Breaking Bad: Component-Wise Parent Deletion for Score-Based Causal Discovery

Min Woo Park, Taehui Yun, YoungIn Jang, Yoonseok Yeom, Jonghwan Kim, Jiyeon Kang, Songseong Kim, Hyemin Jung, Sangmin Lee, Jongseong Jang, Sanghack Lee
Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, PMLR 337:5289-5331, 2026.

Abstract

Directed acyclic graphs (DAGs) are widely used to represent complex causal relationships in real-world systems. The goal of causal discovery is to learn the underlying {DAG} from data generated by these systems. While *Greedy Equivalence Search* (GES) is a well-established score-based algorithm for causal discovery, the GES family often suffers from scalability and sample complexity issues due to its large search space and susceptibility to local optima. In this paper, we introduce *parent deletion*, a novel, simple, yet powerful operator for score-based causal discovery. This operator is theoretically sound and effectively alleviates these limitations. Moreover, its simplicity enables seamless compatibility with existing score-based methods, and extensive experiments demonstrate consistent improvements across a wide range of settings. The source code used in this study is available at https://github.com/LGAI-Research/breaking-bad.

Cite this Paper

BibTeX
@InProceedings{pmlr-v337-park26b, title = {Breaking Bad: Component-Wise Parent Deletion for Score-Based Causal Discovery}, author = {Park, Min Woo and Yun, Taehui and Jang, YoungIn and Yeom, Yoonseok and Kim, Jonghwan and Kang, Jiyeon and Kim, Songseong and Jung, Hyemin and Lee, Sangmin and Jang, Jongseong and Lee, Sanghack}, booktitle = {Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence}, pages = {5289--5331}, year = {2026}, editor = {Perković, Emilija and Malinsky, Daniel}, volume = {337}, series = {Proceedings of Machine Learning Research}, month = {17--21 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v337/main/assets/park26b/park26b.pdf}, url = {https://proceedings.mlr.press/v337/park26b.html}, abstract = {Directed acyclic graphs (DAGs) are widely used to represent complex causal relationships in real-world systems. The goal of causal discovery is to learn the underlying {DAG} from data generated by these systems. While *Greedy Equivalence Search* (GES) is a well-established score-based algorithm for causal discovery, the GES family often suffers from scalability and sample complexity issues due to its large search space and susceptibility to local optima. In this paper, we introduce *parent deletion*, a novel, simple, yet powerful operator for score-based causal discovery. This operator is theoretically sound and effectively alleviates these limitations. Moreover, its simplicity enables seamless compatibility with existing score-based methods, and extensive experiments demonstrate consistent improvements across a wide range of settings. The source code used in this study is available at https://github.com/LGAI-Research/breaking-bad.} }
Endnote
%0 Conference Paper %T Breaking Bad: Component-Wise Parent Deletion for Score-Based Causal Discovery %A Min Woo Park %A Taehui Yun %A YoungIn Jang %A Yoonseok Yeom %A Jonghwan Kim %A Jiyeon Kang %A Songseong Kim %A Hyemin Jung %A Sangmin Lee %A Jongseong Jang %A Sanghack Lee %B Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2026 %E Emilija Perković %E Daniel Malinsky %F pmlr-v337-park26b %I PMLR %P 5289--5331 %U https://proceedings.mlr.press/v337/park26b.html %V 337 %X Directed acyclic graphs (DAGs) are widely used to represent complex causal relationships in real-world systems. The goal of causal discovery is to learn the underlying {DAG} from data generated by these systems. While *Greedy Equivalence Search* (GES) is a well-established score-based algorithm for causal discovery, the GES family often suffers from scalability and sample complexity issues due to its large search space and susceptibility to local optima. In this paper, we introduce *parent deletion*, a novel, simple, yet powerful operator for score-based causal discovery. This operator is theoretically sound and effectively alleviates these limitations. Moreover, its simplicity enables seamless compatibility with existing score-based methods, and extensive experiments demonstrate consistent improvements across a wide range of settings. The source code used in this study is available at https://github.com/LGAI-Research/breaking-bad.
APA
Park, M.W., Yun, T., Jang, Y., Yeom, Y., Kim, J., Kang, J., Kim, S., Jung, H., Lee, S., Jang, J. & Lee, S.. (2026). Breaking Bad: Component-Wise Parent Deletion for Score-Based Causal Discovery. Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research 337:5289-5331 Available from https://proceedings.mlr.press/v337/park26b.html.

Related Material