Dual-level Mixup for Graph Few-shot Learning with Fewer Tasks

TL;DR

This paper introduces SMILE, a dual-level mixup approach for graph few-shot learning that improves generalization with fewer tasks by enriching data and leveraging node degree information.

Contribution

SMILE is a novel method combining within-task and across-task mixup strategies, effectively reducing the need for numerous meta-training tasks in graph few-shot learning.

Findings

SMILE outperforms existing models on multiple datasets.

The dual-level mixup enhances data diversity and model robustness.

Leveraging node degrees improves node representation quality.

Abstract

Graph neural networks have been demonstrated as a powerful paradigm for effectively learning graph-structured data on the web and mining content from it.Current leading graph models require a large number of labeled samples for training, which unavoidably leads to overfitting in few-shot scenarios. Recent research has sought to alleviate this issue by simultaneously leveraging graph learning and meta-learning paradigms. However, these graph meta-learning models assume the availability of numerous meta-training tasks to learn transferable meta-knowledge. Such assumption may not be feasible in the real world due to the difficulty of constructing tasks and the substantial costs involved. Therefore, we propose a SiMple yet effectIve approach for graph few-shot Learning with fEwer tasks, named SMILE. We introduce a dual-level mixup strategy, encompassing both within-task and across-task mixup, to simultaneously enrich the available nodes and tasks in meta-learning. Moreover, we explicitly leverage the prior information provided by the node degrees in the graph to encode expressive node representations. Theoretically, we demonstrate that SMILE can enhance the model generalization ability. Empirically, SMILE consistently outperforms other competitive models by a large margin across all evaluated datasets with in-domain and cross-domain settings. Our anonymous code can be found here.