MetaDMAlgorithm<T, TInput, TOutput>
Models & Types · AiDotNet.MetaLearning.Algorithms
Implementation of Meta-DM: Applications of Diffusion Models on Few-Shot Learning (Hu et al., ICIP 2024).
Inheritance: ModelBase<T, TInput, TOutput> → MetaLearnerBase<T, TInput, TOutput> → MetaDMAlgorithm<T, TInput, TOutput>
Implements: IGeneratedParameterRegistrar<T>
How It Works
Meta-DM uses a DDPM-style diffusion model as a data augmentation module for few-shot learning. Rather than replacing the meta-learning algorithm, it augments the support set by generating synthetic samples conditioned on existing few-shot data, then performs standard gradient-based adaptation on the enriched dataset.
Algorithm:
Key advantage: Modular — can be composed with any gradient-based meta-learner. The diffusion-based augmentation enriches the support set, reducing overfitting.
Constructors
| Constructor | Summary |
|---|---|
MetaDMAlgorithm(MetaDMOptions<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) |