ARIEL: Adversarial Graph Contrastive Learning

TL;DR

ARIEL addresses robustness and sample quality in graph contrastive learning by introducing an adversarial graph view as a new data augmentation, paired with information regularization to stabilize training. It scales via subgraph sampling and unifies node- and graph-level learning by treating graphs as super-nodes in a universal framework. Empirical results show ARIEL consistently improves over state-of-the-art graph CL methods on node and graph classification and exhibits stronger resistance to adversarial attacks. The work highlights a practical, scalable path toward more expressive and robust unsupervised graph representations with potential for broader adoption.

Abstract

Contrastive learning is an effective unsupervised method in graph representation learning, and the key component of contrastive learning lies in the construction of positive and negative samples. Previous methods usually utilize the proximity of nodes in the graph as the principle. Recently, the data-augmentation-based contrastive learning method has advanced to show great power in the visual domain, and some works extended this method from images to graphs. However, unlike the data augmentation on images, the data augmentation on graphs is far less intuitive and much harder to provide high-quality contrastive samples, which leaves much space for improvement. In this work, by introducing an adversarial graph view for data augmentation, we propose a simple but effective method, Adversarial Graph Contrastive Learning (ARIEL), to extract informative contrastive samples within reasonable constraints. We develop a new technique called information regularization for stable training and use subgraph sampling for scalability. We generalize our method from node-level contrastive learning to the graph level by treating each graph instance as a super-node. ARIEL consistently outperforms the current graph contrastive learning methods for both node-level and graph-level classification tasks on real-world datasets. We further demonstrate that ARIEL is more robust in the face of adversarial attacks.

Figures (4)

• Figure 1: Average cosine similarity between the node embeddings of the original graph and the perturbed graph, results are on datasets Amazon-Computers and Amazon-Photo. The shaded area represents the standard deviation.
• Figure 2: The overview of the proposed ArieL framework. For each iteration, two augmented views are generated from the original graph by data augmentation (purple arrows), and then an adversarial view is generated (red arrow) from the original graph by maximizing the contrastive loss against one of the augmented views. Besides, the similarities of the corresponding nodes (dashed lines) will get penalized by the information regularization if they exceed the estimated upper bound. The objective of ArieL is to minimize the contrastive loss (grey arrows) between the augmented views, the adversarial view, and the corresponding augmented view, and the information regularization. Best viewed in color.
• Figure 3: Effect of adversarial contrastive loss coefficient $\epsilon_1$ on Cora and CiteSeer. The dashed line represents the performance of GRACE with subgraph sampling.
• Figure 4: Effect of information regularization on Amazon-Photo. The left figure shows the model performance under different $\epsilon_2$ and the right figure plots the training curve of ArieL under $\epsilon_2=0$ and $\epsilon_2=1.0$.

Theorems & Definitions (1)

• Theorem 1