Image Classification on CIFAR100 (test) (Accuracy and Communication Floats to 45%)
59.9AccuracyFedBCGD+
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FedBCGD+Heterogeneity (rho)=0.1, Total communication floats=1000d, Backbone=ResNet-18, Number of blocks (N)=52026.03 | 59.9 | 200 | |
| FedBCGDHeterogeneity (rho)=0.1, Total communication floats=1000d, Backbone=ResNet-18, Number of blocks (N)=52026.03 | 59.5 | 147 | |
| FedAdamHeterogeneity (rho)=0.1, Total communication floats=1000d, Backbone=ResNet-18, Number of blocks (N)=52026.03 | 49.9 | 610 | |
| FedAvgMHeterogeneity (rho)=0.1, Total communication floats=1000d, Backbone=ResNet-18, Number of blocks (N)=52026.03 | 48.3 | 769 | |
| FedDCHeterogeneity (rho)=0.1, Total communication floats=1000d, Backbone=ResNet-18, Number of blocks (N)=52026.03 | 46.6 | 278 | |
| FedAvgHeterogeneity (rho)=0.1, Total communication floats=1000d, Backbone=ResNet-18, Number of blocks (N)=52026.03 | 45.8 | 741 | |
| SCAFFOLDHeterogeneity (rho)=0.1, Total communication floats=1000d, Backbone=ResNet-18, Number of blocks (N)=52026.03 | 44.3 | — |