Image Classification on PneumoniaMNIST alpha = 0.1 (Severe Non-IID) (test)
89.1AccuracyLocal-Best
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Local-Besttraining_mode=independent hospital training2026.05 | 89.1 | 87.26 | 8.64 | |
| Ours-Dropoutablation=dropout2026.05 | 87.5 | 85.04 | 15.17 | |
| Centralizedtraining_mode=pooled data2026.05 | 86.54 | 83.92 | 9.13 | |
| Oursprotocol=ChainLearn2026.05 | 78.53 | 70.11 | 17.52 | |
| FedMDaggregation=temperature-scaled probabilities2026.05 | 77.76 | 73.5 | 9.93 | |
| FedAvgrounds=5, backbone=ResNet-502026.05 | 75.9 | 63.59 | 21.02 | |
| EqualWt-Ensweighting=equal (Wi=1)2026.05 | 72.44 | 62.32 | 17.99 | |
| FedProxstabilization=proximal2026.05 | 71.35 | 56.49 | 22.16 |