pith. sign in

arxiv: 2404.14792 · v2 · submitted 2024-04-23 · 🧮 math.CO · cs.DM· cs.DS

α_i-Metric Graphs: Hyperbolicity

Pith reviewed 2026-05-24 02:22 UTC · model grok-4.3

classification 🧮 math.CO cs.DMcs.DS
keywords α_i-metric graphshyperbolicityGromov hyperbolicitymetric propertyshortest pathsgraph distancecombinatorics
0
0 comments X

The pith

α_i-metric graphs are f(i)-hyperbolic for a function f linear in i, and α_1-metric graphs are exactly 1-hyperbolic.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that graphs obeying the α_i-metric property, a relaxation requiring that two shortest paths sharing a terminal edge satisfy an approximate additivity condition with slack i, must have hyperbolicity bounded by a linear function of i. A reader would care because this places the new class inside the framework of hyperbolic graphs, which support efficient algorithms and structural results due to their tree-like geometry. When i equals 1 the bound tightens to exactly 1-hyperbolic and the paper proves sharpness. The authors also supply explicit 1-hyperbolic graphs that violate the α_i-metric property for every fixed i, separating the two notions. These results answer questions left from earlier algorithmic study of the same graphs.

Core claim

A graph satisfies the α_i-metric property when, for any vertices u, w, v, x, if a shortest u-w path and a shortest x-v path share the terminal edge vw then d(u,x) is at least d(u,v) plus d(v,x) minus i. The paper proves that every such graph is f(i)-hyperbolic with f linear in i, using several equivalent characterizations of hyperbolicity. In the special case i=1 the graphs are 1-hyperbolic and the constant is best possible. Simple constructions demonstrate that the converse fails: there exist 1-hyperbolic graphs that are not α_i-metric for any constant i.

What carries the argument

The α_i-metric property: whenever two shortest paths share a terminal edge vw the four-point distance inequality holds with additive error at most i.

If this is right

  • α_i-metric graphs are f(i)-hyperbolic for some linear function f of i.
  • α_1-metric graphs are 1-hyperbolic and this bound is sharp.
  • There exist 1-hyperbolic graphs that fail to be α_i-metric for every constant i.
  • Results known for bounded-hyperbolicity graphs transfer to α_i-metric graphs with dependence on i.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Multiple independent proofs via different hyperbolicity definitions indicate the implication is stable across characterizations.
  • The strict separation shows α_i-metric is a proper strengthening of hyperbolicity that can be verified locally via the shared-edge condition.
  • The linear growth in i supplies a tunable parameter for controlling how tree-like the metric must be.

Load-bearing premise

The α_i-metric distance condition on four vertices can be plugged directly into an equivalent definition of hyperbolicity without extra restrictions on the graph.

What would settle it

An explicit α_i-metric graph whose hyperbolicity grows faster than any linear function of i would falsify the claim.

Figures

Figures reproduced from arXiv: 2404.14792 by Feodor F. Dragan, Guillaume Ducoffe.

Figure 1
Figure 1. Figure 1: The triangular (2, n)-grid is α1-metric, but its 1-subdivision graph is not αi-metric for any i ≤ 4n − 5. We now give a direct proof that 1-subdivision graphs of αi-metric graphs have interval thinness at most linear in i. Lemma 4. If G = (V, E) is an αi-metric graph, then κ(Σ(G)) ≤ 2i + 12. 4 [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Thinness in graphs. It is known [10, Proposition 3.1] that in δ-hyperbolic graphs all geodesic triangles are 4δ￾thin. Conversely [20, Lemma 3.1], if all geodesic triangles in a graph are δ-thin, then it is δ￾hyperbolic. Next we show that the thinness of geodesic triangles in αi-metric graphs, and so, their hyperbolicity, is at most 3(i + 1). Theorem 1. All geodesic triangles in an αi-metric graph G are 3(i… view at source ↗
Figure 3
Figure 3. Figure 3: Graph G3 in the proof of Lemma 7. See [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: (a) An α1-metric graph G with interval thinness 2. (b) The injective hull of G which is not α1-metric (contains a C4). We recall the following properties of injective hulls. Lemma 8 ([32]). A graph G is δ-hyperbolic if and only if its injective hull H(G) is δ-hyperbolic. Lemma 9 ([23]). If H is Helly and its interval thinness is at most τ , then it is  τ 2  -hyperbolic. Lemma 10 ([24]). If H is the injec… view at source ↗
Figure 5
Figure 5. Figure 5: A ladder with height ℓ. It is a 1-hyperbolic graph but satisfies αi-metric only for i = 2ℓ as v ∈ I(u, w), w ∈ I(v, x) and 1 = d(u, x) = d(u, v) + d(v, x) − 2ℓ. 4 Graphs with α1-metric By Theorem 2, every α0-metric graph must be 1-hyperbolic, and similarly every α1-metric graph must be 2-hyperbolic. In fact [44, Corollary 30], the hyperbolicity of an α0-metric graph is at most 1 2 . In this section, we imp… view at source ↗
Figure 6
Figure 6. Figure 6: Forbidden isometric subgraph W++ 6 . The next lemma characterizes the situation in which, in an α1-metric graph, the union of two shortest paths with one common terminal edge is not a shortest path. Lemma 13 ([3]). Let G be an α1-metric graph. Let x, y, v, u be vertices of G such that v ∈ I(x, y), x ∈ I(v, u), and x and v are adjacent. Then d(u, y) = d(u, x) + d(v, y) holds if and only if there exist a nei… view at source ↗
Figure 7
Figure 7. Figure 7: To the proof of Lemma 16: construction of a forbidden W++ 6 . – By Lemma 3, d(x, w) ≤ 2 + d(y, w) = rw + 3. Therefore, d(x, y) = 2 and d(x, w) = rw + 3. Let z ∈ N(x) ∩ N(y). By Theorem 3, z ∈ Sru (u, v). – Let au ∈ N(x) ∩ Sru−1(u, v). By Lemma 3, d(au, u′ ) ≤ 2. We prove, as an intermediate claim, that d(x, u′ ) = 3 (and therefore, d(au, u′ ) = 2). Indeed, if it were not the case, then we would obtain d(x,… view at source ↗
Figure 8
Figure 8. Figure 8: To the proof of Lemma 17. Summarizing, in both cases we get wubz ∈ E. Hence, we obtain that there is a C4 with vertex-set {wub, z, y, wu}. Since wuz /∈ E (else, d(u, z) ≤ ru), by Lemma 12, we have wuby ∈ E. As a result, wub ∈ S1(y, b) ⊆ Srb (b, a). By construction, d(u, wub) ≤ ru. Since wub ∈/ Sru (u, v), d(v, wub) = rv + 1. We replace z by wub in the above, and we invert the respective roles of u and v. D… view at source ↗
Figure 9
Figure 9. Figure 9: Case d(wu, w′ u ) = 2. Since we have d(w ′ ua, wub) = 3, one of the edges w ′ uac, wubc must be missing. By symmetry, let us assume w ′ ua, c are non-adjacent. Claim. d(w ′ ua, wu) = 3. Suppose d(w ′ ua, wu) < 3. Then, d(w ′ ua, wu) = 2 (else, d(a, b) ≤ d(a, w′ ua)+1+d(wu, b) = ra+rb+2). Let w ′ ∈ N(w ′ ua) ∩ N(wu) be arbitrary. Observe that (w ′ ua, w′ , wu, wub) and (w ′ ua, x, y, wub) are shortest paths… view at source ↗
Figure 10
Figure 10. Figure 10: Case d(wu, w′ u ) = d(wv, w′ v ) = 1. Let α ∈ N(wu) ∩ N(w ′ u ) ∩ Sru−2(u, v) and β ∈ N(wub) ∩ N(wvb) ∩ Srb−1(b, a), that both exist by Corollary 3. By the choice of α, β, we have N(α) ∩ Y = {wu, w′ u} and N(β) ∩ Y = {wub, wvb}. Therefore, X = {α, β} ∪ Y induces a W++ 6 . Moreover, – d(α, β) ≥ d(u, b) − d(u, α) − d(b, β) = ru + rb − (ru − 2) − (rb − 1) = 3; – d(α, wvb) ≥ d(u, wvb) − d(u, α) = ru + 1 − (ru… view at source ↗
Figure 11
Figure 11. Figure 11: Forbidden isometric subgraphs for 1/2-hyperbolic graphs. Thus, it follows from Theorem 4 and Theorem 5 that the class of α1-metric graphs is a superclass of 1/2-hyperbolic graphs and a subclass of 1-hyperbolic graphs. Theorem 5 describes the place of 1/2-hyperbolic graphs within the class of α1-metric graphs in terms of forbidden isometric subgraphs. Using Theorem 3, Theorem 4 and a recent characterizatio… view at source ↗
Figure 12
Figure 12. Figure 12: The graphs P T, P P1 and P P2. Corollary 4. A graph G is α1-metric if and only if G is 1-hyperbolic, has no isometric C4, C6, C7, W++ 6 or P T, every induced subgraph isomorphic to P P1 has diameter at most 3 and every induced subgraph isomorphic to P P2 has diameter 2 (see [PITH_FULL_IMAGE:figures/full_fig_p018_12.png] view at source ↗
read the original abstract

A graph is called $\alpha_i$-metric ($i \in {\cal N}$) if it satisfies the following $\alpha_i$-metric property for every vertices $u, w, v$ and $x$: if a shortest path between $u$ and $w$ and a shortest path between $x$ and $v$ share a terminal edge $vw$, then $d(u,x) \ge d(u,v) + d(v,x) - i$. The latter is a discrete relaxation of the property that in Euclidean spaces the union of two geodesics sharing a terminal segment must be also a geodesic. Recently in (Dragan & Ducoffe, WG'23) we initiated the study of the algorithmic applications of $\alpha_i$-metric graphs. Our results in this prior work were very similar to those established in (Chepoi et al., SoCG'08) and (Chepoi et al., COCOA'18) for graphs with bounded hyperbolicity. The latter is a heavily studied metric tree-likeness parameter first introduced by Gromov. In this paper, we clarify the relationship between hyperbolicity and the $\alpha_i$-metric property, proving that $\alpha_i$-metric graphs are $f(i)$-hyperbolic for some function $f$ linear in $i$. We give different proofs of this result, using various equivalent definitions to graph hyperbolicity. By contrast, we give simple constructions of $1$-hyperbolic graphs that are not $\alpha_i$-metric for any constant $i$. Finally, in the special case of $i=1$, we prove that $\alpha_1$-metric graphs are $1$-hyperbolic, and the bound is sharp. By doing so, we can answer some questions left open in (Dragan & Ducoffe, WG'23).

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The manuscript defines α_i-metric graphs via a relaxed geodesic-union condition on shortest paths sharing a terminal edge. It proves that every α_i-metric graph is f(i)-hyperbolic for a function f that is linear in i, supplying independent derivations that invoke each of the standard equivalent characterizations of hyperbolicity (four-point condition, thin triangles, etc.). For the special case i=1 the bound is shown to be exactly 1 and sharp. The paper also supplies explicit constructions of 1-hyperbolic graphs that fail to be α_i-metric for any fixed i, thereby separating the two notions, and resolves several questions left open in the authors' prior WG'23 work on algorithmic applications.

Significance. The results give a clean, quantitative link between two distinct notions of tree-likeness that both support efficient algorithms on graphs. Because the proofs are obtained directly from the standard metric characterizations without additional restrictions on the graph class, and because the i=1 case is shown to be tight, the work supplies a precise dictionary that can be used to transfer algorithmic results between the α_i-metric and hyperbolic settings.

minor comments (2)
  1. [Abstract] The abstract states that f is 'linear in i' but does not record the explicit coefficient; if an explicit expression for f appears in the body (e.g., in the proof that invokes the four-point condition), a parenthetical reference to that statement would help readers locate the precise bound.
  2. [Introduction] The citation to Dragan & Ducoffe (WG'23) is used both for the algorithmic motivation and for the open questions that are now settled; a single consolidated sentence in the introduction that lists exactly which questions are answered would improve readability.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive summary, significance assessment, and recommendation to accept. The report contains no major comments requiring response.

Circularity Check

0 steps flagged

No significant circularity; direct derivations from definitions

full rationale

The paper supplies multiple independent proofs deriving the linear-in-i hyperbolicity bound directly from the α_i-metric property by invoking each of the standard equivalent characterizations (four-point condition, thin triangles, etc.). These derivations are explicitly constructed in the text without any reduction to fitted parameters, self-definitional loops, or load-bearing self-citations. The self-citation to Dragan & Ducoffe (WG'23) addresses only prior algorithmic results and open questions answered here; the hyperbolicity proofs stand on the metric definitions themselves. Separate constructions showing 1-hyperbolicity does not imply bounded α_i confirm the notions are properly distinguished. The derivation chain is therefore self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the newly introduced α_i-metric definition together with standard background equivalences among hyperbolicity characterizations; no free parameters or new entities are introduced.

axioms (1)
  • standard math Equivalence of different definitions of graph hyperbolicity
    Invoked to give multiple proofs of the implication from α_i-metric to bounded hyperbolicity.

pith-pipeline@v0.9.0 · 5871 in / 1194 out tokens · 48767 ms · 2026-05-24T02:22:13.209625+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

45 extracted references · 45 canonical work pages

  1. [1]

    Abu-Ata, & F.F

    M. Abu-Ata, & F.F. Dragan. Metric tree-like structures in real-world networks: an empirical study, Networks 67(1) (2016), 49-68

  2. [2]

    Albert, B

    R. Albert, B. DasGupta, & N. Mobasheri. Topological implications of negative curvature for biological and social networks, Physical Review E 89(3) (2014), 032811

  3. [3]

    Bandelt, V

    H.-J. Bandelt, V. Chepoi, 1-Hyperbolic graphs, SIAM J. Discr. Math. 16 (2003), 323–334

  4. [4]

    Bandelt, V

    H.-J. Bandelt, V. Chepoi, Metric graph theory and geometry: a survey, Contemporary Mathematics 453 (2008), 49-86

  5. [5]

    Borassi, P

    M. Borassi, P. Crescenzi, and M. Habib, Into the square: On the complexity of some quadratic-time solvable problems, Electronic Notes in TCS , 322(2016), 51–67

  6. [6]

    Borassi, A

    M. Borassi, A. Chessa, & G. Caldarelli. Hyperbolicity measures democracy in real-world networks, Physical Review E 92(3) (2015), 032812

  7. [7]

    Bondy, U.S.R

    J.A. Bondy, U.S.R. Murty, Graph Theory, Graduate Texts in Mathematics , 244 (2008)

  8. [8]

    Brinkmann, J.H

    G. Brinkmann, J.H. Koolen, & V. Moulton. On the hyperbolicity of chordal graphs, Annals of Combinatorics 5(1) (2001), 61-69

  9. [9]

    Chalopin, M

    J. Chalopin, M. Changat, V. Chepoi, & J. Jacob. First-order logic axiomatization of metric graph theory, arXiv:2203.01070, 2022

  10. [10]

    Chalopin, V

    J. Chalopin, V. Chepoi, F.F. Dragan, G. Ducoffe, A. Mohammed, Y. Vax` es, Fast Approximation and Exact Computation of Negative Curvature Parameters of Graphs, Discret. Comput. Geom., 65(2021), 856–892

  11. [11]

    Chalopin and V

    J. Chalopin and V. Chepoi, U. Giocanti, Graphs with convex balls, Geom. Dedicata (2023) https://doi.org/10.21203/rs.3.rs-2293810/v1

  12. [12]

    Chalopin, V

    J. Chalopin, V. Chepoi, N. Nisse, Y. Vax` es, Cop and robber games when the robber can hide and ride, SIAM J. Discrete Math. 25 (2011) 333–359

  13. [13]

    Chalopin, V

    J. Chalopin, V. Chepoi, P. Papasoglu, T. Pecatte, Cop and robber game and hyperbolicity, SIAM J. Discrete Mathematics 28 (2014), 1987–2007

  14. [14]

    Chepoi, Some d-convexity properties in triangulated graphs, in Mathematical Research, vol

    V. Chepoi, Some d-convexity properties in triangulated graphs, in Mathematical Research, vol. 87, S ¸tiint ¸a, Chi¸ sin˘ au, 1986, pp. 164–177 (Russian)

  15. [15]

    Chepoi, Centers of triangulated graphs, Math

    V. Chepoi, Centers of triangulated graphs, Math. Notes 43 (1988), 143–151

  16. [16]

    Chepoi and F

    V. Chepoi and F. F. Dragan, Finding a central vertex in an HHD-free graph, DAM, 131(1) (2003), 93–111

  17. [17]

    Chepoi, F.F

    V. Chepoi, F.F. Dragan, B. Estellon, M. Habib, Y. Vax´ es, & Y. Xiang. Additive spanners and distance and routing labeling schemes for hyperbolic graphs, Algorithmica 62 (2012), 713-732

  18. [18]

    Chepoi, F.F

    V.D. Chepoi, F.F. Dragan, B. Estellon, M. Habib and Y. Vax` es, Diameters, centers, and approximating trees of δ-hyperbolic geodesic spaces and graphs, Proceedings of the 24th Annual ACM Symposium on Computational Geometry (SoCG 2008) , June 9-11, 2008, College Park, Maryland, USA, pp. 59-68

  19. [19]

    Chepoi, F.F

    V. Chepoi, F.F. Dragan, M. Habib, Y. Vax` es, and H. Alrasheed, Fast approximation of eccentricities and distances in hyperbolic graphs, Journal of Graph Algorithms and Applications , 23(2019), 393–433

  20. [20]

    Chepoi, F.F

    V.D. Chepoi, F.F. Dragan, & Y. Vax´ es. Core congestion is inherent in hyperbolic networks, in Proceedings of the Twenty-Eighth Annual ACM-SIAM Symposium on Discrete Algorithms (SODA) , Society for Industrial and Applied Mathematics (2018), pp. 2264-2279

  21. [21]

    Coudert, G

    D. Coudert, G. Ducoffe. Recognition of C4-free and 1/2-hyperbolic graphs, SIAM Journal on Discrete Mathe- matics 28(3) (2014), 1601-1617. 20

  22. [22]

    Dragan, and G

    F.F. Dragan, and G. Ducoffe, αi-Metric Graphs: Radius, Diameter and all Eccentricities, WG 2023, Lecture Notes in Computer Science 14093, 276–290

  23. [23]

    Dragan, and H.M

    F.F. Dragan, and H.M. Guarnera, Obstructions to a small hyperbolicity in Helly graphs, in Discrete Mathe- matics, vol. 342.2, 2019, pp. 326-338

  24. [24]

    Dragan, and H.M

    F.F. Dragan, and H.M. Guarnera, Helly-gap of a graph and vertex eccentricities, in Theoretical Computer Science, vol. 867, 2021, pp. 68-84

  25. [25]

    Dragan, E

    F.F. Dragan, E. K¨ ohler, H. Alrasheed, Eccentricity Approximating Trees, Discrete Applied Mathematics , 232 (2017), 142–156

  26. [26]

    Dragan, A

    F.F. Dragan, A. Mohammed, Slimness of graphs, Discret. Math. Theor. Comput. Sci. 21(3) (2019)

  27. [27]

    Dress, Trees, tight extensions of metric spaces, and the cohomological dimension of certain groups: A note on combinatorial properties of metric spaces, Adv

    A.W.M. Dress, Trees, tight extensions of metric spaces, and the cohomological dimension of certain groups: A note on combinatorial properties of metric spaces, Adv. Math. , 53 (1984), pp. 321–402

  28. [28]

    Ducoffe and F.F

    G. Ducoffe and F.F. Dragan, A story of diameter, radius, and (almost) Helly property, Networks, 77(2021), 435-453

  29. [29]

    Fomin, P

    F.V. Fomin, P. Golovach, and J. Kratochv´ ıl, On tractability of cops and robbers game, in5th Ifip International Conference On Theoretical Computer Science, vol. 273 (2008), pp. 171-185, Springer, Heidelberg

  30. [30]

    Fomin, P

    F.V. Fomin, P. Golovach, J. Kratochv´ ıl, N. Nisse, and K. Suchan, Pursuing a fast robber on a graph, Theor. Comput. Sci. 411 (2010), 1167–1181

  31. [31]

    Gromov, Hyperbolic Groups, pages 75–263

    M. Gromov, Hyperbolic Groups, pages 75–263. Springer, New York, NY, 1987

  32. [32]

    Guarnera, F.F

    H.M. Guarnera, F.F. Dragan, and A. Leitert. Injective hulls of various graph classes, in Graphs and Combina- torics, vol. 38.4, 2022, 112

  33. [33]

    Howorka, A characterization of distance-hereditary graphs, Quart

    E. Howorka, A characterization of distance-hereditary graphs, Quart. J. Math. Oxford Ser. 28 (1977), 417–420

  34. [34]

    Howorka, On metric properties of certain clique graphs, Journal of Combinatorial Theory, Series B 27(1) (1979), 67–74

    E. Howorka, On metric properties of certain clique graphs, Journal of Combinatorial Theory, Series B 27(1) (1979), 67–74

  35. [35]

    Jonckheere, P

    E. Jonckheere, P. Lohsoonthorn. Geometry of network security, in Proceedings of the American Control Con- ference, IEEE, Vol. 2 (2004), pp. 976-981

  36. [36]

    W. S. Kennedy, I. Saniee, & O. Narayan. On the hyperbolicity of large-scale networks and its estimation, in IEEE International Conference on Big Data (Big Data), IEEE (2016), pp. 3344-3351

  37. [37]

    Koolen, V

    J.H. Koolen, V. Moulton. Hyperbolic bridged graphs, European Journal of Combinatorics23(6) (2002), 683-699

  38. [38]

    de Montgolfier, M

    F. de Montgolfier, M. Soto, L. Viennot, Treewidth and Hyperbolicity of the Internet. NCA 2011: 25–32

  39. [39]

    Nisse and K

    N. Nisse and K. Suchan, Fast robber in planar graphs, in Proceedings of the 34th International Workshop on Graph-Theoretic Concepts in Computer Science (WG), Springer LNCS vol. 5344 (2008), pp. 312–323

  40. [40]

    Nowakowski and P

    R.J. Nowakowski and P. Winkler, Vertex-to-vertex pursuit in a graph, Discr. Math. 43 (1983), 235–239

  41. [41]

    Papasoglu

    P. Papasoglu. Strongly geodesically automatic groups are hyperbolic, Inventiones Math. 121 (1995), 323– 334

  42. [42]

    Quilliot, A short note about pursuit games played on a graph with a given genus, J

    A. Quilliot, A short note about pursuit games played on a graph with a given genus, J. Comb. Theory, Ser. B 38 (1985), 89–92

  43. [43]

    Soltan, V.D

    V.P. Soltan, V.D. Chepoi, Conditions for invariance of set diameters under d-convexification in a graph, Cy- bernetics 19 (1983), 750–756 (Russian, English transl.)

  44. [44]

    Y. Wu, C. Zhang. Hyperbolicity and chordality of a graph, The Electronic Journal of Combinatorics 18(1) (2001), P43

  45. [45]

    Yushmanov, V

    S.V. Yushmanov, V. Chepoi, A general method of investigation of metric graph properties related to the eccentricity, in Mathematical Problems in Cybernetics , vol. 3, Nauka, Moscow, 1991, pp. 217–232 (Russian). 21