BMAMLAlgorithm<T, TInput, TOutput>
Models & Types · AiDotNet.MetaLearning.Algorithms
Implementation of BMAML: Bayesian Model-Agnostic Meta-Learning (Yoon et al., NeurIPS 2018).
Inheritance: ModelBase<T, TInput, TOutput> → MetaLearnerBase<T, TInput, TOutput> → BMAMLAlgorithm<T, TInput, TOutput>
How It Works
BMAML uses Stein Variational Gradient Descent (SVGD) to maintain a particle ensemble {θ¹, …, θᴹ} that approximates the posterior distribution over task-adapted parameters. Each particle is updated using both the task loss gradient and a repulsive kernel term that encourages diversity among particles.
Algorithm:
Constructors
| Constructor | Summary |
|---|---|
BMAMLAlgorithm(BMAMLOptions<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) |