Image Classification on FMNIST (test) with Alpha-Weighted Accuracy
89.02Accuracy (alpha=0.1)FedAvg
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| FedAvgModel=VGG 16, Parameters Up /Down=15.1 /15.1 M, FLOPs=40.3 G, Sys. Heter.=×2026.03 | 89.02 | 93.09 | 93.3 | |
| CA-HFPModel=VGG 16, Parameters Up /Down=3.95 /3.95 M, FLOPs=10.2 G, Sys. Heter.=✓2026.03 | 89 | 93.12 | 93.7 | |
| FedProxModel=Resnet 32, Parameters Up /Down=466 /466 K, FLOPs=4.45 G, Sys. Heter.=×2026.03 | 86.65 | 92.4 | 92.51 | |
| FedProxModel=VGG 16, Parameters Up /Down=15.1 /15.1 M, FLOPs=20.2 G, Sys. Heter.=×2026.03 | 85.46 | 93.47 | 69.48 | |
| CA-HFPModel=Resnet 32, Parameters Up /Down=254 /254 K, FLOPs=5.20 G, Sys. Heter.=✓2026.03 | 80.9 | 90.13 | 90.59 | |
| FedAvgModel=Resnet 32, Parameters Up /Down=466 /466 K, FLOPs=9.01 G, Sys. Heter.=×2026.03 | 80.8 | 92.7 | 91.59 | |
| PruneFLModel=VGG 16, Parameters Up /Down=4.30 /4.30 M, FLOPs=20.1 G, Sys. Heter.=×2026.03 | 77.08 | 92.54 | 92.94 | |
| PruneFLModel=Resnet 32, Parameters Up /Down=261 /261 K, FLOPs=5.45 G, Sys. Heter.=×2026.03 | 73.9 | 85.13 | 81.92 | |
| DapperFLModel=Resnet 32, Parameters Up /Down=254 /466 K, FLOPs=5.23 G, Sys. Heter.=✓2026.03 | 73.53 | 85.11 | 88.98 | |
| FedMPModel=VGG 16, Parameters Up /Down=3.95 /3.95 M, FLOPs=10.2 G, Sys. Heter.=✓2026.03 | 73.34 | 92.47 | 92.79 | |
| FedMPModel=Resnet 32, Parameters Up /Down=254 /254 K, FLOPs=5.20 G, Sys. Heter.=✓2026.03 | 67.49 | 80.01 | 89.99 | |
| DapperFLModel=VGG 16, Parameters Up /Down=3.95 /15.1 M, FLOPs=10.3 G, Sys. Heter.=✓2026.03 | 59.6 | 91.08 | 90.74 |