Research map / Federated learning and privacy / Personalized and heterogeneous FL
Influence-Disentangled Federated Training: Learning Models That Are Easy to Unlearn
Canran Xiao, Qianyu Chen, Liwei Hou
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 shared-subspace routing (r>0) (ablation) | IDFT · CIFAR-10 (ResNet-18), client-wise unlearning (IID, 10 clients) | Avg. Gap: 1.05 -> 0.78 | Improved |
| use influence dimension s larger than 32 (ablation) | IDFT · CIFAR-10 (ResNet-18), client-wise unlearning (IID, 10 clients) | Avg. Gap: 0.86 -> 0.78 | Improved |
| add entanglement shrinkage (lambda>0) (ablation) | IDFT · CIFAR-10 (ResNet-18), client-wise unlearning (IID, 10 clients) | Avg. Gap: 0.84 -> 0.78 | Improved |
| add covariance term to entanglement score (ablation) | IDFT · CIFAR-10 (ResNet-18), client-wise unlearning (IID, 10 clients) | Avg. Gap: 0.89 -> 0.78 | Improved |
| add co-adaptation term to entanglement score (ablation) | IDFT · CIFAR-10 (ResNet-18), client-wise unlearning (IID, 10 clients) | Avg. Gap: 0.88 -> 0.78 | Improved |
| add leakage term to entanglement score (ablation) | IDFT · CIFAR-10 (ResNet-18), client-wise unlearning (IID, 10 clients) | Avg. Gap: 0.90 -> 0.78 | Improved |
| add constrained repair rounds after subtraction (ablation) | IDFT · CIFAR-10 (ResNet-18), client-wise unlearning (IID, 10 clients) | Avg. Gap: 1.22 -> 0.78 | Improved |
| add constrained repair rounds after subtraction (ablation) | IDFT · CIFAR10-ResNet18 (client unlearning, label-flipping) | MIA: 0.56 -> 0.45 | Improved |
| add shared-subspace routing (r>0) (ablation) | IDFT · CIFAR10-ResNet18 (client unlearning, label-flipping) | MIA: 0.50 -> 0.45 | Improved |
| use IDFT instead of NoT | NoT · CIFAR-10 (ResNet-18), client-wise unlearning (IID, 10 clients) | Avg. Gap: 1.78 -> 0.78 | Improved |
| use IDFT instead of NoT | NoT · CIFAR-10 (CNN), client-wise unlearning (IID, 10 clients) | Avg. Gap: 0.29 -> 0.26 | Improved |
| use IDFT instead of FUSED | FUSED · CIFAR100-ResNet18 (client unlearning, label-flipping) | MIA: 0.48 -> 0.10 | Improved |
Closest papers by meaning
- NoT: Federated Unlearning via Weight Negation (CVPR 2025 · similarity 0.89)
- FedImpro: Measuring and Improving Client Update in Federated Learning (ICLR 2024 · similarity 0.87)
- Unlearning through Knowledge Overwriting: Reversible Federated Unlearning via Selective Sparse Adapter (CVPR 2025 · similarity 0.87)
- Efficient Federated Learning against Heterogeneous and Non-stationary Client Unavailability (NeurIPS 2024 · similarity 0.87)
- FLUX: Efficient Descriptor-Driven Clustered Federated Learning under Arbitrary Distribution Shifts (NeurIPS 2025 · similarity 0.87)
- Reviving Stale Updates: Data-Free Knowledge Distillation for Asynchronous Federated Learning (NeurIPS 2026 · 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.
