Image Classification on DomainNet (Efficiency and Cost Metrics)
1.27Efficiency (E)Cluster-based
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Cluster-basedBackbone=cViT-M2026.03 | 1.27 | 52 | 0.41 | |
| FCL-perDeviceBackbone=cResNet-M2026.03 | 1.42 | 61 | 0.43 | |
| PFL-perDeviceBackbone=cResNet-M2026.03 | 1.76 | 60 | 0.34 | |
| Cluster-basedBackbone=cResNet-M2026.03 | 2.03 | 61 | 0.3 | |
| FCL-AveTrigBackbone=cResNet-M2026.03 | 2.1 | 61 | 0.29 | |
| FCL-perDeviceBackbone=cViT-M2026.03 | 2.32 | 58 | 0.25 | |
| PFL-AveTrigBackbone=cResNet-M2026.03 | 2.44 | 61 | 0.25 | |
| FCL-AveTrigBackbone=cViT-M2026.03 | 2.86 | 60 | 0.21 | |
| PFL-perDeviceBackbone=cViT-M2026.03 | 2.9 | 58 | 0.2 | |
| PFL-AveTrigBackbone=cViT-M2026.03 | 3.75 | 60 | 0.16 | |
| DriftGuardBackbone=cResNet-M2026.03 | 3.81 | 61 | 0.16 | |
| DriftGuardBackbone=cViT-M2026.03 | 8.57 | 60 | 0.07 |