FunctionalBase<T>

Base Classes · AiDotNet.CausalDiscovery.Functional

Base class for functional/ICA-based causal discovery algorithms (LiNGAM, ANM, etc.).

Inheritance: CausalDiscoveryBase<T> → FunctionalBase<T>

Known derived types: ANMAlgorithm<T>, CAMAlgorithm<T>, CAMUVAlgorithm<T>, CCDrAlgorithm<T>, DirectLiNGAMAlgorithm<T>, ICALiNGAMAlgorithm<T>, IGCIAlgorithm<T>, PNLAlgorithm<T>, RCDAlgorithm<T>, VARLiNGAMAlgorithm<T>

For Beginners

These methods use clever mathematical properties to figure out which variable causes which. For example, if X causes Y with some noise added, the noise pattern looks different depending on which direction you assume the causation goes. These methods detect that asymmetry.

How It Works

Functional methods exploit properties of the data-generating process to determine causal direction. Unlike constraint-based methods (which only learn equivalence classes), functional methods can often identify the unique causal DAG.

Key assumption families:

  • LiNGAM: Linear model with non-Gaussian noise → uses ICA to identify structure
  • ANM: Additive noise model Y = f(X) + N → exploits asymmetry in residuals
  • PNL: Post-nonlinear model Y = g(f(X) + N) → generalizes ANM

Properties

Property Summary
CausalDiscoveryCategory Category { get; }