Graph Contrastive Learning with Reinforced Augmentation

Ziyang Liu, Chaokun Wang

Keywords: Graph contrastive learning, graph neural network, graph classification, reinforcement learning

Abstract: Graph contrastive learning (GCL), designing contrastive objectives to learn embeddings from augmented graphs, has become a prevailing method for learning embeddings from graphs in an unsupervised manner. As an important procedure in GCL, graph data augmentation (GDA) directly affects the model performance on the downstream task. Currently, there are three types of GDA strategies: trial-and-error, precomputed method, and adversarial method. However, these strategies ignore the connection between the two consecutive augmentation results because GDA is regarded as an independent process. In this paper, we regard the GDA in GCL as a Markov decision process. Based on this point, we propose a reinforced method, i.e., the fourth type of GDA strategy, using a novel Graph Advantage Actor-Critic (GA2C) model for GCL. On 23 graph datasets, the experimental results verify that GA2C outperforms the SOTA GCL models on a series of downstream tasks such as graph classification, node classification, and link prediction.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Unsupervised and Self-supervised learning

TL;DR: In this paper, we design a novel GA2C model making the augmented views evolves well to energize graph contrastive learning and outperforms the SOTA methods..