Federated Unlearning on CIFAR-10 (Retain)
100AccuracyRetrain
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RetrainModel=ResNet-18, Communication Cost (Bytes)=1.23e12, Computation Cost (FLOPs)=5.66e182025.03 | 100 | 0 | |
| FCUModel=ResNet-18, Communication Cost (Bytes)=5.73e11, Computation Cost (FLOPs)=2.64e182025.03 | 100 | 0 | |
| MODEModel=ResNet-18, Communication Cost (Bytes)=5.88e11, Computation Cost (FLOPs)=2.71e182025.03 | 99.8 | 0.2 | |
| NoTModel=ResNet-18, Communication Cost (Bytes)=5.42e11, Computation Cost (FLOPs)=2.49e182025.03 | 99.77 | 0.23 | |
| PGDModel=ResNet-18, Communication Cost (Bytes)=6.04e11, Computation Cost (FLOPs)=2.78e182025.03 | 99.51 | 0.49 | |
| FTModel=ResNet-18, Communication Cost (Bytes)=5.94e11, Computation Cost (FLOPs)=2.73e182025.03 | 99.45 | 0.55 | |
| PGDModel=CNN, Communication Cost (Bytes)=1.19e10, Computation Cost (FLOPs)=5.13e162025.03 | 92.62 | 0.96 | |
| MODEModel=CNN, Communication Cost (Bytes)=1.10e10, Computation Cost (FLOPs)=4.77e162025.03 | 92.56 | 0.9 | |
| FTModel=CNN, Communication Cost (Bytes)=9.39e09, Computation Cost (FLOPs)=4.06e162025.03 | 92.48 | 0.82 | |
| FCUModel=CNN, Communication Cost (Bytes)=1.33e10, Computation Cost (FLOPs)=5.75e162025.03 | 92.46 | 0.8 | |
| NoTModel=CNN, Communication Cost (Bytes)=7.09e09, Computation Cost (FLOPs)=3.06e162025.03 | 91.69 | 0.03 | |
| RetrainModel=CNN, Communication Cost (Bytes)=1.35e10, Computation Cost (FLOPs)=5.81e162025.03 | 91.66 | 0 | |
| FedEraserModel=CNN, Communication Cost (Bytes)=1.34e10, Computation Cost (FLOPs)=5.79e162025.03 | 88.19 | 3.47 | |
| FUKDModel=CNN, Communication Cost (Bytes)=1.33e10, Computation Cost (FLOPs)=5.77e162025.03 | 82.69 | 8.97 |