MOCAAlgorithm<T, TInput, TOutput>
Models & Types · AiDotNet.MetaLearning.Algorithms
Implementation of MOCA: Meta-learning with Online Complementary Augmentation.
Inheritance: ModelBase<T, TInput, TOutput> → MetaLearnerBase<T, TInput, TOutput> → MOCAAlgorithm<T, TInput, TOutput>
Implements: IGeneratedParameterRegistrar<T>
How It Works
MOCA augments the meta-learning task distribution by generating complementary tasks in gradient space. For each real task, it creates augmented versions by perturbing the gradient with a direction orthogonal to the original gradient, scaled by historical gradient statistics. This explores complementary adaptation directions that improve generalization and robustness of the meta-learned initialization.
Algorithm:
Constructors
| Constructor | Summary |
|---|---|
MOCAAlgorithm(MOCAOptions<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) |