
Combining network geometry with Bayesian inference
Ariane Lizotte
2025-09-01
Vertices are placed in metric space. The edge cost increases with its length.

The metric space can be physical (e.g. transportation network, brain network) or not.
Graphs obtained from hyperbolic space reproduce many empirically observed properties.
Krioukov, D., et al. Hyperbolic geometry of complex networks. Phys. Rev. E 82, 036106 (2010).
Zuev, K., et al. Emergence of Soft Communities from Geometric Preferential Attachment. Sci. Rep. 5, 9421 (2015).
Krioukov, D., et al. Clustering Implies Geometry in Networks. Phys. Rev. Lett. 116, 208302 (2016).
The connection probability is
\(\mathbb{S}^1\) model
\[\begin{align} p_{uv} = \frac{1}{1+\left(\frac{d(\theta_u, \theta_v)}{\mu \kappa_u, \kappa_v}\right)^\beta} \end{align}\]
where
We want to represent a given graph using a hyperbolic embedding.
This amounts to:
\[ \text{Pairs of vertices are} \begin{cases} \text{close} & \text{if connected in graph,} \\ \text{far} & \text{if not connected in graph.} \\ \end{cases} \]
Algorithms
yield a single embedding.
We currently ignore
We address both issues using a Bayesian approach.
The likelihood of the \(\mathbb{S}^1\) model is: \[ \mathbb{P}[G|\theta, \kappa, \beta] = \prod_{(u, v) \in E} p_{uv} \prod_{(u, v) \in E^c} (1-p_{uv}), \] where \(p_{uv}=\left[1+\left(\frac{d_\mathbb{S}(\theta_u, \theta_v)}{\mu \kappa_u, \kappa_v}\right)^\beta\right]^{-1}\). The posterior of the Bayesian \(\mathbb{S}^1\) model is
\[\begin{align} p(\theta, \kappa, \beta | G) &\propto \mathbb{P}[G|\theta, \kappa, \beta] \; p(\beta) \prod_{v \in V} p(\theta_v)p(\kappa_v), \\ \end{align}\]
where the priors are
\[\begin{align} \theta_v &\stackrel{iid}{\sim} \text{Uniform}[-\pi, \pi), \\ \kappa_v &\stackrel{iid}{\sim} \text{Cauchy}, \quad \kappa_v >\epsilon,\\ \beta &\sim \text{Normal}, \quad \beta>1. \end{align}\]
We sample embeddings from the posterior using Markov chain Monte Carlo (MCMC).

Since edges are local, groups of nearby vertices should be moved together.
Conflicting ground truth model:
A vertex \(v\) is given two positions \(\theta_v^{(1)}\) and \(\theta_v^{(2)}\).
When generating \(G\) with the \(\mathbb{S}^1\) model, each edge probability including \(v\) uses randomly \(\theta_v^{(1)}\)or \(\theta_v^{(2)}\).
Paper:
Lizotte, A., Young, J.-G. and Allard A. Symmetry-driven embedding of networks in hyperbolic space. Commun. Phys. 8, 199 (2025).
Antoine Allard
Jean-Gabriel Young
Go to the neighbour closest to the destination.
The \(\mathbb{S}^1\) model is not identifiable because of graph and space symmetries.
Comparing embeddings requires alignment.