Evaluate minimum-degree safeguards for thresholded priors
Determine how requiring every node to retain at least one incident edge when the diffusion-derived prior graph is thresholded affects overlapping community-detection performance.
References
One possible safeguard would be to require that every node retain at least one incident edge when the prior graph is thresholded. Such a constraint could reduce the risk of isolating peripheral nodes, but its effect on community-detection performance is included in our future work.
— Diffusion-Induced Spatial Attention Overlapping Community Detection
(2609.26737 - Koistinen et al., 22 Sep 2026) in Appendix A, “Influence-Spreading Model Weights”