Image Classification on CIFAR100 (test) (Accuracy vs Communication Cost)
92Test AccuracyFedBCGD
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FedBCGDBackbone=ViT-Base, Heterogeneity (ρ)=0.6, Number of clients (N)=6, Total communication floats budget=100d, Target Accuracy=88%2026.03 | 92 | 7 | |
| FedBCGD+Backbone=ViT-Base, Heterogeneity (ρ)=0.6, Number of clients (N)=6, Total communication floats budget=100d, Target Accuracy=88%2026.03 | 90.6 | 14 | |
| FedAvgBackbone=ViT-Base, Heterogeneity (ρ)=0.6, Number of clients (N)=6, Total communication floats budget=100d, Target Accuracy=88%2026.03 | 90.4 | 24 | |
| FedAvgMBackbone=ViT-Base, Heterogeneity (ρ)=0.6, Number of clients (N)=6, Total communication floats budget=100d, Target Accuracy=88%2026.03 | 88.7 | 32 | |
| SCAFFOLDBackbone=ViT-Base, Heterogeneity (ρ)=0.6, Number of clients (N)=6, Total communication floats budget=100d, Target Accuracy=88%2026.03 | 88.2 | 88 | |
| FedAdamBackbone=ViT-Base, Heterogeneity (ρ)=0.6, Number of clients (N)=6, Total communication floats budget=100d, Target Accuracy=88%2026.03 | 87.6 | — | |
| FedDCBackbone=ViT-Base, Heterogeneity (ρ)=0.6, Number of clients (N)=6, Total communication floats budget=100d, Target Accuracy=88%2026.03 | 85.8 | 25 | |
| Centralised SGDBackbone=ViT-Base, Heterogeneity (ρ)=0.6, Number of clients (N)=6, Total communication floats budget=100d, Target Accuracy=88%2026.03 | 81.5 | — |