OpenMAMLPlusAlgorithm<T, TInput, TOutput>
Models & Types · AiDotNet.MetaLearning.Algorithms
Implementation of Open-MAML++: MAML with per-parameter learning rates and open-set novelty detection.
Inheritance: ModelBase<T, TInput, TOutput> → MetaLearnerBase<T, TInput, TOutput> → OpenMAMLPlusAlgorithm<T, TInput, TOutput>
Implements: IGeneratedParameterRegistrar<T>
How It Works
Open-MAML++ combines MAML++ improvements (per-parameter learning rates, multi-step loss, learned layer-specific LR) with open-set recognition. The algorithm meta-learns a novelty threshold based on prediction entropy: samples with entropy exceeding the threshold are classified as novel/unknown. The multi-step loss ensures stable intermediate adaptation steps.
Algorithm:
Constructors
| Constructor | Summary |
|---|---|
OpenMAMLPlusAlgorithm(OpenMAMLPlusOptions<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) |