SignatureInformedTransformerOptions<T>
Options & Configuration · AiDotNet.Models.Options
Configuration for SignatureInformedTransformer.
Inheritance: ModelOptions → NeuralNetworkOptions → SignatureInformedTransformerOptions<T>
For Beginners
This model reads a window of price history for a set of assets and outputs how much of the portfolio to put in each. It is trained to keep bad outcomes small rather than to forecast prices accurately.
How It Works
Reference: Yoontae Hwang and Stefan Zohren, “Signature-Informed Transformer for Asset Allocation”, 2025 (arXiv:2510.03129).
Defaults follow that paper and its reference implementation: lookback 60, horizon 20, signature level 2, 8 heads, 1 layer, feed-forward 64, temperature 1.3, and Adam at 1e-3 with batch 64 and dropout 0.1.
ModelDimension is the one default NOT taken directly from the reference. The paper selects
d_model per dataset by validation over {8, 16, 32, 64}, so there is no single paper value; the
reference README’s example of 8 would leave one dimension per head against the default 8 heads.
This class defaults to 64 — see the property for the full reasoning. Stating it here too, because
a class summary that lists a different default than the property is the version a reader trusts.
Constructors
| Constructor | Summary |
|---|---|
SignatureInformedTransformerOptions() |
Initializes a new instance with default values. |
SignatureInformedTransformerOptions(SignatureInformedTransformerOptions<T> other) |
Initializes a new instance by copying every property from another instance. |
Properties
| Property | Summary |
|---|---|
double CVaRAlpha { get; set; } |
Gets or sets the CVaR confidence level alpha. |
double DropoutRate { get; set; } |
Gets or sets the dropout rate. |
double LearningRate { get; set; } |
Gets or sets the learning rate. |
double Temperature { get; set; } |
Gets or sets tau, the softmax temperature controlling allocation concentration. |
double TransactionCostBasisPoints { get; set; } |
Gets or sets the round-trip transaction cost in basis points used when EVALUATING a strategy. |
int BatchSize { get; set; } |
Gets or sets the batch size. |
int EarlyStoppingPatience { get; set; } |
Gets or sets the early-stopping patience, in epochs. |
int FeedForwardDimension { get; set; } |
Gets or sets the feed-forward width. |
int Horizon { get; set; } |
Gets or sets the forecast/holding horizon K, in time steps. |
int LookbackWindow { get; set; } |
Gets or sets the lookback window H, in time steps. |
int MaxEpochs { get; set; } |
Gets or sets the maximum training epochs. |
int ModelDimension { get; set; } |
Gets or sets the model width d_model. |
int NumAssets { get; set; } |
Gets or sets the number of assets in the investment universe. |
int NumHeads { get; set; } |
Gets or sets the attention head count. |
int NumLayers { get; set; } |
Gets or sets the encoder depth. |
int RelationalHiddenDimension { get; set; } |
Gets or sets the hidden width of the MLPs producing the dynamic queries and relational embeddings. |
int SignatureLevel { get; set; } |
Gets or sets the signature truncation level M. |
Methods
| Method | Summary |
|---|---|
void Validate() |
Validates the configuration. |