ACLAlgorithm<T, TInput, TOutput>
Models & Types · AiDotNet.MetaLearning.Algorithms
Implementation of ACL: Adaptive Continual Learning with task-specific parameter importance masks.
Inheritance: ModelBase<T, TInput, TOutput> → MetaLearnerBase<T, TInput, TOutput> → ACLAlgorithm<T, TInput, TOutput>
Implements: IGeneratedParameterRegistrar<T>
How It Works
ACL prevents catastrophic forgetting by maintaining per-parameter importance scores that accumulate across tasks via exponential moving average. Important parameters are protected by reducing their effective learning rate and applying elastic weight consolidation (EWC)-style regularization toward the pre-task initialization.
Algorithm:
Constructors
| Constructor | Summary |
|---|---|
ACLAlgorithm(ACLOptions<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) |