Causal Discovery and Structure Learning

Discovery algorithms and expert-guided workflows for learning graph structure from data.

Constraint and Hybrid Discovery

PC([variant, ci_test, return_type, ...])	The PC algorithm for causal discovery / structure learning.
MmhcEstimator(data, **kwargs)	Implements the MMHC hybrid structure estimation procedure for learning BayesianNetworks from discrete data.

Score-Based and Tree Search

HillClimbSearch([scoring_method, start_dag, ...])	Score-based causal discovery using hill climbing optimization.
GES([scoring_method, return_type, ...])	Score-based causal discovery using Greedy Equivalence Search (GES).
TreeSearch(data[, root_node, n_jobs])	Search class for learning tree related graph structure.
ExhaustiveSearch(data[, scoring_method, ...])	Search class for exhaustive searches over all DAGs with a given set of variables.

Expert-Guided Discovery

ExpertInLoop([data])
ExpertKnowledge([forbidden_edges, ...])	Class to specify expert knowledge for causal discovery / structure learning algorithms.

See also

Conditional Independence (CI) Tests for conditional independence tests used by constraint-based methods.

Structure Scoring for the scoring classes used by score-based structure search.