CAMUVAlgorithm<T>

Models & Types · AiDotNet.CausalDiscovery.Functional

CAM-UV — Causal Additive Model with Unobserved Variables.

Inheritance: CausalDiscoveryBase<T> → FunctionalBase<T> → CAMUVAlgorithm<T>

For Beginners

Sometimes two variables appear related not because one causes the other, but because a hidden third variable causes both. CAM-UV detects these situations by checking: if neither direction X→Y nor Y→X fits cleanly, there might be a hidden common cause. It marks such pairs as “confounded” rather than forcing a causal direction.

How It Works

CAM-UV extends CAM to handle latent (unobserved) confounders. It discovers the causal structure among observed variables even when some common causes are hidden. The algorithm:

  • Fits pairwise additive noise models between all variable pairs.
  • Identifies potential latent confounders by detecting pairs where residuals in

both directions show high dependence (neither direction fits well).

  • Marks bidirectional edges for pairs with suspected latent confounders.
  • Orients remaining edges using the standard ANM asymmetry criterion.

Reference: Maeda and Shimizu (2021), “Causal Additive Models with Unobserved Variables”.

Constructors

Constructor Summary
CAMUVAlgorithm(CausalDiscoveryOptions options)

Properties

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