SDCLAlgorithm<T, TInput, TOutput>
Models & Types · AiDotNet.MetaLearning.Algorithms
Implementation of SDCL: Self-Distillation Collaborative Learning for meta-learning.
Inheritance: ModelBase<T, TInput, TOutput> → MetaLearnerBase<T, TInput, TOutput> → SDCLAlgorithm<T, TInput, TOutput>
Implements: IGeneratedParameterRegistrar<T>
How It Works
SDCL applies knowledge distillation within the meta-learning loop. A teacher model (EMA of student parameters) provides soft targets that regularize the adapted student. The distillation loss (symmetric KL divergence between teacher and student output distributions) stabilizes adaptation and prevents overfitting to the few support examples.
Algorithm:
Constructors
| Constructor | Summary |
|---|---|
SDCLAlgorithm(SDCLOptions<T, TInput, TOutput> options) |
Properties
| Property | Summary |
|---|---|
MetaLearningAlgorithmType AlgorithmType { get; } |
Methods
| Method | Summary |
|---|---|
IModel<TInput, TOutput, ModelMetadata<T>> Adapt(IMetaLearningTask<T, TInput, TOutput> task) |
|
T MetaTrain(TaskBatch<T, TInput, TOutput> taskBatch) |