FedAdamServerOptimizer<T>
Models & Types · AiDotNet.FederatedLearning.ServerOptimizers
FedAdam server optimizer — adaptive learning rates with momentum and second-moment estimation.
Inheritance: FederatedLearningComponentBase<T> → FederatedServerOptimizerBase<T> → FedAdamServerOptimizer<T>
For Beginners
Adam combines the benefits of Adagrad (adaptive learning rates) and momentum (smoothed gradient direction). FedAdam applies this at the server level, treating aggregated client updates as pseudo-gradients for server-side optimization.
How It Works
Reference: Reddi, S., et al. (2021). “Adaptive Federated Optimization.” ICLR 2021.
Constructors
| Constructor | Summary |
|---|---|
FedAdamServerOptimizer(double learningRate, double beta1, double beta2, double epsilon) |
Methods
| Method | Summary |
|---|---|
Vector<T> Step(Vector<T> currentGlobalParameters, Vector<T> aggregatedTargetParameters) |
|
string GetOptimizerName() |