ISyntheticTabularGenerator<T>

Interfaces · AiDotNet.Interfaces

Defines the contract for synthetic tabular data generators that learn the distribution of real tabular data and can produce new, realistic synthetic rows.

Known implementations: AutoDiffTabGenerator<T>, CTABGANPlusGenerator<T>, CTGANGenerator<T>, CausalGANGenerator<T>, CopulaGANGenerator<T>, DPCTGANGenerator<T>, FinDiffGenerator<T>, GOGGLEGenerator<T>, MedGANGenerator<T>, MisGANGenerator<T>, OCTGANGenerator<T>, PATEGANGenerator<T>, REaLTabFormerGenerator<T>, SyntheticTabularGeneratorBase<T>, TVAEGenerator<T>, TabDDPMGenerator<T>, TabFlowGenerator<T>, TabLLMGenGenerator<T>, TabSynGenerator<T>, TabTransformerGenGenerator<T>, TableGANGenerator<T>, TimeGANGenerator<T>

For Beginners

Think of this like a “data factory” that learns from real data:

  1. Fit: You show the generator your real table (like a spreadsheet).

It learns the patterns - ranges of numbers, common categories, relationships between columns.

  1. Generate: The generator creates brand-new rows that look realistic

but aren’t copies of any real data.

Common use cases:

  • Data augmentation: Make a small dataset larger for better ML training
  • Privacy: Share synthetic data instead of real data with sensitive information
  • Testing: Generate realistic test data for development
  • Imbalanced data: Generate more examples of rare categories

Example workflow:

How It Works

Synthetic tabular data generators fit a model to real tabular data (containing a mix of continuous/numerical and categorical/discrete columns), then generate new rows that preserve the statistical properties and inter-column relationships of the original data.

Properties

Property Summary
IReadOnlyList<ColumnMetadata> Columns { get; } Gets the column metadata describing the structure of the data this generator was fitted on.
bool IsFitted { get; } Gets a value indicating whether the generator has been fitted to data and is ready to generate.

Methods

Method Summary
Matrix<T> Generate(int numSamples, Vector<T> conditionColumn, Vector<T> conditionValue) Generates new synthetic tabular data rows.
Task FitAsync(Matrix<T> data, IReadOnlyList<ColumnMetadata> columns, int? epochs, CancellationToken ct) Asynchronously fits the generator model to the provided real tabular data.
void Fit(Matrix<T> data, IReadOnlyList<ColumnMetadata> columns, int? epochs) Fits the generator model to the provided real tabular data.