AutoIntBase<T>

Base Classes · AiDotNet.NeuralNetworks.Tabular

Base class for AutoInt (Automatic Feature Interaction Learning).

Implements: IParameterSource<T>

Known derived types: AutoIntClassifier<T>, AutoIntRegression<T>

For Beginners

AutoInt discovers which features work well together:

  • Without AutoInt: You manually create features like “age * income”
  • With AutoInt: The model automatically learns “age and income interact”

This is especially useful for recommendation systems, click prediction, and any tabular task where feature combinations matter.

How It Works

AutoInt uses multi-head self-attention to automatically learn feature interactions:

  1. Each feature is embedded into a dense vector
  2. Self-attention layers learn interactions between features
  3. Interactions are combined with original embeddings for prediction

Properties

Property Summary
long ParameterCount { get; }

Methods

Method Summary
Tensor<T> GetInteractionWeights() Gets the learned feature interaction importance.
Vector<T> GetParameters() Reads every parameter in traversal order.
void ResetState() Resets internal state.
void SetParameters(Vector<T> parameters) Restores every parameter, in the order GetParameters emitted them.
void UpdateParameters(T learningRate) Updates all parameters.