Interface IFederatedServerOptimizer<T>

Namespace

AiDotNet.Interfaces

Assembly

AiDotNet.dll

Applies a server-side optimization step in federated learning (FedOpt family).

public interface IFederatedServerOptimizer<T>

Type Parameters

T

Numeric type.

Remarks

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

GetOptimizerName()

Gets the name of the server optimizer.

string GetOptimizerName()

Returns

string

Step(Vector<T>, Vector<T>)

Updates global parameters given the current parameters and an aggregated target.

Vector<T> Step(Vector<T> currentGlobalParameters, Vector<T> aggregatedTargetParameters)

Parameters

currentGlobalParameters Vector<T>

The current global parameter vector.

aggregatedTargetParameters Vector<T>

The aggregated target parameter vector (e.g., FedAvg output).

Returns

Vector<T>

The updated global parameter vector.