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

TL;DR

This work tackles graph few-shot learning under the constraint of few meta-training tasks. It introduces SMILE, which combines degree-aware node representations with a dual-level mixup strategy (within-task and across-task) to enrich data and task distributions, backed by theoretical bounds that indicate reduced generalization error. Empirically, SMILE consistently outperforms baselines on in-domain and cross-domain tasks across multiple datasets, with ablations highlighting the importance of degree priors and task/sample augmentation. The approach is simple yet effective, offering practical improvements for graph-based learning when labeled data and task diversity are limited, and it provides a foundation for robust meta-learning in scarce-resource graph settings.

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.

Figures (3)

• Figure 1: Model performance varies with the number of meta-training tasks across different datasets.
• Figure 2: The overall architecture of SMILE.
• Figure 3: Results vary with hyperparameters.

Theorems & Definitions (3)

• Theorem 4.1
• Corollary 4.2
• Theorem 4.3