[Paper Review] Optimal and exact recovery on general non-uniform Hypergraph Stochastic Block Model

Ioana Dumitriu, Hai‐Xiao Wang|arXiv (Cornell University)|Apr 25, 2023

Complex Network Analysis Techniques4 citations

TL;DR

This paper establishes the first sharp threshold for exact recovery in the non-uniform hypergraph stochastic block model (HSBM) with multiple communities (K ≥ 2), proving that exact recovery is possible when information from all hyperedge types is aggregated, even when individual layers are insufficient. The authors propose two efficient algorithms leveraging spectral initialization and local correction, with theoretical guarantees based on concentration and regularization of the non-uniform hypergraph adjacency matrix.

ABSTRACT

Consider the community detection problem in random hypergraphs under the non-uniform hypergraph stochastic block model (HSBM), where each hyperedge appears independently with some given probability depending only on the labels of its vertices. We establish, for the first time in the literature, a sharp threshold for exact recovery under this non-uniform case, subject to minor constraints; in particular, we consider the model with multiple communities. One crucial point here is that by aggregating information from all the uniform layers, we may obtain exact recovery even in cases when this may appear impossible if each layer were considered alone. Besides that, we prove a wide-ranging, information-theoretic lower bound on the number of misclassified vertices \emph{for any algorithm}, depending on a \emph{generalized Chernoff-Hellinger} divergence involving model parameters. We provide two efficient algorithms which successfully achieve exact recovery when above the threshold, and attain the lowest possible mismatch ratio when the exact recovery is impossible, proved to be optimal. The theoretical analysis of our algorithms relies on the concentration and regularization of the adjacency matrix for non-uniform random hypergraphs, which could be of independent interest. We also address some open problems regarding parameter knowledge and estimation.

Motivation & Objective

• To establish a sharp threshold for exact recovery in the non-uniform hypergraph stochastic block model (HSBM) with multiple communities (K ≥ 2).
• To demonstrate that exact recovery is possible by aggregating information across all hyperedge types, even when individual layers are uninformative.
• To develop two efficient algorithms that achieve exact recovery above the derived threshold, with and without prior knowledge of edge probabilities.
• To provide theoretical analysis based on concentration and regularization of the non-uniform hypergraph adjacency matrix, which may be of independent interest.
• To address open problems regarding estimation of the number of communities and parameter knowledge in the HSBM framework.

Proposed method

• Propose a non-uniform HSBM where each hyperedge appears independently with a probability depending only on the community labels of its vertices.
• Use spectral initialization in Stage I to achieve weak consistency, leveraging the structure of the hypergraph adjacency matrix.
• Apply a two-stage algorithm: first achieve weak consistency via spectral methods, then refine via local correction to achieve exact recovery.
• Introduce hypergraph splitting to decompose the non-uniform model into uniform layers, enabling analysis of individual contributions.
• Establish concentration bounds for the adjacency matrix using a generalized Kahn-Szemerédi approach, with separate treatment of light and heavy couples in the variance decomposition.
• Use Chernoff-type bounds and logarithmic binomial approximations to control tail probabilities in the recovery process.

Experimental results

Research questions

• RQ1 What is the sharp threshold for exact recovery in the non-uniform hypergraph stochastic block model with K ≥ 2 communities?
• RQ2 Can exact recovery be achieved in the non-uniform HSBM even when individual hyperedge types are insufficient for recovery?
• RQ3 What are efficient algorithms that achieve exact recovery above the threshold, both with and without prior knowledge of edge probabilities?
• RQ4 How can the concentration and regularization of the non-uniform hypergraph adjacency matrix be rigorously analyzed to support exact recovery guarantees?
• RQ5 What are the implications for estimating the number of communities and unknown model parameters in the non-uniform HSBM?

Key findings

• The paper establishes a sharp threshold for exact recovery in the non-uniform HSBM, proving that exact recovery is possible if and only if the model parameters satisfy a specific information-theoretic condition.
• Aggregating information from all hyperedge types enables exact recovery even when no single layer allows it, demonstrating a synergistic effect across layers.
• Two efficient algorithms are proposed: one requiring prior knowledge of edge probabilities and another achieving strong consistency without such knowledge, both achieving exact recovery above the threshold.
• Theoretical analysis relies on concentration inequalities and a refined variance decomposition (light vs. heavy couples) in the adjacency matrix, extending the Kahn-Szemerédi method to hypergraphs.
• The paper provides a non-asymptotic analysis of the hypergraph adjacency matrix, showing that its spectral properties concentrate around the mean under mild conditions.
• The results resolve open problems regarding parameter estimation and community count recovery in the non-uniform HSBM, particularly in the regime where edge probabilities are unknown.