DynamicTaskSamplingAlgorithm<T, TInput, TOutput>
Models & Types · AiDotNet.MetaLearning.Algorithms
Implementation of Dynamic Task Sampling: difficulty-aware gradient reweighting for meta-learning.
Inheritance: ModelBase<T, TInput, TOutput> → MetaLearnerBase<T, TInput, TOutput> → DynamicTaskSamplingAlgorithm<T, TInput, TOutput>
How It Works
Dynamic Task Sampling maintains running difficulty estimates for tasks in the meta-batch and uses difficulty-proportional weighting on meta-gradients. Tasks with higher post-adaptation query loss (= harder tasks) receive higher gradient weights, focusing meta-learning on areas where the model struggles most. An exploration bonus (UCB-style) ensures that tasks seen less frequently still receive gradient signal.
Algorithm:
Constructors
| Constructor | Summary |
|---|---|
DynamicTaskSamplingAlgorithm(DynamicTaskSamplingOptions<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) |