CTGANOptions<T>

Options & Configuration · AiDotNet.Models.Options

Configuration options for CTGAN (Conditional Tabular GAN), a generative adversarial network specifically designed for generating realistic synthetic tabular data.

Inheritance: ModelOptions → NeuralNetworkOptions → ModelHyperparameterOptions → FinancialNeuralNetworkOptions → RiskModelOptions<T> → SyntheticDataGeneratorOptions<T> → CTGANOptions<T>

For Beginners

CTGAN is like having two neural networks compete to create fake data:

  • The Generator creates fake rows of data from random noise
  • The Discriminator tries to distinguish real rows from fake ones
  • As they compete, the generator learns to produce increasingly realistic data

Special features for tabular data:

  • Handles mixed data types (numbers and categories) in the same table
  • Uses conditional generation to ensure rare categories are well-represented
  • Uses Wasserstein loss with gradient penalty for stable training

Example:

How It Works

CTGAN uses a conditional GAN architecture with several innovations for tabular data:

  • Variational Gaussian Mixture (VGM) mode-specific normalization for continuous columns
  • Training-by-sampling with conditional vectors for handling imbalanced categories
  • WGAN-GP (Wasserstein GAN with Gradient Penalty) for stable training
  • PacGAN packing to prevent mode collapse

Reference: “Modeling Tabular Data using Conditional GAN” (Xu et al., NeurIPS 2019)

Constructors

Constructor Summary
CTGANOptions() Initializes a new instance with the default settings.
CTGANOptions(CTGANOptions<T> other) Initializes a new instance by copying every setting of another instance.

Properties

Property Summary
double DiscriminatorDropout { get; set; } Gets or sets the dropout rate for the discriminator.
double GradientPenaltyWeight { get; set; } Gets or sets the gradient penalty weight (lambda) for WGAN-GP.
double GumbelTemperature { get; set; } Gets or sets the Gumbel-softmax temperature of the generator’s mode and discrete-column heads.
double LearningRate { get; set; } Gets or sets the learning rate for both generator and discriminator optimizers.
int BatchSize { get; set; } Gets or sets the training batch size.
int DiscriminatorSteps { get; set; } Gets or sets the number of discriminator training steps per generator step.
int EmbeddingDimension { get; set; } Gets or sets the dimension of the random noise vector fed to the generator.
int Epochs { get; set; } Gets or sets the number of training epochs.
int PacSize { get; set; } Gets or sets the PacGAN packing size.
int VGMModes { get; set; } Gets or sets the number of Gaussian mixture components for VGM normalization.
int[] DiscriminatorDimensions { get; set; } Gets or sets the hidden layer sizes for the discriminator network.
int[] GeneratorDimensions { get; set; } Gets or sets the hidden layer sizes for the generator network.