Image Classification on CIFAR-100 90% symmetric label noise
58.56AccuracyFedSIR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FedSIRalpha=0.5, # Clean Clients=62026.04 | 58.56 | — | — | |
| FedSIRalpha=0.3, # Clean Clients=72026.04 | 58.43 | — | — | |
| FedSIRalpha=2, # Clean Clients=52026.04 | 58.11 | — | — | |
| FedELCalpha=0.3, # Clean Clients=72026.04 | 57.99 | — | — | |
| FedELCalpha=0.5, # Clean Clients=62026.04 | 57.85 | — | — | |
| FedNoRoalpha=2, # Clean Clients=52026.04 | 57.19 | — | — | |
| FedNoRoalpha=0.5, # Clean Clients=62026.04 | 57.07 | — | — | |
| FedELCalpha=2, # Clean Clients=52026.04 | 57.05 | — | — | |
| FedNedalpha=0.3, # Clean Clients=72026.04 | 56.36 | — | — | |
| FedNedalpha=2, # Clean Clients=52026.04 | 56.26 | — | — | |
| FedNoRoalpha=0.3, # Clean Clients=72026.04 | 55.92 | — | — | |
| FedNedalpha=0.5, # Clean Clients=62026.04 | 55.27 | — | — | |
| FedProxalpha=0.3, # Clean Clients=72026.04 | 54.27 | — | — | |
| FedAvgalpha=0.3, # Clean Clients=72026.04 | 53.1 | — | — | |
| FedProxalpha=0.5, # Clean Clients=62026.04 | 52.14 | — | — | |
| FedProxalpha=2, # Clean Clients=52026.04 | 51.78 | — | — | |
| FedAvgalpha=0.5, # Clean Clients=62026.04 | 51.25 | — | — | |
| FedAvgalpha=2, # Clean Clients=52026.04 | 51.2 | — | — | |
| FedCorralpha=2, # Clean Clients=52026.04 | 50.56 | — | — | |
| FedCorralpha=0.5, # Clean Clients=62026.04 | 49.52 | — | — | |
| FedCorralpha=0.3, # Clean Clients=72026.04 | 47.24 | — | — | |
| RoFLalpha=2, # Clean Clients=52026.04 | 46.69 | — | — | |
| RoFLalpha=0.5, # Clean Clients=62026.04 | 45.24 | — | — | |
| FedLSRalpha=0.3, # Clean Clients=72026.04 | 43.53 | — | — | |
| FedLSRalpha=0.5, # Clean Clients=62026.04 | 42.38 | — | — | |
| FedLSRalpha=2, # Clean Clients=52026.04 | 41.63 | — | — | |
| RoFLalpha=0.3, # Clean Clients=72026.04 | 40.72 | — | — | |
| RHFLalpha=2, # Clean Clients=52026.04 | 16.65 | — | — | |
| RHFLalpha=0.5, # Clean Clients=62026.04 | 15.39 | — | — | |
| RHFLalpha=0.3, # Clean Clients=72026.04 | 13.37 | — | — | |
| C2D (DivideMix with SimCLR)Architecture=PreAct ResNet-182021.03 | — | 58.7 | 58.45 | |
| C2D (DivideMix with SimCLR)Architecture=ResNet-502021.03 | — | 64.3 | 63.91 | |
| C2D (ELR+ with SimCLR)Architecture=ResNet-342021.03 | — | 55.08 | 54.06 | |
| CE+mixup with SimCLRArchitecture=ResNet-342021.03 | — | 55.51 | 54.64 | |
| DivideMixArchitecture=PreAct ResNet-182021.03 | — | 31.5 | 31 | |
| ELR+Architecture=ResNet-342021.03 | — | — | 33.4 | |
| Meta-learningArchitecture=PreAct ResNet-322021.03 | — | 19.5 | 14.3 |