyond graphs. The study then investigates stability: small changes in the input should move the summary only slightly. A characterization identifies exactly which features induce balanced behavior in both persistence types.
A working implementation for hypergraphs illustrates the machinery. Hypergraphs, where edges join many vertices, are a natural test case beyond ordinary graphs. The results widen topological data analysis while keeping its reliability guarantees in view.
Catalog record
| CREATOR | Yann-Situ Gazull |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 94 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-07911
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Stability and Extension of Steady and Ranging Persistence Yann-Situ Gazull Abstract Persistent homology is a topological data analysis tool that has been widely generalized, extending its scope outside the field of topology. Among its extensions, steady and ranging persistence was developed to study a wide variety of graph properties. Precisely, given a feature of interest on graphs, it is possible to build two types of persistence (steady and ranging persistence) that follow the evolution of the feature along graph filtrations. This study extends steady and ranging persistence to other objects using category theory and investigates the stability of such persistence. In particular, a characterization of the features that induce balanced steady and ranging persistence is provided. The main results of this study are illustrated using a practical implementation for hypergraphs. Keywords: persistence, categorical persistence, generalized persistence, hypergraph, stability 1 Introduction Topological data analysis has been extensively developed in recent years. Persistent homology, one of its main tools, has become a significant research topic. For this reason, persistent homology has been extended in several ways. A natural extension of persistent homology is to ease the restriction that a filtration is an increasing sequence of objects. One such extension is multidimensional persistent homology ( [ 10 , 6 ] ), which aims to generalize persistent homology to multidimensional filtrations and is currently a fruitful topic with applications in topological data analysis. Another extension is zigzag persistence ( [ 11 ] ), where the maps between two consecutive spaces in the filtration can point in either direction. A theoretical generalization of persistent homology was to theoretically develop the concept of persistence. This induced the generalization of persistence using category theory ( [ 9 , 8 , 25 ] ). Within this new paradigm, concepts related to homology have been replaced by other abstract categorical concepts, and the focus has moved from homology to the persistence function itself. From this viewpoint, persistence is seen as a process that takes a filtration and produces a persistence diagram by analyzing how a specific concept persists along the filtration. Another extension was driven by the will to use a persistence diagram on objects other than topological sets and for features other than homology, typically to study certain properties of graphs. An approach is to build a topological complex whose homology corresponds to the characteristics studied (see, for example, [ 18 , 17 , 19 , 7 ] ). Another approach is to use the aforementioned categorical persistence framework to study persistent features without using homology and topological objects. An example is the study of persistent hypergraph attributes in [ 22 , 23 ] and in [ 1 ] . Rank-based persistence ( [ 5 ] ) and graph persistence ( [ 2 ] ) were also developed in this context, as well as steady and ranging persistence ( [ 3 ] ), which constitutes the starting point of the present work. Steady and ranging persistence is a method developed by Bergomi et al. to study the evolution of features over graphs (or digraphs) along a filtration. Specifically, given a feature on graphs (e.g., Eulerian sets or hubs in [ 3 ] ), they define two ways of studying the persistence of this feature, namely steady and ranging persistence. In their work, they presented some examples of graph features and studied the stability of steady and ranging persistence induced by these features. Our theoretical work is twofold. First, we use category theory to extend the steady and ranging persistence paradigm to a larger variety of objects (e.g., hypergraphs). Second, we address the problem of stability in the context of steady and ranging persistence. Precisely, we extend the notion of balancedness and discuss its relationship with stability. We then provide a characterization of features that induce balanced steady (respectively, ranging) persistence (see our main result Theorem 4.8 ). The remainder of this paper is organized as follows. In Section 2 , a mathematical background is introduced. Precisely, we present categorical persistence, ip-generators and steady and ranging persistence for graphs. Then we extend the scope of steady and ranging persistence in Section 3 . In particular, we investigate to what extent balancedness is equivalent to stability in this context. In Section 4 , convex features are introduced, and our main result (which is the characterization of balanced features) is established. Finally, in Section 5 , we present experiments on hypergraphs to illustrate our two theoretical contributions. 2 Related Works This section introduces previous works that are important for the present work, and is organized as follows: we first introduce the categorical persistent framework, then provide useful definitions concerning the stability of persistence, and finally, we present the steady and ranging persistence theory for graphs, which corresponds to the starting point of our work. 2.1 Categorical persistence In the categorical persistent framework, category theory is used to define the generalization of filtrations (which are no longer growing topological objects), persistence functions (which correspond to persistence diagrams), and ip-generators (which are functions that associate a filtration with a persistence diagram). Categorical persistence has been formalized in various studies (see, for example, [ 9 ] ). Here we follow the formalism used in [ 3 ] . In this study, we do not introduce category theory. See [ 28 , 20 ] for a complete introduction to category theory. In the present work, we mainly consider categories where every morphism is a monomorphism. Such categories will be called mono categories . Given a category \mathbf{C} , we denote \mathbf{C}_{m} as the category whose objects are the objects of \mathbf{C} but whose morphisms are the monomorphisms of \mathbf{C} . In particular, we write \mathbb{R}_{m} for the category whose objects are real numbers and arrows correspond to the \leq relation (usually denoted (\mathbb{R},\leq) ). We use the notation X\overset{\iota}{\rightarrowtail}X^{\prime} for a monomorphism \iota from X to X^{\prime} . Definition 1 (persistence function [ 3 , Definition 1] ) . A persistence function is a function p:\mathrm{morph}(\mathbb{R}_{m})\to\mathbb{Z} that satisfies the following properties: for every u_{1}\leq u_{2}\leq v_{1}\leq v_{2} : 1. p(u_{1}\leq v_{1})\leq p(u_{2}\leq v_{1}) 2. p(u_{2}\leq v_{2})\leq p(u_{2}\leq v_{1}) 3. p(u_{2}\leq v_{1})-p(u_{1}\leq v_{1})\geq p(u_{2}\leq v_{2})-p(u_{1}\leq v_{2}) Persistence functions are as informative as persistence diagrams. Indeed, persistence functions give rise - and can be recovered from - persistence diagrams (see Figure 1 ). See [ 13 ] for the formal definition of a persistence diagram. To recover the multiset definition of a persistence diagram, we can define the multiplicity as follows: Definition 2 ( [ 5 ] , Definition 5) . Given u<v\in\mathbb{R}\cup\{-\infty,+\infty\} we define the multiplicity \mu(u,v) of (u,v) as the minimum of the following expression, over I_{u} , I_{v} disjoint connected neighborhoods of u and v respectively: p(\sup I_{u}\leq\inf I_{v})-p(\inf I_{u}\leq\inf I_{v})-p(\sup I_{u}\leq\sup I_{v})+p(\inf I_{u}\leq\sup I_{v}) A point (u,v) is said to be a cornerpoint of p if \mu(u,v)>0 . By convention we also define \mu(u,u)=+\infty . Definition 3 (bottleneck distance, [ 12 ] ) . Given two persistence functions p and p^{\prime} , we denote Dgm(p) and Dgm(p^{\prime}) their corresponding diagram seen as multisets and \Gamma the set of all bijections from Dgm(p) to Dgm(p^{\prime}) . The bottleneck distance d_{B}(p,p^{\prime}) between p and p^{\prime} of is defined as d_{B}(p,p^{\prime})=\inf_{\gamma\in\Gamma}\sup_{c\in Dgm(p)}||c-\gamma(c)||_{\infty} Figure 1 : A