COSTA: Covariance-Preserving Feature Augmentation for Graph Contrastive Learning

TL;DR

COSTA introduces a novel feature augmentation method for graph contrastive learning that preserves covariance, leading to improved or comparable performance with reduced memory and computation compared to traditional graph augmentation.

Contribution

The paper proposes COSTA, a covariance-preserving feature augmentation framework for GCL, shifting focus from input graph augmentation to hidden feature augmentation.

Findings

COSTA achieves comparable or better results than graph augmentation methods.

Feature augmentation with COSTA conserves memory and computational resources.

COSTA enhances the discriminative power of learned graph representations.

Abstract

Graph contrastive learning (GCL) improves graph representation learning, leading to SOTA on various downstream tasks. The graph augmentation step is a vital but scarcely studied step of GCL. In this paper, we show that the node embedding obtained via the graph augmentations is highly biased, somewhat limiting contrastive models from learning discriminative features for downstream tasks. Thus, instead of investigating graph augmentation in the input space, we alternatively propose to perform augmentations on the hidden features (feature augmentation). Inspired by so-called matrix sketching, we propose COSTA, a novel COvariance-preServing feaTure space Augmentation framework for GCL, which generates augmented features by maintaining a "good sketch" of original features. To highlight the superiority of feature augmentation with COSTA, we investigate a single-view setting (in addition to multi-view one) which conserves memory and computations. We show that the feature augmentation with COSTA achieves comparable/better results than graph augmentation based models.