Cross-Domain Few-Shot Graph Classification

TL;DR

This paper introduces new benchmarks and an attention-based graph encoder for cross-domain few-shot graph classification, demonstrating improved performance with metric-based meta-learning strategies.

Contribution

It presents the first benchmarks for cross-domain few-shot graph classification and proposes a novel attention-based encoder for effective knowledge transfer.

Findings

Proposed encoder improves classification accuracy across benchmarks.

Coupling with metric-based meta-learning yields best results.

Benchmarks facilitate future research in cross-domain graph tasks.

Abstract

We study the problem of few-shot graph classification across domains with nonequivalent feature spaces by introducing three new cross-domain benchmarks constructed from publicly available datasets. We also propose an attention-based graph encoder that uses three congruent views of graphs, one contextual and two topological views, to learn representations of task-specific information for fast adaptation, and task-agnostic information for knowledge transfer. We run exhaustive experiments to evaluate the performance of contrastive and meta-learning strategies. We show that when coupled with metric-based meta-learning frameworks, the proposed encoder achieves the best average meta-test classification accuracy across all benchmarks. The source code and data will be released here: https://github.com/kavehhassani/metagrl