Graph Contrastive Learning Meets Graph Meta Learning: A Unified Method for Few-shot Node Tasks

Hao Liu,Jiarui Feng,Lecheng Kong,Dacheng Tao,Yixin Chen,Muhan Zhang

2023-09-19

Abstract:Graph Neural Networks (GNNs) have become popular in Graph Representation Learning (GRL). One fundamental application is few-shot node classification. Most existing methods follow the meta learning paradigm, showing the ability of fast generalization to few-shot tasks. However, recent works indicate that graph contrastive learning combined with fine-tuning can significantly outperform meta learning methods. Despite the empirical success, there is limited understanding of the reasons behind it. In our study, we first identify two crucial advantages of contrastive learning compared to meta learning, including (1) the comprehensive utilization of graph nodes and (2) the power of graph augmentations. To integrate the strength of both contrastive learning and meta learning on the few-shot node classification tasks, we introduce a new paradigm: Contrastive Few-Shot Node Classification (COLA). Specifically, COLA employs graph augmentations to identify semantically similar nodes, which enables the construction of meta-tasks without the need for label information. Therefore, COLA can utilize all nodes to construct meta-tasks, further reducing the risk of overfitting. Through extensive experiments, we validate the essentiality of each component in our design and demonstrate that COLA achieves new state-of-the-art on all tasks.

Machine Learning,Artificial Intelligence

What problem does this paper attempt to address?

The paper attempts to address the problem of Few-shot Node Classification (FSNC) in Graph Neural Networks (GNN). Specifically, the paper focuses on how to improve the performance of node classification tasks when there are only a few labeled samples for each category. Most current methods adopt the Meta-Learning paradigm to solve this problem, but recent studies have shown that Graph Contrastive Learning (GCL) combined with fine-tuning can significantly outperform Meta-Learning methods.

The main contributions of the paper are:
1. Analyzing two key advantages of contrastive learning over meta-learning: (1) comprehensive utilization of all nodes in the graph; (2) the power of graph augmentation techniques.
2. Proposing a new framework—Contrastive Few-Shot Node Classification (COLA), which combines the strengths of contrastive learning and meta-learning. This framework identifies semantically similar nodes through graph augmentation techniques, thereby constructing meta-tasks without the need for label information.

Experimental results show that COLA achieves the latest State-of-the-Art (SOTA) performance on multiple benchmark datasets, validating its effectiveness and potential application value.