Phase Transition for Stochastic Block Model with more than $\sqrt{n}$ Communities
A new research paper explores the phase transition for the Stochastic Block Model (SBM) when the number of communities exceeds the square root of the number of nodes. The study provides evidence supporting a new threshold proposed by Chin et al. (2025) for this many-communities regime. The authors prove that low-degree polynomial methods fail below this threshold across all graph densities and demonstrate that community recovery is possible above it, even in moderately sparse regimes, by analyzing specific motif occurrences. AI
IMPACT This research refines understanding of community detection in complex networks, potentially impacting graph-based machine learning algorithms.