TopoGCL: Topological Graph Contrastive Learning

Yuzhou Chen,Jose Frias,Yulia R. Gel

2024-06-25

Abstract:Graph contrastive learning (GCL) has recently emerged as a new concept which allows for capitalizing on the strengths of graph neural networks (GNNs) to learn rich representations in a wide variety of applications which involve abundant unlabeled information. However, existing GCL approaches largely tend to overlook the important latent information on higher-order graph substructures. We address this limitation by introducing the concepts of topological invariance and extended persistence on graphs to GCL. In particular, we propose a new contrastive mode which targets topological representations of the two augmented views from the same graph, yielded by extracting latent shape properties of the graph at multiple resolutions. Along with the extended topological layer, we introduce a new extended persistence summary, namely, extended persistence landscapes (EPL) and derive its theoretical stability guarantees. Our extensive numerical results on biological, chemical, and social interaction graphs show that the new Topological Graph Contrastive Learning (TopoGCL) model delivers significant performance gains in unsupervised graph classification for 11 out of 12 considered datasets and also exhibits robustness under noisy scenarios.

Machine Learning,Artificial Intelligence

What problem does this paper attempt to address?

### Problems Addressed by the Paper

The paper "TopoGCL: Topological Graph Contrastive Learning" aims to address a key issue in Graph Contrastive Learning (GCL): existing methods often overlook important latent information in higher-order substructures of graphs.

#### Main Contributions:
1. **Introduction of Topological Invariance and Extended Persistence**: By introducing the concepts of topological invariance and Extended Persistence (Persistent Homology, PH), the paper addresses this shortcoming in graph contrastive learning. Specifically, these concepts are used to extract latent shape properties of graphs at different resolutions.

2. **Proposing a New Contrastive Mode**: A new contrastive mode is proposed, which involves contrasting topological representations between two augmented views of the same graph (referred to as topological-topological contrastive learning). This enables the model to capture not only key topological and geometric information but also to enhance the learning of its latent representations.

3. **Extended Persistence Landscape (EPL)**: A new extended persistence summary called the Extended Persistence Landscape is proposed, and its theoretical stability guarantee is demonstrated. It is compared with another extended persistence summary (EPI) and shown to outperform it across different datasets.

#### Experimental Results:
- The effectiveness of the TopoGCL model in unsupervised graph classification tasks was validated on 12 benchmark datasets.
- It outperformed current state-of-the-art baseline models in 11 datasets and showed statistically significant improvements in 8 datasets.
- Demonstrated high robustness in noisy environments.

Through the above methods, the paper effectively enhances the performance of graph contrastive learning and provides a new perspective for graph representation learning.