Research map / Graph learning / GNN architectures
Self-supervised Graph Neural Networks via Low-Rank Decomposition
Liang Yang, Runjie Shi, Qiuliang Zhang, Bingxin Niu, Zhen Wang, Xiaochun Cao, Chuan Wang
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 tensor decomposition with similar ego-networks instead of matrix decomposition (ablation) | LRD-GNN · Wiki-CS | accuracy (%): 80.19±0.18 -> 81.43±0.13 | Improved |
| use tensor decomposition instead of matrix decomposition (ablation) | LRD-GNN · Squirrel | accuracy (%): 47.64±1.21 -> 55.91±0.91 | Improved |
| use tensor decomposition instead of matrix decomposition (ablation) | LRD-GNN · Chameleon | accuracy (%): 60.71±2.21 -> 66.27±1.27 | Improved |
| use tensor decomposition instead of matrix decomposition (ablation) | LRD-GNN · Actor | accuracy (%): 37.22±0.83 -> 37.07±0.83 | No clear effect |
| use tensor decomposition instead of matrix decomposition (ablation) | LRD-GNN · Cornell | accuracy (%): 78.48±0.36 -> 84.73±2.83 | Improved |
| LRD-GNN-Tensor vs best SSL baseline GRACE-FA | LRD-GNN-Tensor · Chameleon | accuracy (%): 52.68±2.14 -> 66.27±1.27 | Improved |
| LRD-GNN-Tensor vs BGRL | LRD-GNN-Tensor · Cora | accuracy (%): 81.08±0.17 -> 83.74±0.61 | Improved |
| LRD-GNN-Tensor vs BGRL | LRD-GNN-Tensor · Amazon Photo | accuracy (%): 93.24±0.29 -> 93.26±0.15 | No clear effect |
| LRD-GNN-Tensor vs BGRL | LRD-GNN-Tensor · PubMed | accuracy (%): 79.97±0.36 -> 79.71±0.75 | Hurt |
| build tensor from attribute-similar nodes instead of structure-similar nodes (ablation) | LRD-GNN · Cora | accuracy (%): 78.20±0.65 -> 83.74±0.61 | Improved |
| drop A2X (higher-order information) from tensor variants (ablation) | LRD-GNN · Cora | accuracy (%): 83.74±0.61 -> 83.55±0.27 | No clear effect |
| remove raw attributes X and use only propagated features AX, A2X (ablation) | LRD-GNN · Cora | accuracy (%): 83.74±0.61 -> 80.32±0.49 | Hurt |
Closest papers by meaning
- Predicting Global Label Relationship Matrix for Graph Neural Networks under Heterophily (NeurIPS 2023 · similarity 0.89)
- High-Order Pooling for Graph Neural Networks with Tensor Decomposition (NeurIPS 2022 · similarity 0.88)
- DEGREE: Decomposition Based Explanation for Graph Neural Networks (ICLR 2022 · similarity 0.88)
- SLOG: An Inductive Spectral Graph Neural Network Beyond Polynomial Filter (ICML 2024 · similarity 0.87)
- Uncovering the Redundancy in Graph Self-supervised Learning Models (NeurIPS 2024 · similarity 0.87)
- Graph Neural Networks Are More Than Filters: Revisiting and Benchmarking from A Spectral Perspective (ICLR 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.
