A Stable and Theoretically Grounded Gromov-Wasserstein Distance for Reeb Graph Comparison using Persistence Images
Erin W. Chambers, Guangyu Meng · 2025
Read A Stable and Theoretically Grounded Gromov-Wasserstein Distance for Reeb Graph Comparison using Persistence Images by Erin W. Chambers, Guangyu Meng (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-2507-01171.txt.
Introduction
Chambers and Meng build a stable way to compare Reeb graphs, which sum up the shape of a data field in twenty twenty-five. Such graphs trace how a field's level sets split and merge. Comparing two graphs calls for a distance that respects both shape and structure. Past distances were slow or missed key traits.
Their distance adapts a broad tool for comparing measured spaces. A fair form of the graph's radius handles shape. Weights from persistence images mark which parts matter most in structure. They prove small wobbles in the field cause only small shifts in the distance. Tests on several data sets beat rival methods, they report.
Catalog record
| CREATOR | Erin W. Chambers, Guangyu Meng |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 101 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-01171
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A Stable and Theoretically Grounded Gromov-Wasserstein Distance for Reeb Graph Comparison using Persistence Images Erin W. Chambers, Guangyu Meng Abstract Reeb graphs are a fundamental structure for analyzing the topological and geometric properties of scalar fields. Comparing Reeb graphs is crucial for advancing research in this domain, yet existing metrics are often computationally prohibitive or fail to capture essential topological features effectively. In this paper, we explore the application of the Gromov-Wasserstein distance, a versatile metric for comparing metric measure spaces, to Reeb graphs. We propose a framework integrating a symmetric variant of the Reeb radius for robust geometric comparison, and a novel probabilistic weighting scheme based on Persistence Images derived from extended persistence diagrams to effectively incorporate topological significance. A key contribution of this work is the rigorous theoretical proof of the stability of our proposed Reeb Gromov-Wasserstein distance with respect to perturbations in the underlying scalar fields. This ensures that small changes in the input data lead to small changes in the computed distance between Reeb graphs, a critical property for reliable analysis. We demonstrate the advantages of our approach, including its enhanced ability to capture topological features and its proven stability, through comparisons with other alternatives on several datasets, showcasing its practical utility and theoretical soundness. 1 Introduction In computational topology and topological data analysis (TDA), Reeb graphs have emerged as powerful tools for analyzing scalar fields—fundamental for understanding structured data [ 8 , 40 ] . Together with their discrete approximations, mapper graphs, they encode the changing component structure of level sets, thereby providing a compact yet expressive representation of the underlying topology. The versatility of Reeb graphs has spurred their widespread adoption in diverse applications, including shape analysis in computer graphics [ 21 , 16 ] , feature extraction in scientific visualization [ 38 , 18 ] , and as topological summaries in deep learning [ 23 , 11 ] . Despite the utility of Reeb graphs in TDA, a significant challenge lies in effectively quantifying their similarity. Numerous metrics have been developed, broadly categorized into direct calculations and compound methods. Direct calculation techniques include the bottleneck distance [ 10 , 7 ] , interleaving distance [ 13 ] , functional distortion distance [ 5 ] , Reeb graph edit distance [ 15 ] , and the universal distance [ 6 ] , each offering unique perspectives and properties [ 8 ] . Compound methods, such as Decorated Reeb Graphs (DRGs) [ 12 ] and Multi-Dimensional Reeb Graphs (MDRGs) [ 32 ] , aim to enhance Reeb graph comparisons by incorporating additional structures. DRGs, for instance, utilize barcode decorations with multiscale bottleneck and Gromov-Hausdorff distances [ 27 ] to provide robust multiscale topological summaries and improve sensitivity to local geometric variations. MDRGs, conversely, rely on bottleneck and Laplace-Beltrami-based extended bottleneck distances [ 33 ] , demonstrating strengths in capturing richer geometric and spectral properties of the underlying data. However, these existing methods face several limitations. The bottleneck distance, for example, can overlook structural context, as its value is determined by the single most prominent difference between persistence diagrams, thereby ignoring the collective information from all other topological features and their interrelationships. Other direct metrics are often computationally intractable because they may require solving complex combinatorial optimization problems [ 8 ] .Compound methods frequently suffer from scalability issues due to the high memory and computational costs associated with added decorations or operators. Furthermore, hyperparameter tuning presents a significant hurdle for these compound metrics, as achieving an effective balance among their constituent components is inherently complex. To address these limitations and offer a theoretically robust framework, we propose a novel approach for comparing Reeb graphs that leverages the Gromov-Wasserstein (GW) distance. The GW distance, a metric from optimal transport (OT) theory, is designed to compare metric measure spaces [ 22 , 29 , 27 ] . It is an appealing prospect as it captures both local and global structural differences by aligning intrinsic metric structures. Moreover, the GW distance provides node-to-node correspondences, enabling interpretable insights into graph similarities [ 27 ] . Note that our formulation, defined via the exact GW problem (which is NP-hard), is made computationally feasible by adopting a practical and standard approach in the field [ 31 , 17 ] : leveraging efficient iterative solvers. These solvers are widely recognized for providing high-quality approximate solutions in polynomial time (with respect to the number of samples), with demonstrated acceptable approximation error for many applications [ 31 , 36 ] . Our main contributions involve tailoring the GW distance framework specifically for Reeb graphs and establishing its theoretical underpinnings. First, to define the metric structure on the Reeb graph, we introduce a symmetric variant of the Reeb radius. This variant addresses the asymmetry of the traditional Reeb radius [ 12 ] and the noise sensitivity of alternatives like shortest path distances [ 24 ] , thereby providing a robust and valid distance metric between nodes. Second, for the measure component, we propose a Persistence Image (PI) [ 2 ] -based probability measure. This approach transforms the extended persistence diagram [ 10 , 3 ] associated with the scalar field into a Persistence Image [ 2 ] , from which node weights are subsequently derived. Such a measure reflects the global topological context and significance of features, offering a more nuanced weighting than uniform or simpler heuristic distributions commonly employed [ 12 , 24 ] . A cornerstone of this work is the formal proof of stability for our proposed Reeb Gromov-Wasserstein ( RGW_{p} ) distance. We demonstrate that the RGW_{p} distance between two Reeb graphs—equipped with our symmetric Reeb radius and PI-based measure—is bounded by the supremum norm difference between their generating scalar functions. This stability guarantee is crucial for the metric’s reliability in practical applications where data can be noisy or subject to minor variations. Through experiments on several benchmark datasets, we demonstrate the practical advantages of our RGW_{p} distance, including its high accuracy, enhanced ability to capture topological features, and competitive computational efficiency. The combination of a robust geometric metric, a topologically informed measure, and proven stability establishes our framework as a compelling new tool for Reeb graph analysis, showcasing its practical utility and theoretical soundness. The remainder of this paper is organized as follows: Section 2 reviews background concepts. Section 3 details our methodology, covering the symmetric Reeb radius and the PI-based probability measure. Section 4 presents the stability analysis of our RGW_{p} distance. Section 5 presents an example for the RGW_{p} workflow. Section 6 describes our experimental setup and results. Finally, Section 7 concludes the paper and discusses future work. 2 Background In this section, we provide a brief review of foundational concepts essential for understanding our work. These include scalar fields, Morse functions, Reeb graphs, various distance metrics applicable to Reeb graphs (such as the Reeb Radius, which forms a basis for our symmetric variant), the theory of extended persistence, and the construction of Persistence Images (PIs). For readers seeking a more comprehensive treatment of Reeb graphs and their associated metrics, we recommend the survey papers by Yan et al. [ 43 ] and Bollen et al. [ 8 ] . For a thorough exploration of persistent homology, the texts