Score-Based vs Constraint-Based Causal Learning in the Presence of Confounders
Abstract
We compare score-based and constraint-based learning in the presence of latent confounders. We use a greedy search strategy to identify the best fitting maximal ancestral graph (MAG) from continuous data, under the assumption of multivariate normality. Scoring maximal ancestral graphs is based on (a) residual iterative conditional fitting for obtaining maximum likelihood estimates for the parameters of a given MAG and (b) factorization and score decomposition results for mixed causal graphs. We compare the score-based approach in simulated settings with two standard constraint-based algorithms: FCI and conservative FCI. Results show a promising performance of the greedy search algorithm.
Cite
Text
Triantafillou and Tsamardinos. "Score-Based vs Constraint-Based Causal Learning in the Presence of Confounders." Conference on Uncertainty in Artificial Intelligence, 2016.Markdown
[Triantafillou and Tsamardinos. "Score-Based vs Constraint-Based Causal Learning in the Presence of Confounders." Conference on Uncertainty in Artificial Intelligence, 2016.](https://mlanthology.org/uai/2016/triantafillou2016uai-score/)BibTeX
@inproceedings{triantafillou2016uai-score,
title = {{Score-Based vs Constraint-Based Causal Learning in the Presence of Confounders}},
author = {Triantafillou, Sofia and Tsamardinos, Ioannis},
booktitle = {Conference on Uncertainty in Artificial Intelligence},
year = {2016},
pages = {59-67},
url = {https://mlanthology.org/uai/2016/triantafillou2016uai-score/}
}