Federated Learning Convergence Analysis on CIFAR10 Extreme non-i.i.d. partition
2.46Global Cycles to 30% AccpDFL
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| pDFLtau_r=5, Heterogeneous ML optimizers=true2026.06 | 2.46 | 5.02 | 9.42 | 17.31 | |
| SSD-FLtau_r=5, Heterogeneous ML optimizers=true2026.06 | 2.6 | 4.84 | 8.91 | 15.38 | |
| sDFLtau_r=5, Heterogeneous ML optimizers=true2026.06 | 2.63 | 5.13 | 9.91 | 16.98 | |
| SSD-FLtau_r=3, Heterogeneous ML optimizers=true2026.06 | 2.86 | 5.88 | 11.21 | — | |
| sDFLtau_r=3, Heterogeneous ML optimizers=true2026.06 | 2.91 | 6.42 | 12.42 | — | |
| cSTCtau_r=5, Heterogeneous ML optimizers=true2026.06 | 2.92 | 7.52 | 13.28 | — | |
| pDFLtau_r=3, Heterogeneous ML optimizers=true2026.06 | 3 | 6.47 | 12.6 | — | |
| cSTCtau_r=3, Heterogeneous ML optimizers=true2026.06 | 3.06 | 7.64 | 14.06 | — | |
| STCtau_r=5, Heterogeneous ML optimizers=true2026.06 | 3.37 | 6.67 | 15.31 | — | |
| STCtau_r=3, Heterogeneous ML optimizers=true2026.06 | 3.4 | 8.47 | 17.38 | — |