Probing Negative Sampling Strategies to Learn GraphRepresentations via Unsupervised Contrastive Learning

TL;DR

This paper investigates negative sampling strategies in unsupervised contrastive learning for graph representations, addressing class collision and data imbalance issues, and achieves state-of-the-art results on real-world datasets.

Contribution

It introduces novel negative sampling techniques for node-wise contrastive learning, improving graph representation quality in unsupervised settings.

Findings

Addresses class collision in contrastive learning.

Resolves negative data imbalance issues.

Achieves state-of-the-art performance on three datasets.

Abstract

Graph representation learning has long been an important yet challenging task for various real-world applications. However, their downstream tasks are mainly performed in the settings of supervised or semi-supervised learning. Inspired by recent advances in unsupervised contrastive learning, this paper is thus motivated to investigate how the node-wise contrastive learning could be performed. Particularly, we respectively resolve the class collision issue and the imbalanced negative data distribution issue. Extensive experiments are performed on three real-world datasets and the proposed approach achieves the SOTA model performance.