Multi-class Classification on EMNIST (val)
86.2AccuracyI-CGM-RG-SAGA
Evaluation results
| Method | Links | |
|---|---|---|
| I-CGM-RG-SAGABackbone=6-layer residual CNN, Outer iterations=100, Local stepsize (1/eta)=0.02, lambda=0.01, p=0.01, beta=0.2, q=0.001, t0=0, Batch size=1282025.12 | 86.2 | |
| I-CGM-RG-SVRGBackbone=6-layer residual CNN, Outer iterations=100, Local stepsize (1/eta)=0.02, lambda=0.01, p=0.01, beta=0.2, q=0.001, Batch size=1282025.12 | 86 | |
| SCAFFOLDBackbone=6-layer residual CNN, Outer iterations=100, Local stepsize (1/eta)=0.02, K=100, q=0.001, Batch size=1282025.12 | 85.9 | |
| FEDAVGBackbone=6-layer residual CNN, Outer iterations=100, Local stepsize (1/eta)=0.02, K=100, Batch size=1282025.12 | 85.6 | |
| SABER-FULLBackbone=6-layer residual CNN, Outer iterations=100, Local stepsize (1/eta)=0.02, lambda=0.01, p=0.01, beta=0.2, q=0.001, Batch size=1282025.12 | 85.3 | |
| SCAFFNEWBackbone=6-layer residual CNN, Outer iterations=100, Local stepsize (1/eta)=0.02, p=0.01, q=0.001, Batch size=1282025.12 | 84.9 |