DPCTGANOptions<T>

Options & Configuration · AiDotNet.Models.Options

Configuration options for DP-CTGAN, a differentially private version of CTGAN that provides formal privacy guarantees while generating synthetic tabular data.

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

For Beginners

DP-CTGAN generates fake data while mathematically guaranteeing that the synthetic data doesn’t reveal too much about any individual in the real data.

Think of it like CTGAN with a “privacy filter”:

  1. During training, gradients are clipped (bounded) so no single person’s data

has too much influence on the model

  1. Random noise is added to further obscure individual contributions
  2. A “privacy budget” (epsilon) tracks how much privacy has been spent

Lower epsilon = more privacy but lower data quality. Typical values: epsilon 1-10 for reasonable utility.

Example:

How It Works

DP-CTGAN adds differential privacy to CTGAN through:

  • Per-sample gradient clipping: Bounds the sensitivity of each training sample
  • Gaussian noise injection: Adds calibrated noise to clipped gradients
  • Privacy accountant: Tracks cumulative privacy loss (epsilon)

Constructors

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

Properties

Property Summary
double ClipNorm { get; set; } Gets or sets the L2 norm clipping bound for per-sample gradients.
double Delta { get; set; } Gets or sets the delta parameter for (epsilon, delta)-differential privacy.
double DiscriminatorDropout { get; set; } Gets or sets the discriminator dropout rate.
double Epsilon { get; set; } Gets or sets the total privacy budget (epsilon) for the training process.
double GradientPenaltyWeight { get; set; } Gets or sets the gradient penalty weight.
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.
double NoiseMultiplier { get; set; } Gets or sets the noise multiplier for Gaussian mechanism.
int BatchSize { get; set; } Gets or sets the training batch size.
int DiscriminatorSteps { get; set; } Gets or sets the number of discriminator steps per generator step.
int EmbeddingDimension { get; set; } Gets or sets the dimension of the random noise vector.
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 VGM modes.
int[] DiscriminatorDimensions { get; set; } Gets or sets the hidden layer sizes for the discriminator.
int[] GeneratorDimensions { get; set; } Gets or sets the hidden layer sizes for the generator.