MASTrainerOptions<T>
Options & Configuration · AiDotNet.ContinualLearning.Trainers
Configuration options for the MAS trainer.
For Beginners
These options control how the MAS trainer operates. MAS protects important parameters by measuring how sensitive the network output is to each parameter, without needing task labels.
Constructors
| Constructor | Summary |
|---|---|
MASTrainerOptions() |
Properties
| Property | Summary |
|---|---|
bool? ComputeValidationMetrics { get; set; } |
Whether to compute validation metrics after each epoch. |
bool? UseBatchedImportance { get; set; } |
Whether to compute importance in batches for efficiency. |
bool? UseExperienceReplay { get; set; } |
Whether to use experience replay alongside MAS regularization. |
double? GradientClipThreshold { get; set; } |
Gradient clipping threshold. |
double? RegularizationWeight { get; set; } |
Weight for the regularization loss relative to task loss. |
double? ReplayLearningRateFactor { get; set; } |
Fraction of learning rate to use for replay samples. |
int? ImportanceBatchSize { get; set; } |
Batch size for importance computation. |
int? ImportanceSamples { get; set; } |
Number of samples to use for importance estimation after task completion. |