MMHCAlgorithm<T>

Models & Types · AiDotNet.CausalDiscovery.Hybrid

MMHC — Max-Min Hill-Climbing, a hybrid constraint-based + score-based algorithm.

Inheritance: CausalDiscoveryBase<T> → HybridBase<T> → MMHCAlgorithm<T>

For Beginners

MMHC first quickly identifies which variables MIGHT be related (using statistical tests), then carefully determines the exact direction and strength of those relationships (using a scoring approach). This makes it both fast and accurate.

How It Works

MMHC combines two phases:

  • MMPC phase (constraint-based): For each variable, identify candidate parents/children

using the Max-Min Parents and Children heuristic. This restricts the search space.

  • HC phase (score-based): Run greedy Hill Climbing search within the restricted

space defined by the MMPC skeleton, optimizing BIC score.

Reference: Tsamardinos et al. (2006), “The Max-Min Hill-Climbing Bayesian Network Structure Learning Algorithm”, Machine Learning.

Constructors

Constructor Summary
MMHCAlgorithm(CausalDiscoveryOptions options)

Properties

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