Research map / Graph learning / Graph self-supervised learning
Joint Learning of Label and Environment Causal Independence for Graph Out-of-Distribution Generalization
Shurui Gui, Meng Liu, Xiner Li, Youzhi Luo, Shuiwang Ji
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 |
|---|---|---|---|
| full LECI (label + environment adversarial independence) vs ERM | GNN · GOOD-Motif basis | accuracy (%): 60.93 -> 84.56 | Improved |
| add both label and environment adversarial training to basic network (ablation) | subgraph discovery network · GOOD-Motif basis | accuracy (%): 58.38 -> 84.56 | Improved |
| add both LA and EA to basic network (ablation) | subgraph discovery network · GOOD-Motif size | accuracy (%): 65.17 -> 71.43 | Improved |
| add only label adversarial (LA) to basic network (ablation) | subgraph discovery network · GOOD-Motif size | accuracy (%): 65.17 -> 53.57 | Hurt |
| add only environment adversarial (EA) to basic network (ablation) | subgraph discovery network · GOOD-Motif size | accuracy (%): 65.17 -> 38.69 | Hurt |
| add only EA to basic network (ablation) | subgraph discovery network · GOOD-Motif basis | accuracy (%): 58.38 -> 64.02 | Mixed |
| add pure feature shift consideration (PFSC) alone (ablation) | subgraph discovery network · GOOD-CMNIST color | accuracy (%): 33.41 -> 19.33 | Hurt |
| add LA, EA and PFSC (full LECI) to basic network (ablation) | subgraph discovery network · GOOD-CMNIST color | accuracy (%): 33.41 -> 51.80 | Improved |
| LECI environment exploitation vs GIL (IGA) on same GSAT network (ablation) | GSAT · GOOD-Motif basis | accuracy (%): 55.99 -> 84.56 | Improved |
| LECI vs GIL (IGA) exploitation (ablation) | GSAT · GOOD-SST2 | accuracy (%): 75.04 -> 83.44 | Improved |
| LECI vs GIL (IGA) exploitation (ablation) | GSAT · GOOD-Twitter | accuracy (%): 60.01 -> 59.64 | Hurt |
| LECI vs best baseline on real-world datasets | GNN · GOOD-SST2 | accuracy (%): 81.50 -> 83.44 | Improved |
Closest papers by meaning
- Subgraph Aggregation for Out-of-Distribution Generalization on Graphs (AAAI 2025 · similarity 0.90)
- Quantifying Distributional Invariance in Causal Subgraph for IRM-Free Graph Generalization (NeurIPS 2025 · similarity 0.90)
- Dissecting the Failure of Invariant Learning on Graphs (NeurIPS 2024 · similarity 0.90)
- Out-of-Distribution Generalization on Graphs via Progressive Inference (AAAI 2025 · similarity 0.88)
- Disentangled Graph Self-supervised Learning for Out-of-Distribution Generalization (ICML 2024 · similarity 0.88)
- Improving Generalization of Dynamic Graph Learning via Environment Prompt (NeurIPS 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.
