CSGCL: Community-Strength-Enhanced Graph Contrastive Learning

TL;DR

This paper introduces CSGCL, a novel graph contrastive learning framework that incorporates community strength to improve graph representations, achieving state-of-the-art results in node classification, clustering, and link prediction.

Contribution

The paper proposes a new framework that models community influence differences and integrates community strength into graph contrastive learning, with novel augmentation methods and a dynamic contrastive scheme.

Findings

CSGCL outperforms existing GCL methods on multiple tasks.

Community strength enhances the quality of learned graph representations.

The proposed methods effectively leverage community information for better learning.

Abstract

Graph Contrastive Learning (GCL) is an effective way to learn generalized graph representations in a self-supervised manner, and has grown rapidly in recent years. However, the underlying community semantics has not been well explored by most previous GCL methods. Research that attempts to leverage communities in GCL regards them as having the same influence on the graph, leading to extra representation errors. To tackle this issue, we define ''community strength'' to measure the difference of influence among communities. Under this premise, we propose a Community-Strength-enhanced Graph Contrastive Learning (CSGCL) framework to preserve community strength throughout the learning process. Firstly, we present two novel graph augmentation methods, Communal Attribute Voting (CAV) and Communal Edge Dropping (CED), where the perturbations of node attributes and edges are guided by community strength. Secondly, we propose a dynamic ''Team-up'' contrastive learning scheme, where community strength is used to progressively fine-tune the contrastive objective. We report extensive experiment results on three downstream tasks: node classification, node clustering, and link prediction. CSGCL achieves state-of-the-art performance compared with other GCL methods, validating that community strength brings effectiveness and generality to graph representations. Our code is available at https://github.com/HanChen-HUST/CSGCL.