NODEOptions<T>

Options & Configuration · AiDotNet.Models.Options

Configuration options for NODE (Neural Oblivious Decision Ensembles).

Inheritance: ModelOptions → NeuralNetworkOptions → ModelHyperparameterOptions → FinancialNeuralNetworkOptions → RiskModelOptions<T> → NODEOptions<T>

For Beginners

NODE brings the interpretability of decision trees to deep learning:

  • Oblivious trees: Simpler trees that are faster to evaluate
  • Soft splits: Instead of hard left/right decisions, gradual transitions
  • End-to-end training: Trees are trained with gradient descent like neural networks

This makes NODE both interpretable AND trainable with standard deep learning tools.

Example:

How It Works

NODE combines differentiable oblivious decision trees with neural network training:

  1. Oblivious trees: All nodes at the same depth use the same splitting feature
  2. Soft splits: Differentiable split decisions using entmax for sparse attention
  3. Ensemble: Multiple trees aggregated for the final prediction

Reference: “Neural Oblivious Decision Ensembles for Deep Learning on Tabular Data” (2019)

Constructors

Constructor Summary
NODEOptions()

Properties

Property Summary
IActivationFunction<T> HiddenActivation { get; set; } Gets or sets the hidden layer activation function for the MLP head.
IVectorActivationFunction<T> HiddenVectorActivation { get; set; } Gets or sets the hidden layer vector activation function (alternative to scalar activation).
bool UseBatchNorm { get; set; } Gets or sets whether to use batch normalization on inputs.
bool UseFeaturePreprocessing { get; set; } Gets or sets whether to use feature preprocessing before trees.
double DropoutRate { get; set; } Gets or sets the dropout rate.
double EntmaxAlpha { get; set; } Gets or sets the entmax alpha parameter for sparse attention.
double InitScale { get; set; } Gets or sets the initialization scale for tree parameters.
double Temperature { get; set; } Gets or sets the temperature for soft tree splits.
int NumLeaves { get; } Gets the number of leaf nodes per tree.
int NumTrees { get; set; } Gets or sets the number of trees in the ensemble.
int TreeDepth { get; set; } Gets or sets the depth of each tree.
int TreeOutputDimension { get; set; } Gets or sets the output dimension for each tree.