MetaOptNetModel<T, TInput, TOutput>

Models & Types · AiDotNet.MetaLearning.Models

The linear classifier MetaOptNet’s convex base learner produced for one task.

Implements: IModel<TInput, TOutput, ModelMetadata<T>>

How It Works

Holds its own copy of the embedding network and one weight row per class. Predict returns each example’s score for every class, [rows, NumClasses], scaled by the learned logit scale of eq. 12, for Tensor and Matrix outputs, and the highest-scoring class for a Vector output. It used to flatten the batch into one vector, so a batch of four examples produced one score per class instead of four.

Constructors

Constructor Summary
MetaOptNetModel(IFullModel<T, TInput, TOutput> featureEncoder, Matrix<T> classifierWeights, T temperature, MetaOptNetOptions<T, TInput, TOutput> options) Initializes a new instance of the MetaOptNetModel.

Properties

Property Summary
Matrix<T> ClassifierWeights { get; } Gets the classifier the base learner produced, one row per class.
ModelMetadata<T> Metadata { get; }
T Temperature { get; } Gets the scale that multiplies the logits.
int NumClasses { get; } Gets the number of classes.

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
TOutput Predict(TInput input)
Vector<T> GetParameters()
void Train(TInput inputs, TOutput targets)
void UpdateParameters(Vector<T> parameters)