TransferEntropyAlgorithm<T>

Models & Types · AiDotNet.CausalDiscovery.InformationTheoretic

Transfer Entropy — information-theoretic measure of directed information flow.

Inheritance: CausalDiscoveryBase<T> → InfoTheoreticBase<T> → TransferEntropyAlgorithm<T>

For Beginners

Transfer Entropy is like Granger causality but works for nonlinear relationships too. It asks: “Does knowing X’s past reduce my uncertainty about Y’s future, beyond what Y’s own past already tells me?” If yes, X transfers information to Y.

How It Works

Transfer entropy quantifies the amount of directed information transfer from one process to another. It measures the reduction in uncertainty of Y’s future given the past of both X and Y, compared to only Y’s past. It is a nonlinear generalization of Granger causality.

Algorithm:

  • For each pair (X→Y), construct lagged variables: Y_past (lag 1..L of Y), X_past (lag 1..L of X)
  • Compute TE(X→Y) = H(Y_future | Y_past) - H(Y_future | Y_past, X_past)
  • Using Gaussian approximation: TE = 0.5 * log(var(Y_future|Y_past) / var(Y_future|Y_past,X_past))
  • Apply score threshold: only keep edges where TE exceeds threshold
  • Direction is inherent: TE(X→Y) ≠ TE(Y→X) in general

Reference: Schreiber (2000), “Measuring Information Transfer”, Physical Review Letters.

Constructors

Constructor Summary
TransferEntropyAlgorithm(CausalDiscoveryOptions options)

Properties

Property Summary
bool SupportsNonlinear { get; }
bool SupportsTimeSeries { get; }
string Name { get; }