Inference via Message Passing on Partially Labeled Stochastic Block Models
This work addresses the fundamental limitations of local algorithms for community recovery in networks, with incremental improvements in error bounds for partially observed data.
The paper tackles the community detection problem in partially-labeled stochastic block models by developing a linearized message-passing algorithm, achieving an exponential mis-classification rate of exp(-(SNR-1)/2) when SNR>1, which improves upon the previous rate of (SNR-1)^{-1}.
We study the community detection and recovery problem in partially-labeled stochastic block models (SBM). We develop a fast linearized message-passing algorithm to reconstruct labels for SBM (with $n$ nodes, $k$ blocks, $p,q$ intra and inter block connectivity) when $δ$ proportion of node labels are revealed. The signal-to-noise ratio ${\sf SNR}(n,k,p,q,δ)$ is shown to characterize the fundamental limitations of inference via local algorithms. On the one hand, when ${\sf SNR}>1$, the linearized message-passing algorithm provides the statistical inference guarantee with mis-classification rate at most $\exp(-({\sf SNR}-1)/2)$, thus interpolating smoothly between strong and weak consistency. This exponential dependence improves upon the known error rate $({\sf SNR}-1)^{-1}$ in the literature on weak recovery. On the other hand, when ${\sf SNR}<1$ (for $k=2$) and ${\sf SNR}<1/4$ (for general growing $k$), we prove that local algorithms suffer an error rate at least $\frac{1}{2} - \sqrt{δ\cdot {\sf SNR}}$, which is only slightly better than random guess for small $δ$.