Research map / Graph learning / GNN architectures
Towards Label Position Bias in Graph Neural Networks
Haoyu Han, Xiaorui Liu, Feng Shi, MohamadAli Torkamani, Charu C. Aggarwal, Jiliang Tang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use LPSL-learned graph structure with APPNP | APPNP · Cora, 5 labels/class | accuracy (%): 75.86 ± 2.34 -> 77.24 ± 2.18 | Improved |
| Use LPSL-learned graph structure with APPNP | APPNP · CiteSeer, 20 labels/class | accuracy (%): 70.85 ± 1.45 -> 71.25 ± 1.14 | Improved |
| Use LPSL-learned graph structure with APPNP | APPNP · PubMed, 60% labels/class | accuracy (%): 78.93 ± 2.11 -> 87.96 ± 0.57 | Hurt |
| Use LPSL-learned graph structure with GCN | GCN · Amazon Computers, 20 labels/class | accuracy (%): 79.17 ± 1.92 -> 80.05 ± 2.92 | Improved |
| Use LPSL-learned graph structure with GCN | GCN · ogbn-arxiv, 54% labels | accuracy (%): 71.91 ± 0.15 -> 72.04 ± 0.12 | Improved |
| Use LPSL-learned graph structure | GCN and APPNP · Cora and CiteSeer | LPS-group bias (WDP): Cora GCN 0.0494 -> 0.0361; CiteSeer APPNP 0.0344 -> 0.0275 | Mixed |
| Use LPSL-learned graph structure | GCN and APPNP · Cora and CiteSeer | degree bias (WDP): Cora APPNP 0.0408 -> 0.0349; CiteSeer APPNP 0.0499 -> 0.0316 | Improved |
| Use LPSL-learned graph structure | GCN and APPNP · Cora and CiteSeer | shortest-path-distance bias (WDP): Cora GCN 0.0237 -> 0.0150; CiteSeer GCN 0.0296 -> 0.0246 | Mixed |
| Vary smoothing parameter λ within the selected ranges (ablation) | LPSL-APPNP and LPSL-GCN · Cora and CiteSeer, 20 labels/class | accuracy | No clear effect |
| Vary uniform proximity constraint c within the selected ranges (ablation) | LPSL-APPNP and LPSL-GCN · Cora and CiteSeer, 20 labels/class | accuracy | No clear effect |
Closest papers by meaning
- Combining Label Propagation and Simple Models out-performs Graph Neural Networks (ICLR 2021 · similarity 0.89)
- Leveraging Label Non-Uniformity for Node Classification in Graph Neural Networks (ICML 2023 · similarity 0.89)
- NoisyGL: A Comprehensive Benchmark for Graph Neural Networks under Label Noise (NeurIPS 2024 · similarity 0.88)
- On Which Nodes Does GCN Fail? Enhancing GCN From the Node Perspective (ICML 2024 · similarity 0.87)
- Label-free Node Classification on Graphs with Large Language Models (LLMs) (ICLR 2024 · similarity 0.87)
- NLGT: Neighborhood-based and Label-enhanced Graph Transformer Framework for Node Classification (AAAI 2025 · similarity 0.87)
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.
