An Approach for Addressing Internally-Disconnected Communities in Louvain Algorithm

TL;DR

This paper introduces GSP-Louvain, a parallel community detection algorithm based on Louvain, which effectively addresses internally-disconnected communities and significantly outperforms existing Leiden-based algorithms in speed.

Contribution

GSP-Louvain is a novel parallel algorithm that mitigates the Louvain algorithm's issue of internally-disconnected communities and achieves substantial performance improvements.

Findings

GSP-Louvain outperforms Leiden, NetworKit Leiden, and cuGraph Leiden by up to 391x, 6.9x, and 2.6x respectively.

It processes 410 million edges per second on a 3.8 billion edge graph.

Performance improves by 1.5x with each doubling of threads.

Abstract

Community detection is the problem of identifying densely connected clusters within a network. While the Louvain algorithm is commonly used for this task, it can produce internally-disconnected communities. To address this, the Leiden algorithm was introduced. This technical report introduces GSP-Louvain, a parallel algorithm based on Louvain, which mitigates this issue. Running on a system with two 16-core Intel Xeon Gold 6226R processors, GSP-Louvain outperforms Leiden, NetworKit Leiden, and cuGraph Leiden by 391x, 6.9x, and 2.6x respectively, processing 410M edges per second on a 3.8B edge graph. Furthermore, GSP-Louvain improves performance at a rate of 1.5x for every doubling of threads.