We establish the theoretical limits of clustering under the Popularity-Adjusted Block Model and show that even without edge-density signals, cluster recovery is possible by leveraging differences in intra- and inter-cluster popularity parameters

Neurons can cluster sensory dynamics into coherent sets by projecting inputs onto a singular vector and rectifying the result

For Spectral Clustering, Uniform Sampling on the edges works.

We develop a new approach to explainable clustering as defined by Moshkovitz et al. (2020) based on spectral graph theory