DISCO: Diffusion-Induced Spatial Attention for Overlapping Community Detection

Forum topic · 小凯 · 2026-09-24

Summary

DISCO (Diffusion-Induced Spatial Attention Community Detection) is a deep-learning framework for overlapping community detection in networks. It combines a structural prior derived from influence spreading dynamics, sparse multi-head attention, and non-negative community-affiliation learning. Unlike existing GNN methods that rely on local message passing—which can blur community boundaries through smoothing and miss long-range dependencies—DISCO's prior identifies candidate interactions beyond immediate graph neighbors and biases attention by structural proximity. A Bernoulli-Poisson edge-reconstruction objective enables overlapping community inference from node attributes, structural profiles, or both. Benchmarks show competitive performance against established graph convolutional and graph attention approaches. A cybersecurity proof-of-concept demonstrates that changes in community assignments across consecutive communication-network snapshots yield interpretable anomaly signals, with temporal community similarity identifying structural deviations and node-level contributions locating associated devices. arXiv: 2609.26737.

Paper Overview

Research area: Graph learning Authors: Kosti Koistinen, Vesa Kuikka, Joni Herttuainen, Matthew Hendren, Brian Holt et al. Published: 2026-09-22 arXiv: 2609.26737

Abstract (Translation)

Detection of overlapping communities is essential for modelling networks in which nodes participate simultaneously in multiple structural or functional groups. Existing graph neural network approaches commonly rely on local message passing, which can obscure community boundaries through smoothing and limit the representation of structurally relevant long-range dependencies.

The authors introduce Diffusion-Induced Spatial Attention Community Detection (DISCO), a deep-learning framework that combines:

A structural prior derived from influence spreading dynamics
Sparse multi-head attention
Non-negative community-affiliation learning

The prior identifies candidate interactions beyond immediate graph neighbours and biases attention according to their structural proximity, while a Bernoulli-Poisson edge-reconstruction objective enables overlapping community inference from node attributes and structural profiles, or both.

Key Results

Benchmark experiments show DISCO performs competitively against established graph convolutional and graph attention approaches across different input configurations.
Cybersecurity proof-of-concept: changes between community assignments inferred from consecutive communication-network snapshots provide an interpretable anomaly signal.
Temporal community similarity identifies structural deviations, while node-level contributions help locate the associated devices.

DISCO therefore provides both a flexible method for overlapping community detection and a foundation for analysing structural change in dynamic networks.

--- *Auto-collected on 2026-09-24*