TabTransformer<T>

Models & Types · AiDotNet.Finance.Risk

TabTransformer model for tabular data, using transformers for categorical features.

Inheritance: NeuralNetworkBase<T> → FinancialModelBase<T> → RiskModelBase<T> → TabTransformer<T>

For Beginners

TabTransformer applies the power of Transformers (like BERT/GPT) specifically to the categorical parts of your data (e.g., “Sector”, “Country”, “Rating”). It learns deep contextual embeddings for these categories before combining them with numerical data.

Example

// Define architecture for categorical-rich tabular risk data (10 features, single output)
var architecture = new NeuralNetworkArchitecture<double>(
    inputType: InputType.OneDimensional,
    taskType: NeuralNetworkTaskType.Regression,
    inputSize: 10, outputSize: 1);
            
// Training mode: transformer contextualizes categorical embeddings before MLP
var model = new TabTransformer<double>(architecture);
            
// Parameterless constructor with default architecture
var defaultModel = new TabTransformer<double>();

Constructors

Constructor Summary
TabTransformer() Initializes a new instance of the TabTransformer model.
TabTransformer(NeuralNetworkArchitecture<T> architecture, TabTransformerOptions<T> options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer, ILossFunction<T> lossFunction)
TabTransformer(NeuralNetworkArchitecture<T> architecture, string onnxModelPath, TabTransformerOptions<T> options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer, ILossFunction<T> lossFunction) Initializes a new instance from ONNX.

Methods

Method Summary
T CalculateRisk(Tensor<T> input) Calculates risk using TabTransformer.
Tensor<T> AdjustForRisk(Tensor<T> action, T riskBudget) Adjusts action for risk.