English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Recent studies have shown that graph neural networks (GNNs) exhibit strongbiases towards the node degree: they usually perform satisfactorily onhigh-degree nodes with rich neighbor information but struggle with low-degreenodes. Existing works tackle this problem by deriving either designated GNNarchitectures or training strategies specifically for low-degree nodes. Thougheffective, these approaches unintentionally create an artificialout-of-distribution scenario, where models mainly or even only observelow-degree nodes during the training, leading to a downgraded performance forhigh-degree nodes that GNNs originally perform well at. In light of this, wepropose a test-time augmentation framework, namely GraphPatcher, to enhancetest-time generalization of any GNNs on low-degree nodes. Specifically,GraphPatcher iteratively generates virtual nodes to patch artificially createdlow-degree nodes via corruptions, aiming at progressively reconstructing targetGNN's predictions over a sequence of increasingly corrupted nodes. Through thisscheme, GraphPatcher not only learns how to enhance low-degree nodes (when theneighborhoods are heavily corrupted) but also preserves the original superiorperformance of GNNs on high-degree nodes (when lightly corrupted).Additionally, GraphPatcher is model-agnostic and can also mitigate the degreebias for either self-supervised or supervised GNNs. Comprehensive experimentsare conducted over seven benchmark datasets and GraphPatcher consistentlyenhances common GNNs' overall performance by up to 3.6% and low-degreeperformance by up to 6.5%, significantly outperforming state-of-the-artbaselines. The source code is publicly available athttps://github.com/jumxglhf/GraphPatcher.

GraphPatcher: Mitigating Degree Bias for Graph Neural Networks via Test-time Augmentation