FedDyn

Implementation of the FedDyn [FedDyn21] algorithm.

References

[FedDyn21]

Durmus Alp Emre Acar, Yue Zhao, Ramon Matas, Matthew Mattina, Paul Whatmough, and Venkatesh Saligrama. Federated Learning with Dynamic Regularization. In ICLR (2021). URL: https://openreview.net/pdf?id=B7v4QMR6Z9w

Classes included in fluke.algorithms.feddyn

FedDynClient

FedDynServer

FedDyn

Classes

class fluke.algorithms.feddyn.FedDynClient(index: int, train_set: FastDataLoader, test_set: FastDataLoader, optimizer_cfg: OptimizerConfigurator, loss_fn: Module, local_epochs: int, alpha: float, fine_tuning_epochs: int = 0, clipping: float = 5, **kwargs)[source]

Bases: Client

class fluke.algorithms.feddyn.FedDynServer(model: Module, test_set: FastDataLoader, clients: Collection[Client], weighted: bool = True, alpha: float = 0.01)[source]

Bases: Server

class fluke.algorithms.feddyn.FedDyn(n_clients: int, data_splitter: DataSplitter, hyper_params: DDict | dict[str, Any], **kwargs)[source]

Bases: CentralizedFL