Multi-label classification on CIFAR-10 multi-label 28 (beta=0.1, gamma=0.5)
0.922Macro AUCCentralized
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Centralizedbeta (Dirichlet concentration parameter)=0.1, gamma (class-presence ratio)=0.5, Backbone=ResNet-18, Communication rounds=100, Local epochs=1, Batch size=32, Optimizer=AdamW2025.09 | 0.922 | 0.613 | 0.9004 | 0.6202 | |
| FedNCA-MLbeta (Dirichlet concentration parameter)=0.1, gamma (class-presence ratio)=0.5, Backbone=ResNet-18, Communication rounds=100, Local epochs=1, Batch size=32, Optimizer=AdamW2025.09 | 0.838 | 0.3809 | 0.789 | 0.416 | |
| SphereFedbeta (Dirichlet concentration parameter)=0.1, gamma (class-presence ratio)=0.5, Backbone=ResNet-18, Communication rounds=100, Local epochs=1, Batch size=32, Optimizer=AdamW2025.09 | 0.8062 | 0.3683 | 0.7837 | 0.3818 | |
| FedLGTbeta (Dirichlet concentration parameter)=0.1, gamma (class-presence ratio)=0.5, Backbone=ResNet-18, Communication rounds=100, Local epochs=1, Batch size=32, Optimizer=AdamW2025.09 | 0.8054 | 0.363 | 0.8065 | 0.3924 | |
| FedCurvbeta (Dirichlet concentration parameter)=0.1, gamma (class-presence ratio)=0.5, Backbone=ResNet-18, Communication rounds=100, Local epochs=1, Batch size=32, Optimizer=AdamW2025.09 | 0.7906 | 0.31 | 0.7746 | 0.3503 | |
| SCAFFOLDbeta (Dirichlet concentration parameter)=0.1, gamma (class-presence ratio)=0.5, Backbone=ResNet-18, Communication rounds=100, Local epochs=1, Batch size=32, Optimizer=AdamW2025.09 | 0.79 | 0.3138 | 0.7772 | 0.3554 | |
| FedAvgbeta (Dirichlet concentration parameter)=0.1, gamma (class-presence ratio)=0.5, Backbone=ResNet-18, Communication rounds=100, Local epochs=1, Batch size=32, Optimizer=AdamW2025.09 | 0.7892 | 0.3117 | 0.7762 | 0.3507 | |
| FedProxbeta (Dirichlet concentration parameter)=0.1, gamma (class-presence ratio)=0.5, Backbone=ResNet-18, Communication rounds=100, Local epochs=1, Batch size=32, Optimizer=AdamW2025.09 | 0.7882 | 0.3074 | 0.774 | 0.3502 |