Research map / Graph learning / Graph self-supervised learning
Compositional Fairness Constraints for Graph Embeddings
Avishek Bose, William Hamilton
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 compositional adversarial filters for gender, age, and occupation | MovieLens1M embeddings · MovieLens1M | RMSE: 0.865 -> 1.01 | Hurt |
| Add compositional adversarial filters for sensitive attributes | MovieLens1M embeddings · MovieLens1M | Gender AUC for sensitive-attribute prediction: 0.712 -> 0.511 | Improved |
| Add compositional adversarial filters for sensitive attributes | Freebase15k-237 TransD · Freebase15k-237 | Mean rank: 285 -> 542 | Hurt |
| Add compositional adversarial filters rather than separate per-attribute adversarial models | MovieLens1M embeddings · MovieLens1M | Gender AUC for sensitive-attribute prediction: 0.532 -> 0.511 | Improved |
| Add compositional adversarial filters rather than separate per-attribute adversarial models | Freebase15k-237 TransD · Freebase15k-237 | Mean rank: 320 -> 542 | Hurt |
| Increase adversarial regularization strength | Compositional graph embeddings · MovieLens1M | RMSE | Hurt |
| Train with random combinations of sensitive-attribute filters | Compositional encoder · Reddit, held-out attribute combinations | Sensitive-attribute AUC: 0.025 | Improved |
| Increase the number of sensitive attributes from 20 to 50 | Compositional adversary · Reddit, held-out combinations | Held-out AUC: 0.569; 0.569; 0.556; 0.519 | Mixed |
| Use compositional rather than single-attribute adversarial training | Reddit graph embeddings · Reddit | Sensitive-attribute prediction | Hurt |
| Use compositional rather than single-attribute adversarial training | Reddit graph embeddings · Reddit | Edge-prediction performance | Hurt |
Closest papers by meaning
- Stable Fair Graph Representation Learning with Lipschitz Constraint (ICML 2025 · similarity 0.87)
- DANCE: Dual Unbiased Expansion with Group-acquired Alignment for Out-of-distribution Graph Fairness Learning (ICML 2025 · similarity 0.85)
- Chasing Fairness in Graphs: A GNN Architecture Perspective (AAAI 2024 · similarity 0.84)
- Fair Graph U-Net: A Fair Graph Learning Framework Integrating Group and Individual Awareness (AAAI 2025 · similarity 0.84)
- FairSIN: Achieving Fairness in Graph Neural Networks through Sensitive Information Neutralization (AAAI 2024 · similarity 0.84)
- Adversarial Fairness Network (AAAI 2024 · similarity 0.84)
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.
