Research map / Graph learning / Graph self-supervised learning
Why Propagate Alone? Parallel Use of Labels and Features on Graphs
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Add trainable label-mapping weights to label propagation | Label Propagation · Cora-full, Pubmed, Arxiv, Products | accuracy: 66.44 ± 0.93 -> 67.23 ± 0.66; 83.45 ± 0.63 -> 83.52 ± 0.59; 67.11 ± 0.00 -> 68.42 ± 0.01; 74.24 ± 0.00 -> 75.61 ± 0.21 | Helped |
| Apply deterministic self-excluded label trick (ablation) | SGC · Pubmed, Arxiv | accuracy: 85.02 ± 0.43 -> 85.23 ± 0.57; 69.07 ± 0.01 -> 70.22 ± 0.03 | Helped |
| Apply deterministic self-excluded label trick (ablation) | SIGN · Pubmed, Arxiv | accuracy: 87.94 ± 0.52 -> 88.09 ± 0.59; 69.97 ± 0.16 -> 70.98 ± 0.21 | Helped |
| Apply deterministic self-excluded label trick (ablation) | TWIRLS · Pubmed, Arxiv | accuracy: 89.81 ± 0.56 -> 90.08 ± 0.52; 72.93 ± 0.19 -> 73.22 ± 0.10 | Helped |
| Apply deterministic self-excluded label trick (ablation) | SIGN and TWIRLS · Cora-full | accuracy: 68.54 ± 0.76 -> 68.44 ± 0.88; 70.36 ± 0.51 -> 70.40 ± 0.71 | No clear effect |
| Train C&S post-processing weights | MLP+C&S · Cora-full, Pubmed | accuracy: 66.95 ± 1.46 -> 67.89 ± 1.37; 89.12 ± 0.27 -> 89.76 ± 0.17 | Helped |
| Train C&S post-processing weights | MLP+C&S · Arxiv, Products | accuracy: 73.05 ± 0.35 -> 73.03 ± 0.18; 83.16 ± 0.13 -> 83.10 ± 0.15 | Mixed |
| Train C&S weights with label trick | GBDT+C&S · House, County, VK | MSE: 0.51 ± 0.01 -> 0.45 ± 0.01; 1.42 ± 0.14 -> 1.13 ± 0.09; 7.02 ± 0.20 -> 6.95 ± 0.22 | Helped |
| Train C&S weights with label trick | GBDT+C&S · Avazu | MSE: 0.106 ± 0.014 -> 0.106 ± 0.014 | No clear effect |
| Increase the label-splitting probability α (ablation) | Linear feature-and-label propagation on Arxiv · Arxiv validation | accuracy | Helped |
| Increase the label-splitting probability α (ablation) | Three-layer GNN · Arxiv validation | accuracy | Hurt |
Closest papers by meaning
- Combining Label Propagation and Simple Models out-performs Graph Neural Networks (ICLR 2021 · similarity 0.93)
- Meta Propagation Networks for Graph Few-shot Semi-supervised Learning (AAAI 2022 · similarity 0.91)
- NLGT: Neighborhood-based and Label-enhanced Graph Transformer Framework for Node Classification (AAAI 2025 · similarity 0.89)
- Multi-Label Node Classification with Label Influence Propagation (ICLR 2025 · similarity 0.89)
- Label Attentive Distillation for GNN-Based Graph Classification (AAAI 2024 · similarity 0.88)
- Label-free Node Classification on Graphs with Large Language Models (LLMs) (ICLR 2024 · similarity 0.88)
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.