FGES Algorithm

Overview

Fast Greedy Equivalence Search (FGES) is a score-based algorithm that searches over Markov equivalence classes rather than individual DAGs. It uses greedy edge additions and removals to optimise a chosen score (such as BIC), which makes it practical for larger variable sets where exhaustive search is infeasible.

In causaliq-discovery, FGES is currently provided via the Tetrad causal-cmd backend and returns a PDAG representing an equivalence class of candidate causal structures. FGES is often a strong baseline for high-dimensional settings, but like other greedy methods it can still depend on score choice and data quality.

Class: score · PDAG Package: Tetrad/causal-cmd

Reference

Ramsey J. et al. (2017) – A million variables and more: the Fast Greedy Equivalence Search algorithm for learning high-dimensional graphical causal models. Int. J. Data Sci. Anal. 3, 121–129. https://doi.org/10.1007/s41060-016-0032-z

Hyperparameters

Hyperparameter	Type	Default	Values	Description
iss	float	1.0	—	Imaginary Sample Size weighting the prior in Bayesian scores.
max_elapsed	int	No limit	—	Maximum allowed execution time in seconds.
penalty_weight	float	1.0	—	Weight of the penalty component in AIC and BIC scores.
score	str	bic	aic, bdeu, bge, bic, k2, loglik	Scoring function for score-based learning.

Variants

Variant	Package
tetrad	Tetrad/causal-cmd