MisGANOptions<T>
Options & Configuration · AiDotNet.Models.Options
Configuration options for MisGAN, a GAN for learning from incomplete data with dual generator/discriminator pairs for data and missingness patterns.
Inheritance: ModelOptions → NeuralNetworkOptions → ModelHyperparameterOptions → FinancialNeuralNetworkOptions → RiskModelOptions<T> → SyntheticDataGeneratorOptions<T> → MisGANOptions<T>
For Beginners
MisGAN handles datasets with missing values:
Real datasets often have missing values (e.g., patients who skip tests). MisGAN learns both:
- What the data looks like when complete
- Which values tend to be missing (and why)
This produces synthetic data with realistic patterns of completeness.
Example:
How It Works
MisGAN uses four networks:
- Data generator: Generates complete data samples
- Mask generator: Generates realistic missingness patterns
- Data discriminator: Judges realism of data values
- Mask discriminator: Judges realism of missingness patterns
Reference: “MisGAN: Learning from Incomplete Data with GANs” (Li et al., ICLR 2019)
Constructors
| Constructor | Summary |
|---|---|
MisGANOptions() |
Properties
| Property | Summary |
|---|---|
double DiscriminatorDropout { get; set; } |
Gets or sets the dropout rate for discriminator hidden layers. |
double GradientPenaltyWeight { get; set; } |
Gets or sets the weight for the WGAN-GP gradient penalty term. |
double LearningRate { get; set; } |
Gets or sets the learning rate. |
double MissingRate { get; set; } |
Gets or sets the expected rate of missing values in the data. |
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 noise dimension. |
int Epochs { get; set; } |
Gets or sets the number of training epochs. |
int VGMModes { get; set; } |
Gets or sets the number of VGM modes for continuous column encoding. |
int[] HiddenDimensions { get; set; } |
Gets or sets the hidden layer sizes. |