Federated Learning on CIFAR-100 (Dirichlet 0.3, Specific Accuracy Convergence)
32.7Accuracy (500R)FedACG
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| FedACGparticipation ratio=4%, clients=250, dynamic updates=true2022.01 | 32.7 | 41.51 | 376 | 769 | |
| FedDCparticipation ratio=4%, clients=250, dynamic updates=true, communication cost=2x2022.01 | 31.35 | 36.82 | 469 | 1,000 | |
| MOONparticipation ratio=4%, clients=250, dynamic updates=true2022.01 | 30.88 | 39.57 | 430 | 852 | |
| FedDynparticipation ratio=4%, clients=250, dynamic updates=true2022.01 | 29.45 | 38.47 | 517 | 941 | |
| FedAvgparticipation ratio=4%, clients=250, dynamic updates=true2022.01 | 28.61 | 35.87 | 577 | 1,000 | |
| FedNTDparticipation ratio=4%, clients=250, dynamic updates=true2022.01 | 28.45 | 36.26 | 578 | 1,000 | |
| FedProxparticipation ratio=4%, clients=250, dynamic updates=true2022.01 | 28.17 | 35.89 | 602 | 1,000 |