Personalized Federated Learning Unlearning on FEMNIST (natural partition)
100PPSRetrain
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RetrainpFL Architecture=FedPer2026.06 | 100 | 100 | |
| RetrainpFL Architecture=FedRep2026.06 | 100 | 100 | |
| RetrainpFL Architecture=Ditto2026.06 | 100 | 100 | |
| RetrainpFL Architecture=FedBN2026.06 | 100 | 100 | |
| pFedULpFL Architecture=Ditto2026.06 | 98.4 | 97.5 | |
| pFedULpFL Architecture=FedRep2026.06 | 98.1 | 97.1 | |
| pFedULpFL Architecture=FedPer2026.06 | 97.8 | 96.6 | |
| pFedULpFL Architecture=FedBN2026.06 | 97.4 | 96 | |
| Fine-tunepFL Architecture=Ditto2026.06 | 96.1 | 93.4 | |
| Fine-tunepFL Architecture=FedRep2026.06 | 95.6 | 92.8 | |
| Fine-tunepFL Architecture=FedPer2026.06 | 95.2 | 92.1 | |
| Fine-tunepFL Architecture=FedBN2026.06 | 94.7 | 91.4 | |
| FedEraser-adapt.pFL Architecture=Ditto2026.06 | 89.8 | 87.3 | |
| FedEraser-adapt.pFL Architecture=FedRep2026.06 | 89.1 | 86.4 | |
| FedEraser-adapt.pFL Architecture=FedPer2026.06 | 88.4 | 85.6 | |
| FedEraser-adapt.pFL Architecture=FedBN2026.06 | 87.6 | 84.5 | |
| GA-NaivepFL Architecture=Ditto2026.06 | 86.6 | 83.2 | |
| GA-NaivepFL Architecture=FedRep2026.06 | 85.8 | 82.1 | |
| GA-NaivepFL Architecture=FedPer2026.06 | 85.1 | 81.2 | |
| GA-NaivepFL Architecture=FedBN2026.06 | 84.3 | 80.1 |