Spectral redemption: clustering sparse networks
This addresses the challenge of community detection in sparse networks for researchers and practitioners, representing a novel method rather than an incremental improvement.
The paper tackled the problem of suboptimal spectral clustering in sparse networks by introducing a non-backtracking walk-based algorithm, which achieves optimal community detection down to the theoretical limit in stochastic block models.
Spectral algorithms are classic approaches to clustering and community detection in networks. However, for sparse networks the standard versions of these algorithms are suboptimal, in some cases completely failing to detect communities even when other algorithms such as belief propagation can do so. Here we introduce a new class of spectral algorithms based on a non-backtracking walk on the directed edges of the graph. The spectrum of this operator is much better-behaved than that of the adjacency matrix or other commonly used matrices, maintaining a strong separation between the bulk eigenvalues and the eigenvalues relevant to community structure even in the sparse case. We show that our algorithm is optimal for graphs generated by the stochastic block model, detecting communities all the way down to the theoretical limit. We also show the spectrum of the non-backtracking operator for some real-world networks, illustrating its advantages over traditional spectral clustering.