Research map / Federated learning and privacy / Personalized and heterogeneous FL
Forgetting Through Transforming: Enabling Federated Unlearning via Class-Aware Representation Transformation
Qi Guo, Zhen Tian, Minghao Yao, Saiyu Qi, Yong Qi, Bingyi Liu
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 transformation class selection (vs random label) (ablation) | FUCRT · CIFAR10 (Non-IID) | remaining-class accuracy: 89.24 -> 89.82 | Improved |
| add transformation class selection (vs random label) (ablation) | FUCRT · FMNIST (Non-IID) | remaining-class accuracy: 86.20 -> 90.95 | Improved |
| add transformation class selection (vs random label) (ablation) | FUCRT · EuroSAT (IID) | remaining-class accuracy: 90.19 -> 94.92 | Improved |
| add local class-aware contrastive loss (ablation) | FUCRT · CIFAR10 (Non-IID) | remaining-class accuracy: 20.87 -> 89.82 | Improved |
| add local class-aware contrastive loss (ablation) | FUCRT · EuroSAT (IID) | remaining-class accuracy: 23.10 -> 94.92 | Improved |
| add local class-aware contrastive loss (ablation) | FUCRT · CIFAR100 (IID) | remaining-class accuracy: 75.00 -> 75.25 | No clear effect |
| add global class-aware contrastive loss (ablation) | FUCRT · CIFAR10 (Non-IID) | remaining-class accuracy: 89.76 -> 89.82 | No clear effect |
| add global class-aware contrastive loss (ablation) | FUCRT · EuroSAT (Non-IID) | remaining-class accuracy: 83.41 -> 87.38 | Improved |
| add global class-aware contrastive loss (ablation) | FUCRT · CIFAR100 (IID) | remaining-class F1: 81.35 -> 85.70 | Improved |
| use FUCRT instead of fine-tuning on remaining data | Fine-tune · FMNIST (IID) | unlearning-class accuracy: 91.79 -> 0.00 | Improved |
| raise unlearning proportion from 10% to 50% | FUCRT · CIFAR10 (IID) | FUCRT minus From-scratch remaining accuracy: 90.06 vs 87.99 at 10%; 94.15 vs 95.45 at 50% | Hurt |
Closest papers by meaning
- Federated Unlearning with Gradient Adaptive Shaping (NeurIPS 2026 · similarity 0.90)
- Unlearning through Knowledge Overwriting: Reversible Federated Unlearning via Selective Sparse Adapter (CVPR 2025 · similarity 0.90)
- NoT: Federated Unlearning via Weight Negation (CVPR 2025 · similarity 0.89)
- Federated Unlearning with Gradient Descent and Conflict Mitigation (AAAI 2025 · similarity 0.88)
- Efficient Verified Unlearning For Distillation (NeurIPS 2025 · similarity 0.87)
- Federated Class-Incremental Learning: A Hybrid Approach Using Latent Exemplars and Data-Free Techniques to Address Local and Global Forgetting (ICLR 2025 · similarity 0.87)
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.
