GESAlgorithm<T>

Models & Types · AiDotNet.CausalDiscovery.ScoreBased

GES (Greedy Equivalence Search) — score-based causal discovery over equivalence classes.

Inheritance: CausalDiscoveryBase<T> → ScoreBasedBase<T> → GESAlgorithm<T>

For Beginners

GES builds a causal graph by first adding edges that improve the model fit, then removing edges that are unnecessary. It uses a score (BIC) that rewards fitting the data well while penalizing too many edges.

How It Works

GES searches over Markov equivalence classes of DAGs using two phases:

  • Forward phase: Greedily adds edges that most improve the BIC score.
  • Backward phase: Greedily removes edges that improve the BIC score.

Reference: Chickering (2002), “Optimal Structure Identification with Greedy Search”, JMLR.

Constructors

Constructor Summary
GESAlgorithm(CausalDiscoveryOptions options)

Properties

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