TabDPTBase<T>

Base Classes · AiDotNet.NeuralNetworks.Tabular

Base class for TabDPT (Tabular Data Pre-Training) foundation model.

Implements: IParameterSource<T>

Known derived types: TabDPTClassifier<T>, TabDPTRegression<T>

For Beginners

TabDPT is like a “GPT for tables”:

  • Pre-training: Model learns patterns from many different tabular datasets
  • Transfer learning: These learned patterns help on new, unseen data
  • In-context learning: Given a few examples, it adapts to new tasks
  • Feature-wise attention: Understands relationships between columns

The model processes features as tokens and uses transformer architecture to capture complex interactions, similar to how language models process words.

How It Works

TabDPT applies foundation model concepts to tabular data, using pre-training on diverse datasets to learn transferable representations that can adapt to new tasks through in-context learning.

Reference: “TabDPT: Scaling Tabular Foundation Models” (2025)

Properties

Property Summary
int NumNumericalFeatures { get; } Gets the number of numerical features.
long ParameterCount { get; }

Methods

Method Summary
Vector<T> GetParameters() Reads every parameter in traversal order.
void ResetState() Resets internal state and caches.
void SetParameters(Vector<T> parameters) Restores every parameter, in the order GetParameters emitted them.
void UpdateParameters(T learningRate) Updates all trainable parameters.