Hyperbolic Geometric Graph¶
A graph model that positions vertices in hyperbolic space and makes edges depend on the vertices' hyperbolic distance.
Core Idea¶
A hyperbolic geometric graph places vertices in a negatively curved hyperbolic space and makes edges depend on the hyperbolic distance between vertex pairs. A hard rule links pairs within a cutoff; a soft rule assigns a distance-dependent connection probability. The defining move is the coupling of latent geometry to adjacency, not merely drawing a graph in a curved-looking disk. The original \(\mathbb{H}^{2}\) model gives explicit radial/angular placement and both edge-rule forms.[^ref-25236c359f5a]
Heavy-tailed degrees, clustering, sparse structure and effective greedy routing are possible results in specified regimes or fitted maps, not guarantees of every hyperbolic geometric graph. Parameter choices and how coordinates were obtained matter.[ref-25236c359f5a][ref-fe3aeaa6d728]
Scope of Application¶
The model supports theoretical study of network topology and comparison with observed networks. In Krioukov and colleagues' cold-disk construction, the radial-density/curvature ratio \(\alpha/\zeta\) controls a power-law exponent under stated conditions, while temperature softens the connection threshold and changes clustering behavior. An observed graph can also be mapped back into hyperbolic coordinates: Boguñá and colleagues fitted an autonomous-system Internet map and evaluated greedy forwarding on that map. Fitting is an inverse application, not proof of how the observed graph was originally generated.[ref-25236c359f5a][ref-fe3aeaa6d728]
Clarity¶
Ask whether vertices have hyperbolic coordinates and whether hyperbolic distance controls edges. Both are necessary. A Euclidean random geometric graph uses a different metric; a disk visualization with arbitrary edges is only an embedding. In the hard construction, \(p(x)=\Theta(R-x)\) gives a categorical edge decision. In a soft finite-temperature construction, \(p(x)=\bigl[1+\exp\bigl(\beta\zeta(x-R)/2\bigr)\bigr]^{-1}\) gives a probability, so two realizations can differ even with the same coordinates.[^ref-25236c359f5a]
Manages Complexity¶
Latent positions plus a distance-to-edge law replace many unconstrained pairwise link choices with a compact geometric model. This can explain why some sampled topologies exhibit hubs or clustering, but only after the placement law, radius and connection regime are specified. A good fit to observed links is informative, yet it cannot by itself prove a unique latent geometry or erase substantive causes of those links.[ref-25236c359f5a][ref-fe3aeaa6d728]
Abstract Reasoning¶
For forward generation, sample vertex coordinates in a hyperbolic disk, compute each pair distance, and apply either a hard cutoff or a declining edge probability. Changing radial placement while holding the edge rule fixed alters the distance distribution; softening the rule while holding positions fixed alters which links are certain. For inverse mapping, begin instead with observed links and infer coordinates that make them plausible. Mixing these directions produces false claims that a useful coordinate map proves literal historical generation by the model.[ref-25236c359f5a][ref-fe3aeaa6d728]
Knowledge Transfer¶
The hard and soft constructions instantiate the same hyperbolic-distance edge pattern through different link mechanisms. An Internet map illustrates another use of that family: inference and navigation on a fitted observed graph; its reported routing success is map-specific. The broader skeleton “latent metric conditions relations” is a future-prime question, not a strict live parent established here. The draft stays unparented: live Network names the general connection-pattern object, whereas this entry names a geometric graph-generating ensemble.[^ref-fe3aeaa6d728]
[^ref-25236c359f5a]: Dmitri Krioukov et al., “Hyperbolic Geometry of Complex Networks”, Physical Review E 82, 036106 (2010), Secs. IV and VI–VIII, including hard/soft connection rules and regime-dependent topology. [^ref-fe3aeaa6d728]: Marián Boguñá, Fragkiskos Papadopoulos and Dmitri Krioukov, “Sustaining the Internet with hyperbolic mapping”, Nature Communications 1, 62 (2010), author-hosted original PDF, Secs. IV–V on mapping and routing; journal version.
Neighborhood in Abstraction Space¶
Hyperbolic Geometric Graph sits in a sparse region of the domain-specific corpus (72nd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Graph Structures & Combinatorial Objects (44 abstractions)
Nearest neighbors
- Force-Directed Graph Drawing — 0.85
- Graph Embedding — 0.84
- Quantum Walk — 0.83
- Hypercube Graph — 0.83
- Metric dimension (graph theory) — 0.83
Computed from structural-signature embeddings · 2026-10-08