Fundamentals of Theory of Continuous Gromov–Hausdorff distance.

Semeon A. Bogaty, Alexey A. Tuzhilin · 2025

Read Fundamentals of Theory of Continuous Gromov–Hausdorff distance. by Semeon A. Bogaty, Alexey A. Tuzhilin (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-2512-02611.txt.

Introduction

Semeon A. Bogaty and Alexey A. Tuzhilin extend a classic way to compare shapes and spaces. The Gromov-Hausdorff distance measures how far two metric spaces are from matching exactly. Shape comparison tools across geometry and data science rest on this idea.

Their continuous version also accounts for topological differences, not just distances. Two spaces can sit close in distance yet differ in how they connect. A distance that sees both gaps gives a fuller test of sameness. The paper lays foundations for this refined measure, which may serve geometry, imaging, and shape analysis. A sharper yard stick for shape helps fields as far apart as drug scans and parts checks for planes.

Catalog record

CREATORSemeon A. Bogaty, Alexey A. Tuzhilin
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE90 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-02611

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Fundamentals of Theory of Continuous Gromov–Hausdorff distance. Semeon A. Bogaty, Alexey A. Tuzhilin Abstract The Gromov–Hausdorff distance (hereinafter referred to as the GH-distance) is a measure of non-isometricity of metric spaces. In this paper, we study a modification of this distance that also takes topological differences into account. The resulting function of pairs of metric spaces is called the continuous GH-distance. We show that many basic properties of the classical GH-distance also hold in the continuous case. However, the continuous GH-distance, distinguishing between topologies, can differ significantly from the classical one. We will provide numerous examples of this distinction and demonstrate the role of topological dimension here. In particular, we will prove that the continuous GH-distance, like the classical one, is intrinsic, but, unlike the classical one, it is incomplete. Since we are dealing with all metric spaces, we will show, within the framework of the von Neumann-Bernays-Gödel set theory, how topological concepts can be transferred to proper classes. Keywords : metric space, Hausdorff distance, Gromov–Hausdorff distance, topological dimension, small and large inductive dimensions, hyperspace, continuum. 1 Introduction. This paper develops the theory of a variation on Gromov–Hausdorff distance geometry. Classical Gromov–Hausdorff distance measures the difference between two metric spaces: if the spaces are isometric, then they are at zero distance from each other, and the greater the distance, the greater their metric differences. The classical definition of the Gromov-Hausdorff distance does not take into account additional structures that metric spaces may possess. Even the topology generated by the metric is ignored by this distance. This feature of the Gromov-Hausdorff distance was modified, for example, by Rieffel [ 1 ] , who proposed a more subtle comparison of the so-called quantum metric spaces. In the paper citeLimMemoliSmith, another modification of the Gromov-Hausdorff distance was proposed that takes continuity into account. It was noted that comparing spheres in Euclidean space, endowed with the standard intrinsic metric, the classical Gromov-Hausdorff distance differs from those obtained with its continuous analog. Another version of the continuous Gromov–Hausdorff distance was proposed by [ 3 ] for comparing dynamical systems and solutions of partial differential equations. However, these authors’ approach significantly complicates the technique, since their version of the distance does not satisfy the triangle inequality. We decided to use the definition from [ 2 ] and begin a more fundamental study of the continuous Gromov–Hausdorff distance. We do not restrict ourselves to compact metric spaces, which constitute the traditional domain of the classical Gromov–Hausdorff distance, but consider all spaces, which also leads us to some difficulties associated with the paradoxes of set theory. For the sake of completeness, we present in the supplementary Section 11 a number of technical results that make it possible to work with such ‘‘monsters’’ as proper classes in the sense of von Neumann–Bernays–Gödel [ 4 , 5 ] . To formulate the main results of this paper, we begin with a reminder of the necessary definitions. Let (X,\rho) be a metric space, and x , y be its points, then |xy|=\rho(x,y) will denote the distance between these points, and if A and B are nonempty subsets of X , then we put |AB|=|BA|=\inf\bigl\{|ab|:a\in A,\,b\in B\bigr\} . If A=\{a\} , then instead of \bigl|\{a\}B\bigr|=\bigl|B\{a\}\bigr| we will write |aB|=|Ba| . For a positive real s and a non-negative real r we define • the open ball with center a\in X and radius s as U_{s}(a)=\{x\in X:|xa|<s\}; • the open s -neighborhood of non-empty A\subset X as U_{s}(A)=\{x\in X:|xA|<s\}; • the closed ball with center a\in X and radius r as B_{r}(a)=\{x\in X:|xa|\leq r\}; • the closed r -neighborhood of non-empty A\subset X as B_{r}(A)=\{x\in X:|xA|\leq r\}. For non-empty subsets A,B\subset X , the value d_{H}(A,B)=\max\bigl\{\sup_{a\in A}|aB|,\,\sup_{b\in B}|Ab|\bigr\} is called the Hausdorff distance between A and B . An equivalent definition is: d_{H}(A,B)=\inf\bigl\{r:A\subset B_{r}(B)\ \ \text{and}\ \ B\subset B_{r}(A)\bigr\}. It is well known that d_{H} is a generalized pseudometric on the set of all non-empty subsets of a metric space X . Here, the word ‘‘generalized’’ means that d_{H} may be equal to infinity on some pairs of subsets (for example, on a bounded and an unbounded set), and the word ‘‘pseudometric’’ means that d_{H} may be equal to zero on some pairs of distinct subsets (for example, on a non-closed set and its closure). M. Gromov defined the distance between two non-empty metric spaces [ 6 , 7 , 8 ] . The Gromov–Hausdorff distance d_{GH}(X,Y) is the greatest lower bound of the Hausdorff distances between the images of the spaces X and Y under their isometric embeddings in all possible metric spaces. In this paper, we study a continuous analogue of the Gromov–Hausdorff distance d_{GH}^{c}(X,Y) (Formula ( 7 )). We show that • the continuous Gromov–Hausdorff distance d_{GH}^{c}(X,Y) is a generalized pseudometric 1 1 1 In [ 3 ] , some analogue of the continuous Gromov–Hausdorff distance is also considered, but there this distance does not satisfy the triangle inequality, which leads to numerous technical difficulties. Instead of that, we follow the approach of [ 2 ] . (Proposition 3.1 ) and it dominates the Gromov–Hausdorff metric d_{GH}(X,Y)\leq d_{GH}^{c}(X,Y) (Property ( 2 ) of Proposition 3.2 ); • for zero-dimensional metric spaces these distances coincide (Corollary 4.6 ), and the continuous distance from a connected space X to any zero-dimensional space is not less than the diameter of this connected space (Proposition 4.9 ); • although every continuum in the Euclidean space \mathbb{R}^{n} , n\geq 2 , is at zero distance from the set of all pseudo-arcs (Bing’s Theorem [ 9 ] ), its continuous Gromov–Hausdorff distance to the set of all pseudo-arcs is equal to half the diameter of this continuum (Proposition 8.5 ); • Cook’s continuum provides a contrasting example of the difference between the Hausdorff and Gromov–Hausdorff metrics on one hand and the continuous Gromov–Hausdorff metric on the other (Corollary 8.9 ); • a metric space Y at zero continuous Gromov–Hausdorff distance from a sphere S^{n} , n\geq 1 , with an arbitrary metric, is isometric to this sphere (Corollary 5.5 ). But for every metric on the sphere S^{n} , n\geq 1 , there are at least continuum pairwise non-isometric metric spaces at zero classical Gromov–Hausdorff distance from this sphere S^{n} (for example, obtained from the sphere by discarding of finite sets of points that are pairwise nonisometric to each other). We also show that, unlike the classical Gromov–Hausdorff distance [ 10 , 11 , 12 ] , the continuous Gromov–Hausdorff distance is not complete (Theorem 10.5 ). Here it should be emphasized that all metric spaces do not form a set, and therefore do not form a topological space. For complete rigor of exposition, a transition to the theory of von Neumann–Bernays–Gödel classes [ 4 ] is necessary. Taking this into account, the concept of completeness of the (continuous) Gromov–Hausdorff distance is well defined [ 10 ] . The assertion we proved that the continuous Gromov–Hausdorff distance is intrinsic (Theorem 9.1 ) also has a correct meaning. However, in order to remain within the framework of set theory, one can only consider metric spaces of some cardinalities bounded from above by some fixed one. The work of A.A. Tuzhilin was supported by grant No. 25-21-00152 of

Read the complete text