IFederatedServerOptimizer<T>
Interfaces · AiDotNet.Interfaces
Applies a server-side optimization step in federated learning (FedOpt family).
Known implementations: FederatedServerOptimizerBase<T>
How It Works
For Beginners: In classic FedAvg, the server simply replaces the global model with the averaged client model. In FedOpt, the server treats the aggregated update like a “gradient” and applies an optimizer step (like Adam), which can improve stability and convergence, especially with non-IID data.
Methods
| Method | Summary |
|---|---|
Vector<T> Step(Vector<T> currentGlobalParameters, Vector<T> aggregatedTargetParameters) |
Updates global parameters given the current parameters and an aggregated target. |
string GetOptimizerName() |
Gets the name of the server optimizer. |