Image Classification on CIFAR-10 (Directional Accuracy Evaluation)
90.96Accuracy (Dir-0.6)FedACT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FedACTBackbone=ResNet-18, Communication Rounds=3002026.07 | 90.96 | 89.83 | 84.49 | |
| FedAdamWBackbone=ResNet-18, Communication Rounds=3002026.07 | 90.69 | 89.37 | 84.27 | |
| FedLADABackbone=ResNet-18, Communication Rounds=3002026.07 | 89.78 | 88.77 | 83.82 | |
| FedAvgBackbone=ResNet-18, Communication Rounds=3002026.07 | 89.76 | 88.29 | 83.52 | |
| FedProxBackbone=ResNet-18, Communication Rounds=3002026.07 | 89.64 | 87.76 | 82.91 | |
| SCAFFOLDBackbone=ResNet-18, Communication Rounds=3002026.07 | 89.45 | 88.3 | 83.69 | |
| LocalAdamWBackbone=ResNet-18, Communication Rounds=3002026.07 | 89.06 | 87.08 | 80.86 | |
| LocalAdamBackbone=ResNet-18, Communication Rounds=3002026.07 | 88.59 | 87.29 | 82.05 | |
| FedAdamBackbone=ResNet-18, Communication Rounds=3002026.07 | 87.6 | 86.87 | 77.92 |