Research map / Federated learning and privacy / Communication-efficient FL
Computation and Communication Efficient Federated Unlearning via On-server Gradient Conflict Mitigation and Expression
Minh-Duong Nguyen, Senura Wanasekara, Le-Tuan Nguyen, Quoc-Viet Pham, Ken-Tye Yong, Nguyen H. Tran, Dung D. Le
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 on-server gradient-matching unlearning after the L2U stage (ablation) | FOUL · PACS | forget accuracy: 69.53 -> 70.97 | Hurt |
| add on-server gradient-matching unlearning after the L2U stage (ablation) | FOUL · PACS | retain accuracy: 93.11 -> 92.33 | No clear effect |
| add on-server gradient-matching unlearning after the L2U stage (ablation) | FOUL · TerraIncognita | test accuracy: 38.16 -> 39.16 | Improved |
| use FOUL instead of retraining from scratch on retain clients | FedAvg · PACS | retain accuracy: 82.84 -> 92.33 | Improved |
| use FOUL instead of retraining from scratch on retain clients | FedAvg · PACS | test accuracy: 77.45 -> 76.43 | No clear effect |
| use FOUL instead of approximate unlearning baselines such as FUSED | FedAvg · PACS | forget accuracy: 75.94 -> 70.97 | Improved |
| use FOUL instead of approximate unlearning baselines | FedAvg · PACS | MIA: 50.02 -> 51.93 | Hurt |
| use FOUL instead of FUSED | FedAvg · PACS | communication per round (MB): 0.98 -> 16.02 | Hurt |
| use FOUL instead of retraining | FedAvg · PACS | computation (FLOPs): 5.81e16 -> 2.35e16 | Improved |
| use on-server gradient matching instead of naive FedAvg retraining | FedAvg · PACS | time-to-forget: 0.13 -> 0.32 | Improved |
| unlearn only the non-causal featurizer during unlearning | FOUL · PACS | retain accuracy: 82.84 -> 92.33 | Improved |
Closest papers by meaning
- Federated Unlearning with Gradient Adaptive Shaping (NeurIPS 2026 · similarity 0.92)
- NoT: Federated Unlearning via Weight Negation (CVPR 2025 · similarity 0.91)
- Federated Unlearning with Gradient Descent and Conflict Mitigation (AAAI 2025 · similarity 0.91)
- Unlearning through Knowledge Overwriting: Reversible Federated Unlearning via Selective Sparse Adapter (CVPR 2025 · similarity 0.90)
- Efficient Federated Learning via Clients-to-Server Knowledge Distillation (Student Abstract) (AAAI 2025 · similarity 0.86)
- Efficient Verified Unlearning For Distillation (NeurIPS 2025 · similarity 0.86)
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.
