Research map / Graph learning / Graph self-supervised learning
Improving Subgraph Extraction for Graph Invariant Learning via Graph Sinkhorn Attention
Junchi Yan, Fangyu Ding, J. F. Sun, Zhaoping Hu, Yunyi Zhou, Lei Zhu
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 |
|---|---|---|---|
| add Gumbel noise (ablation) | GSINA with GIN · SPMotif b=0.5 | accuracy: 48.27 -> 55.16 | Improved |
| add Gumbel noise (ablation) | GSINA with PNA · SPMotif b=0.5 | accuracy: 69.95 -> 76.39 | Improved |
| add Gumbel noise (ablation) | GSINA with GIN · SPMotif b=0.9 | accuracy: 49.86 vs 50.28 | Hurt |
| add node attention (ablation) | GSINA with PNA · SPMotif b=0.7 | accuracy: 58.70 -> 73.96 | Improved |
| add node attention (ablation) | GSINA with GIN · SPMotif b=0.7 | accuracy: 54.63 -> 56.83 | Improved |
| add both Gumbel noise and node attention (ablation) | GSINA with GIN · SPMotif b=0.5 | accuracy: 46.28 -> 55.16 | Improved |
| add both Gumbel noise and node attention (ablation) | GSINA with PNA · SPMotif b=0.5 | accuracy: 71.75 -> 76.39 | Improved |
| use GSINA instead of GSAT attention | PNA backbone · SPMotif b=0.5 | accuracy: 68.74 -> 76.39 | Improved |
| use GSINA instead of GSAT attention | GIN backbone · OGBG-Molhiv | AUC: 76.47 -> 77.99 | Improved |
| use GSINA instead of CIGAv2 top-k | CIGAv2 · SPMotif-Mixed b=0.9 | accuracy: 54.25 -> 68.89 | Improved |
| use GSINA instead of CIGAv2 top-k | CIGAv2 · SPMotif-Struc b=0.33 | accuracy: 77.33 -> 75.49 | Hurt |
| use GSINA instead of CIGAv2 top-k | CIGAv2 · DrugOOD-Assay | ROC-AUC: 73.17 -> 72.84 | No clear effect |
Closest papers by meaning
- Subgraph Invariant Learning Towards Large-Scale Graph Node Classification (AAAI 2025 · similarity 0.91)
- Disentangling Invariant Subgraph via Variance Contrastive Estimation under Distribution Shifts (ICML 2025 · similarity 0.90)
- Graph Invariant Learning with Subgraph Co-mixup for Out-of-Distribution Generalization (AAAI 2024 · similarity 0.89)
- Subgraph Aggregation for Out-of-Distribution Generalization on Graphs (AAAI 2025 · similarity 0.89)
- A Flexible, Equivariant Framework for Subgraph GNNs via Graph Products and Graph Coarsening (NeurIPS 2024 · similarity 0.89)
- Dissecting the Failure of Invariant Learning on Graphs (NeurIPS 2024 · similarity 0.89)
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.
