Typing assumptions improve identification in causal discovery

PHILIPPE BROUILLARD, Perouz Taslakian, Alexandre Lacoste, Sebastien Lachapelle, Alexandre Drouin
Proceedings of the First Conference on Causal Learning and Reasoning, PMLR 177:162-177, 2022.

Abstract

Causal discovery from observational data is a challenging task that can only be solved up to a set of equivalent solutions, called an equivalence class. Such classes, which are often large in size, encode uncertainties about the orientation of some edges in the causal graph. In this work, we propose a new set of assumptions that constrain possible causal relationships based on the nature of variables, thus circumscribing the equivalence class. Namely, we introduce typed directed acyclic graphs, in which variable types are used to determine the validity of causal relationships. We demonstrate, both theoretically and empirically, that the proposed assumptions can result in significant gains in the identification of the causal graph. We also propose causal discovery algorithms that make use of these assumptions and demonstrate their benefits on simulated and pseudo-real data.

Cite this Paper

BibTeX
@InProceedings{pmlr-v177-brouillard22a, title = {Typing assumptions improve identification in causal discovery}, author = {BROUILLARD, PHILIPPE and Taslakian, Perouz and Lacoste, Alexandre and Lachapelle, Sebastien and Drouin, Alexandre}, booktitle = {Proceedings of the First Conference on Causal Learning and Reasoning}, pages = {162--177}, year = {2022}, editor = {Schölkopf, Bernhard and Uhler, Caroline and Zhang, Kun}, volume = {177}, series = {Proceedings of Machine Learning Research}, month = {11--13 Apr}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v177/brouillard22a/brouillard22a.pdf}, url = {https://proceedings.mlr.press/v177/brouillard22a.html}, abstract = {Causal discovery from observational data is a challenging task that can only be solved up to a set of equivalent solutions, called an equivalence class. Such classes, which are often large in size, encode uncertainties about the orientation of some edges in the causal graph. In this work, we propose a new set of assumptions that constrain possible causal relationships based on the nature of variables, thus circumscribing the equivalence class. Namely, we introduce typed directed acyclic graphs, in which variable types are used to determine the validity of causal relationships. We demonstrate, both theoretically and empirically, that the proposed assumptions can result in significant gains in the identification of the causal graph. We also propose causal discovery algorithms that make use of these assumptions and demonstrate their benefits on simulated and pseudo-real data.} }
Endnote
%0 Conference Paper %T Typing assumptions improve identification in causal discovery %A PHILIPPE BROUILLARD %A Perouz Taslakian %A Alexandre Lacoste %A Sebastien Lachapelle %A Alexandre Drouin %B Proceedings of the First Conference on Causal Learning and Reasoning %C Proceedings of Machine Learning Research %D 2022 %E Bernhard Schölkopf %E Caroline Uhler %E Kun Zhang %F pmlr-v177-brouillard22a %I PMLR %P 162--177 %U https://proceedings.mlr.press/v177/brouillard22a.html %V 177 %X Causal discovery from observational data is a challenging task that can only be solved up to a set of equivalent solutions, called an equivalence class. Such classes, which are often large in size, encode uncertainties about the orientation of some edges in the causal graph. In this work, we propose a new set of assumptions that constrain possible causal relationships based on the nature of variables, thus circumscribing the equivalence class. Namely, we introduce typed directed acyclic graphs, in which variable types are used to determine the validity of causal relationships. We demonstrate, both theoretically and empirically, that the proposed assumptions can result in significant gains in the identification of the causal graph. We also propose causal discovery algorithms that make use of these assumptions and demonstrate their benefits on simulated and pseudo-real data.
APA
BROUILLARD, P., Taslakian, P., Lacoste, A., Lachapelle, S. & Drouin, A.. (2022). Typing assumptions improve identification in causal discovery. Proceedings of the First Conference on Causal Learning and Reasoning, in Proceedings of Machine Learning Research 177:162-177 Available from https://proceedings.mlr.press/v177/brouillard22a.html.

Related Material