TimeGANOptions<T>
Options & Configuration · AiDotNet.Models.Options
Configuration options for TimeGAN, a generative adversarial network designed specifically for generating realistic time-series data while preserving temporal dynamics.
Inheritance: ModelOptions → NeuralNetworkOptions → ModelHyperparameterOptions → FinancialNeuralNetworkOptions → RiskModelOptions<T> → SyntheticDataGeneratorOptions<T> → TimeGANOptions<T>
For Beginners
TimeGAN generates fake time-series data that looks realistic.
Regular GANs struggle with sequential data because they don’t understand “time.” TimeGAN solves this with a clever multi-step training approach:
- Embedding training: Learn to compress real data into a simpler form
- Supervised training: Learn the rules of how data changes over time
- Joint training: Train everything together — generator, discriminator,
embedder, recovery, and supervisor — all at once
Example:
How It Works
TimeGAN combines autoencoding and adversarial training in a shared latent space. It uses five components:
- Embedder: Maps real data to a latent embedding space
- Recovery: Reconstructs data from the latent space
- Generator: Produces synthetic latent embeddings
- Supervisor: Learns temporal dynamics in the latent space
- Discriminator: Distinguishes real from synthetic embeddings
Reference: “Time-series Generative Adversarial Networks” (Yoon et al., NeurIPS 2019)
Constructors
| Constructor | Summary |
|---|---|
TimeGANOptions() |
Properties
| Property | Summary |
|---|---|
double DiscriminatorDropout { get; set; } |
Gets or sets the dropout rate for discriminator hidden layers. |
double LearningRate { get; set; } |
Gets or sets the learning rate. |
double ReconstructionWeight { get; set; } |
Gets or sets the weight for the reconstruction loss. |
double SupervisedWeight { get; set; } |
Gets or sets the weight for the supervised loss in the generator objective. |
int BatchSize { get; set; } |
Gets or sets the training batch size. |
int Epochs { get; set; } |
Gets or sets the total number of training epochs. |
int HiddenDimension { get; set; } |
Gets or sets the hidden dimension for the RNN components. |
int NumLayers { get; set; } |
Gets or sets the number of RNN layers in each component. |
int SequenceLength { get; set; } |
Gets or sets the length of each time-series sequence. |