GraphMixup: Improving Class-Imbalanced Node Classification by Reinforcement Mixup and Self-Supervised Context Prediction

Abstract

Data imbalance, i.e., some classes may have much fewer samples than others, is a serious problem that can lead to unfavorable node classification. However, most existing GNNs are based on the assumption that node samples for different classes are balanced. In this case, directly training a GNN classifier with raw data would under-represent samples from those minority classes and result in sub-optimal performance. This paper proposes GraphMixup, a novel mixup-based framework for improving class-imbalanced node classification on graphs. However, directly performing mixup in the input space or embedding space may produce out-of-domain samples due to the extreme sparsity of minority classes; hence we construct semantic relation spaces that allow Feature Mixup to be performed at the semantic level. Moreover, we apply two context-based self-supervised techniques to capture both local and global information in the graph structure and specifically propose Edge Mixup to handle graph data. Finally, we develop a Reinforcement Mixup mechanism to adaptively determine how many samples are to be generated by mixup for those minority classes. Extensive experiments on three real-world datasets have shown that GraphMixup yields truly encouraging results for the task of class-imbalanced node classification. Codes are available at: https://github.com/LirongWu/GraphMixup .