FederatedDistillationOptions
Options & Configuration · AiDotNet.Models.Options
Configuration options for federated knowledge distillation, enabling model-heterogeneous FL.
Inheritance: ModelOptions → FederatedDistillationOptions
For Beginners
In standard federated learning, all clients must use the exact same model architecture. Knowledge distillation removes this constraint — a phone can use a tiny model while a server uses a large model, and they still learn from each other.
Constructors
| Constructor | Summary |
|---|---|
FederatedDistillationOptions() |
Properties
| Property | Summary |
|---|---|
FederatedDistillationStrategy Strategy { get; set; } |
Gets or sets the distillation strategy. |
double DistillationAlpha { get; set; } |
Gets or sets the weight for distillation loss vs task-specific loss. |
double Temperature { get; set; } |
Gets or sets the softmax temperature for soft label generation. |
int DistillationEpochs { get; set; } |
Gets or sets the number of local distillation epochs per round. |
int FeatureDimension { get; set; } |
Gets or sets the feature dimensionality for FedGEN synthetic data. |
int NumClasses { get; set; } |
Gets or sets the number of output classes for FedGEN. |