AI Paper Atlas

Research map / Graph learning / GNN architectures

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

Mingxuan Ju, Tong Zhao, Wenhao Yu, Neil Shah, Yanfang Ye

NeurIPS 2023 · Graph learning › GNN architectures · approach: GNN generalization theory · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add GraphPatcher test-time patching to GCNGCN · Cora (low-degree nodes)accuracy: 73.27 -> 78.08Improved
add GraphPatcher test-time patching to GCNGCN · Cora (overall)accuracy: 81.22 -> 84.17Improved
add GraphPatcher to GCNGCN · Cora (high-degree nodes)accuracy: 86.83 -> 88.02Improved
add TuneUp to GCNGCN · Amazon-Photo (high-degree nodes)accuracy: 84.07 -> 81.57Hurt
add ColbBrew to GCNGCN · Wiki.CS (high-degree nodes)accuracy: 83.40 -> 71.98Hurt
add Tail-GNN to GCNGCN · Cora (high-degree nodes)accuracy: 86.83 -> 84.13Hurt
add EERM or GTrans (OOD frameworks) to GCNGCN · degree-bias benchmarksaccuracyNo clear effect
apply GraphPatcher to GraphSAGEGraphSAGE · Cora (low-degree nodes)accuracy: 70.57 -> 72.33Improved
apply GraphPatcher to GATGAT · Cora (low-degree nodes)accuracy: 73.27 -> 76.88Improved
apply GraphPatcher to self-supervised DGIDGI · Pubmed (low-degree nodes)accuracy: 75.63 -> 78.04Improved
apply GraphPatcher to GRANDGRAND · Pubmed (low-degree nodes)accuracy: 80.48 -> 84.68Improved
increase number of test-time patching nodes (ablation)GraphPatcher · Cora, Citeseer, Pubmedoverall accuracyImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy