Research map / Federated learning and privacy / Personalized and heterogeneous FL
PAGE: A Unified Approach for Federated Graph Unlearning
Yuming Ai, Xunkai Li, Jiaqi Chao, Bowen Fan, Zhengyu Wu, Yinlin Zhu, Rong-Hua Li, Guoren 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 PAGE instead of the best baseline for client unlearning | federated graph unlearning · 8 datasets, client unlearning | accuracy: +5.08% average over SOTA | Improved |
| use PAGE instead of the best baseline for meta unlearning | federated graph unlearning · 8 datasets, meta unlearning | accuracy: +1.50% average over SOTA | Improved |
| scale up graph size (small to large datasets) | PAGE client unlearning · Photo vs ogbn-arxiv | accuracy gain over SOTA: 2.13% on Photo vs 11.84% on ogbn-arxiv | Improved |
| add the influenced unlearn component (adversarial graph + negative distillation) to existing graph unlearning methods (ablation) | GIF, D2DGN, MEGU, FedLU, FedDM · Cora, meta unlearning | prediction performance: +2.82% average | Improved |
| add the influenced unlearn component to existing graph unlearning methods (ablation) | GIF, D2DGN, MEGU, FedLU, FedDM · Cora, meta unlearning | unlearning performance (AUC-ROC): +4.64% average | Improved |
| add the influenced unlearn component to existing graph unlearning methods (ablation) | GIF, D2DGN, MEGU, FedLU, FedDM · CS, meta unlearning | prediction performance: +4.49% | Improved |
| add the influenced unlearn component to existing graph unlearning methods (ablation) | GIF, D2DGN, MEGU, FedLU, FedDM · CS, meta unlearning | unlearning performance (AUC-ROC): +7.22% | Improved |
| use PAGE under increasing edge attack ratio | meta unlearning on Photo (GAT, SGC) · Photo with poisoned edges | accuracy: 2.94% to 8.62% higher than SOTA | Improved |
| use PAGE instead of FedLU for edge and feature meta unlearning | Physics and Computers, Metis and Metis-Plus splits · Table 2 | accuracy | Mixed |
| raise the client unlearn ratio from 0.1 to 0.9 (ablation) | PAGE · Chameleon, Minesweeper (GraphSAGE, louvain) | accuracy | No clear effect |
| use PAGE on small-scale datasets versus baselines | client unlearning · small datasets (Cora, PubMed) | accuracy | No clear effect |
Closest papers by meaning
- Erase Then Rectify: A Training-Free Parameter Editing Approach for Cost-Effective Graph Unlearning (AAAI 2025 · similarity 0.90)
- Community-Centric Graph Unlearning (AAAI 2025 · similarity 0.89)
- FedSPA: Generalizable Federated Graph Learning under Homophily Heterogeneity (CVPR 2025 · similarity 0.89)
- Virtual Nodes Can Help: Tackling Distribution Shifts in Federated Graph Learning (AAAI 2025 · similarity 0.88)
- Federated Graph Learning under Domain Shift with Generalizable Prototypes (AAAI 2024 · similarity 0.88)
- $S^2$FGL: Spatial Spectral Federated Graph Learning (ICML 2025 · 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.
