From Canonical Correlation Analysis to Self-supervised Graph Neural Networks

TL;DR

This paper presents a simple self-supervised graph representation learning method inspired by Canonical Correlation Analysis, avoiding complex components like negative samples, and achieving competitive results across multiple datasets.

Contribution

It introduces a novel feature-level objective for self-supervised graph learning that simplifies existing methods and provides theoretical insights based on the Information Bottleneck Principle.

Findings

Performs competitively on seven public graph datasets.

Does not require negative samples or additional projectors.

Provides theoretical understanding via the Information Bottleneck Principle.

Abstract

We introduce a conceptually simple yet effective model for self-supervised representation learning with graph data. It follows the previous methods that generate two views of an input graph through data augmentation. However, unlike contrastive methods that focus on instance-level discrimination, we optimize an innovative feature-level objective inspired by classical Canonical Correlation Analysis. Compared with other works, our approach requires none of the parameterized mutual information estimator, additional projector, asymmetric structures, and most importantly, negative samples which can be costly. We show that the new objective essentially 1) aims at discarding augmentation-variant information by learning invariant representations, and 2) can prevent degenerated solutions by decorrelating features in different dimensions. Our theoretical analysis further provides an understanding for the new objective which can be equivalently seen as an instantiation of the Information Bottleneck Principle under the self-supervised setting. Despite its simplicity, our method performs competitively on seven public graph datasets. The code is available at: https://github.com/hengruizhang98/CCA-SSG.