Subgraph Augmentation with Application to Graph Mining

Abstract

Graph classification, which aims to identify the category labels of graphs, plays a significant role in drug classification, toxicity detection, protein analysis etc. However, the limitation of the general scale of benchmark datasets makes easily causes graph classification models to fall into overfitting and undergeneralization. In this chapter, the M-Evolve framework is introduced for graph classification (Zhou et al., Data augmentation for graph classification. In: Proceedings of the 29th ACM International Conference on Information & Knowledge Management, pp. 2341–2344, 2020; Zhou et al., M-evolve: structural-mapping-based data augmentation for graph classification. In: IEEE Transactions on Network Science and Engineering, pp. 1–1, 2020), a novel technique for expanding graph structured data spaces and optimizing graph classifiers. Typical graph tasks such as node classification and link prediction are unified to generate graph classification patterns, demonstrating some applications to multiple tasks in graph mining. One of the main contributions of this chapter is to apply the technique of subgraph augmentation for various tasks. The M-Evolve is general and flexible, which can be easily combined with existing graph classification models. Extensive experiments are conducted on real datasets to illustrate the effectiveness of our framework.