Next character prediction on Shakespeare realistic non-IID LEAF benchmark (test)
54.52Top-1 AccSuPerFed-MM
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SuPerFed-MM# clients=660, Backbone=two stacked LSTM layers2021.09 | 54.52 | 84.27 | |
| SuPerFed-LM# clients=660, Backbone=two stacked LSTM layers2021.09 | 54.52 | 83.97 | |
| APFL# clients=660, Backbone=two stacked LSTM layers2021.09 | 54.08 | 83.32 | |
| FedProx# clients=660, Backbone=two stacked LSTM layers2021.09 | 51.33 | 80.31 | |
| FedAvg# clients=660, Backbone=two stacked LSTM layers2021.09 | 50.9 | 80.15 | |
| SCAFFOLD# clients=660, Backbone=two stacked LSTM layers2021.09 | 50.76 | 80.43 | |
| Ditto# clients=660, Backbone=two stacked LSTM layers2021.09 | 49.04 | 78.14 | |
| FedPer# clients=660, Backbone=two stacked LSTM layers2021.09 | 45.82 | 75.68 | |
| FedRep# clients=660, Backbone=two stacked LSTM layers2021.09 | 38.15 | 68.65 | |
| LG-FedAvg# clients=660, Backbone=two stacked LSTM layers2021.09 | 33.88 | 62.84 | |
| pFedMe# clients=660, Backbone=two stacked LSTM layers2021.09 | 32.29 | 63.12 |