---
title: Multi-Hypergraph Stochastic Block Model
url: https://www.emergentmind.com/topics/multi-hypergraph-stochastic-block-model-mhsbm
type: topic
---

# Multi-Hypergraph Stochastic Block Model

The Multi-Hypergraph Stochastic Block Model (MHSBM) is a probabilistic framework for modeling and inferring latent community structure in settings where multiple hypergraphs—potentially of heterogeneous type, node set, or domain—encode higher-order relations among elements. The model generalizes the classical stochastic block model (SBM) for graphs to collective ensembles of hypergraphs, accounting both for high-order intrarelationships (hyperedges of arbitrary cardinality) and inter-hypergraph couplings, thus providing a unifying mechanism to analyze integrative, multi-relational datasets encountered in biological, social, and information systems [2301.11657], [2505.04967].

## 1. Model Specification: Structural and Generative Elements

MHSBM operates over $L$ hypergraphs $H^{1}, \dots, H^{L}$, with each $H^{l} = (V^{l}, E^{l})$ and $E^{l}$ containing hyperedges of arbitrary cardinality. Nodes $v_{i}^{l}$ possess community memberships $u_{i}^{l} \in \Delta_{K^{l}}$ (the unit simplex in $\mathbb{R}^{K^{l}}$), allowing for mixed-membership assignments.

Each hypergraph $l$ is parameterized by an affinity matrix $w^{l} \in \mathbb{R}_{\ge 0}^{K^{l} \times K^{l}}$ dictating community-level interaction strengths, while each ordered hypergraph pair $(l, l')$ is associated with an inter-hypergraph affinity $w^{ll'} \in \mathbb{R}_{\ge 0}^{K^{l} \times K^{l'}}$. Realizations of intra-hypergraph hyperedges $A^l_{e}$ and inter-hypergraph connections $S^{ll'}_{ij}$ are modeled as independent Poisson random variables with rates defined by the corresponding membership and affinity parameters.

To capture node-specific influence in hyperedge formation and model preferential attachment, MHSBM incorporates a hyperedge-internal-degree matrix $\theta^l \in \mathbb{R}_{+}^{N^{l} \times M^{l}}$, where $\theta_{ie}^l$ quantifies node $i$’s weighted participation in hyperedge $e$. For $e$, $\sum_{i \in e} \theta^l_{ie} = |e|$ always holds.

The generative process for hyperedge weights is:
\[
A^{l}_{e} \sim \mathrm{Pois}\left( \lambda_{e}^{l} / \mu_{e}^{l} \right), \quad \lambda_{e}^{l} = \sum_{i<j \in e} \theta^{l}_{ie} (u_{i}^{l} w^{l} (u_{j}^{l})^{T}) \theta^{l}_{je},
\]
with $\mu_{e}^{l}$ the normalization for node pairs.

Cross-hypergraph edges are realized as:
\[
S^{ll'}_{ij} \sim \mathrm{Pois}(\lambda_{ij}^{ll'}), \quad \lambda_{ij}^{ll'} = u_{i}^{l} w^{ll'} (u_{j}^{l'})^{T}.
\]
This structure allows MHSBM to encode arbitrary interdependence between hypergraphs and flexible within-hypergraph connectivity, subsuming both assortative and disassortative regimes.

## 2. Likelihood, Posterior, and Variational Inference

Given observed $A, S$, the likelihood under latent parameters $\Phi=\{ u^{l}, w^{l}, w^{ll'} \}$ factorizes as:
\[
P(A,S|\Phi) = \prod_{l=1}^{L} \prod_{e \in \Omega^{l}} \mathrm{Pois}\left(A^{l}_{e};\, \lambda_{e}^{l} / \mu_{e}^{l}\right) \cdot \prod_{l<l'}\prod_{i \in V^{l}}\prod_{j \in V^{l'}} \mathrm{Pois}\left(S_{ij}^{ll'};\, \lambda_{ij}^{ll'}\right).
\]
In the maximum-likelihood setting, parameters are estimated directly. A Bayesian formulation introduces Dirichlet/Gamma priors on $u$, $w$ and yields a posterior proportional to $P(A,S|\Phi)P(\Phi)$.

Inference employs expectation–maximization (EM) with variational approximations. Due to combinatorial explosion in possible hyperedges, negative sampling is used: for each $l$, only the observed ($E^{l+}$) and sampled “negative” ($E^{l-}$) hyperedges contribute to the likelihood.

E-step updates variational weights for each hyperedge and cross-hypergraph edge; M-step computes closed-form parameter updates given current variational parameters. Multiple EM restarts are advocated to mitigate local optima [2505.04967].

## 3. Aggregate Similarity Matrix and SDP-Based Recovery

When restricting to multilayer, $d$-uniform, single-node-set hypergraphs, MHSBM reduces to the multilayer hypergraph SBM analyzed via semidefinite programming (SDP) [2301.11657]. Each layer $\ell$ possesses adjacency tensor $A^{(\ell)}$. The aggregate similarity matrix $S=W\in\mathbb{Z}^{N\times N}_{\ge0}$, with $W_{ij}$ the total number of hyperedges (across all layers) containing both $i$ and $j$, forms the weighted adjacency matrix for clustering.

Community recovery is formulated as the NP-hard quadratic program:
\[
\max_{x \in \{\pm 1\}^N,\, \mathbf{1}^T x = 0} x^T W x \quad (\text{assortative}),
\]
relaxed to an SDP over $X=xx^T$, with constraints $X_{ii} = 1$, $\mathbf{1}^T X \mathbf{1} = 0$, and $X \succeq 0$. Disassortative cases replace $W\to -W$. Exact recovery conditions are characterized in terms of an information rate parameter $I$; if $I>1$, exact recovery of the true labeling (up to global sign) occurs with high probability as $N\to\infty$ [2301.11657].

Proofs leverage dual-certificate construction, matrix concentration, and diagonal lower bounds to establish uniqueness of the SDP optimum.

## 4. Performance Evaluation and Empirical Results

Empirical assessment of MHSBM targets three axes: community detection, hyperedge reconstruction, and inter-hypergraph edge inference. Experimental datasets span single-type multi-hypergraphs (e.g., face-to-face contact, high school, hospital ward), multi-domain instances (gene–disease, arXiv co-authorship, Author–Citation), and legislative cosponsorship.

Performance is measured via F1-score, normalized mutual information (NMI), cosine similarity (CS), and area under curve (AUC) for link and edge prediction. MHSBM consistently outperforms single-layer baselines (Hy-MMSBM [Contisciani et al., 2022], Hypergraph-MT [De Bacco et al., 2022]) by 10–20% (CS), achieves AUC $>0.80$ for hyperedge prediction, and attains AUC in [0.77, 0.97] on cross-hypergraph edge prediction tasks. Removal of inter-hypergraph couplings leads to smooth degradation to the single-hypergraph baseline, evidencing the benefit of integrative modeling [2505.04967].

## 5. Extensions, Scalability, and Modeling Flexibility

MHSBM readily accommodates extensions: multi-view settings (where $V^l$ are identical but $E^l$ differ), multi-domain set-ups (heterogeneous node and edge types), dynamic or temporal multi-hypergraphs, and fully Bayesian versions with parameter priors. For more than two communities, unbalanced partitions, or non-uniform hyperedge sizes, analogous aggregate or tensor-based relaxations and inference mechanisms apply [2301.11657], [2505.04967].

When $L$ (number of hypergraphs/layers) is fixed or grows slowly ($L = o((\log N)^{1/2})$), current theoretical results on exact recovery and SDP concentration continue to hold. Computationally, SDP solvers require $O(N^6)$ (interior-point), though practical methods exploit low-rank or first-order algorithms for $O(N^3)$ cost.

MHSBM’s hyperedge-internal-degree $\theta_{ie}^l$ parameter allows nuanced modeling of node importance and preferential attachment, increasing fidelity to empirical phenomena such as “star” nodes or organizers in social/event hypergraphs. The framework handles both assortative and disassortative structures: nonzero off-diagonal $w^l$ or $w^{ll'}$ influence whether communities prefer internal or cross-type connections.

## 6. Applications and Practical Implications

MHSBM enables integrative discovery of latent communities, robust prediction of missing hyperedges, and accurate inference of cross-graph edges in complex systems. Integrating multiple hypergraphs (via $w^{ll'}$) substantially improves interpretability and predictive power over single-layer approaches, particularly in heterogeneous or multi-modal situations such as gene–protein, author–citation, or hybrid communication networks. In practice, MHSBM can serve as a foundation for high-order clustering, multi-relational link prediction, and structural analysis across domains [2301.11657], [2505.04967].

## 7. Summary Table: Key Structural Objects in MHSBM

| Object                     | Definition/Role                                                       | Domain                         |
|----------------------------|-----------------------------------------------------------------------|--------------------------------|
| $u^{l}_{i}$                | Mixed-membership vector for $v^{l}_i$                                 | $\Delta_{K^{l}}$               |
| $w^{l}$, $w^{ll'}$         | Intra- and inter-hypergraph community affinities                      | $\mathbb{R}_{\ge 0}$ matrices  |
| $\theta^{l}_{ie}$          | Internal degree of node $i$ in hyperedge $e$ (preferential weight)    | $\mathbb{R}_{+}$               |
| $A^l_{e}, S^{ll'}_{ij}$    | Hyperedge / cross-hypergraph edge random variables                    | $\mathbb{N}$                   |
| $E^{l+}, E^{l-}$           | Observed / negative-sampled hyperedges for computational tractability | Hyperedge sets                 |

The MHSBM unifies high-order community detection and link prediction with rigorous likelihood-based and spectral inference, scales to complex multi-relational settings, and is empirically validated for real-world, multigraph data sources [2301.11657], [2505.04967].

Source: https://www.emergentmind.com/topics/multi-hypergraph-stochastic-block-model-mhsbm