Image Classification on CIFAR-100 (Test Accuracy and Train Loss)
85.85Accuracy (Test)FedAdamW
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FedAdamWBackbone=Swin Transformer, Data Distribution (Dirichlet)=Dir-0.1, Communication Rounds=100, Number of Clients=100, Participation Rate=5%, Batch Size=16, K=50, Evaluation Protocol=Fine-tuning2025.10 | 85.85 | 0.285 | |
| Local AdamWBackbone=Swin Transformer, Data Distribution (Dirichlet)=Dir-0.1, Communication Rounds=100, Number of Clients=100, Participation Rate=5%, Batch Size=16, K=50, Evaluation Protocol=Fine-tuning2025.10 | 83.35 | 0.381 | |
| FedCMBackbone=Swin Transformer, Data Distribution (Dirichlet)=Dir-0.1, Communication Rounds=100, Number of Clients=100, Participation Rate=5%, Batch Size=16, K=50, Evaluation Protocol=Fine-tuning2025.10 | 82.38 | 0.565 | |
| SCAFFOLDBackbone=Swin Transformer, Data Distribution (Dirichlet)=Dir-0.1, Communication Rounds=100, Number of Clients=100, Participation Rate=5%, Batch Size=16, K=50, Evaluation Protocol=Fine-tuning2025.10 | 81.3 | 0.514 | |
| FedAvgBackbone=Swin Transformer, Data Distribution (Dirichlet)=Dir-0.1, Communication Rounds=100, Number of Clients=100, Participation Rate=5%, Batch Size=16, K=50, Evaluation Protocol=Fine-tuning2025.10 | 80.02 | 0.588 | |
| Local AdamBackbone=Swin Transformer, Data Distribution (Dirichlet)=Dir-0.1, Communication Rounds=100, Number of Clients=100, Participation Rate=5%, Batch Size=16, K=50, Evaluation Protocol=Fine-tuning2025.10 | 79.75 | 0.534 | |
| FedAdamBackbone=Swin Transformer, Data Distribution (Dirichlet)=Dir-0.1, Communication Rounds=100, Number of Clients=100, Participation Rate=5%, Batch Size=16, K=50, Evaluation Protocol=Fine-tuning2025.10 | 77.48 | 0.651 | |
| FedLADABackbone=Swin Transformer, Data Distribution (Dirichlet)=Dir-0.1, Communication Rounds=100, Number of Clients=100, Participation Rate=5%, Batch Size=16, K=50, Evaluation Protocol=Fine-tuning2025.10 | 74.64 | 0.598 |