Zero-Shot Generalization of GNNs over Distinct Attribute Domains

Yangyi Shen, Beatrice Bevilacqua, Joshua Robinson, Charilaos Kanatsoulis, Jure Leskovec, Bruno Ribeiro

Keywords: Graph neural networks, zero-shot, attributed graphs, graph foundation models

TL;DR: We propose the first pure graph-based method that can zero-shot generalize to test attributed graphs with different features than those seen in train.

Abstract: There are no known graph machine learning methods that can zero-shot generalize across attributed graphs with very different node attribute domains and consistently outperform methods that ignore node attributes. For instance, no method can significantly outperform structure-only predictions in zero-shot link prediction by pretraining on online appliance store datasets (with node attributes such as brand, model, capacity, dimension, has ice maker, energy rating for refrigerators) and zero-shot at test on an electronics store dataset for smartphones (with attributes such as processor type, display type, storage, and battery capacity). In this work, we leverage concepts in statistical theory to design STAGE, a universally applicable approach for encoding node attributes in _any GNN_ that facilitates such generalization. Empirically, we show that STAGE outperforms its natural baselines and can accurately make predictions when presented with completely new feature domains.

Submission Number: 41