Node Embedding via Word Embedding for Network Community Discovery

TL;DR

This paper introduces a novel unsupervised community discovery algorithm using neural node embeddings inspired by word embedding techniques, demonstrating superior performance and theoretical limits on real and simulated graph data.

Contribution

The paper presents a new algorithm that applies word embedding methods to node embeddings for community detection, achieving state-of-the-art results and theoretical optimality.

Findings

Outperforms spectral clustering and belief propagation in community recovery

Achieves information-theoretic limits for community detection under stochastic block models

Demonstrates robustness and accuracy on real-world and simulated datasets

Abstract

Neural node embeddings have recently emerged as a powerful representation for supervised learning tasks involving graph-structured data. We leverage this recent advance to develop a novel algorithm for unsupervised community discovery in graphs. Through extensive experimental studies on simulated and real-world data, we demonstrate that the proposed approach consistently improves over the current state-of-the-art. Specifically, our approach empirically attains the information-theoretic limits for community recovery under the benchmark Stochastic Block Models for graph generation and exhibits better stability and accuracy over both Spectral Clustering and Acyclic Belief Propagation in the community recovery limits.