Discovering Cyclic Causal Models by Independent Components Analysis
Abstract
We generalize Shimizu et al's (2006) ICA-based approach for discovering linear non-Gaussian acyclic (LiNGAM) Structural Equation Models (SEMs) from causally sufficient, continuous-valued observational data. By relaxing the assumption that the generating SEM's graph is acyclic, we solve the more general problem of linear non-Gaussian (LiNG) SEM discovery. LiNG discovery algorithms output the distribution equivalence class of SEMs which, in the large sample limit, represents the population distribution. We apply a LiNG discovery algorithm to simulated data. Finally, we give sufficient conditions under which only one of the SEMs in the output class is 'stable'.
Cite
Text
Lacerda et al. "Discovering Cyclic Causal Models by Independent Components Analysis." Conference on Uncertainty in Artificial Intelligence, 2008.Markdown
[Lacerda et al. "Discovering Cyclic Causal Models by Independent Components Analysis." Conference on Uncertainty in Artificial Intelligence, 2008.](https://mlanthology.org/uai/2008/lacerda2008uai-discovering/)BibTeX
@inproceedings{lacerda2008uai-discovering,
title = {{Discovering Cyclic Causal Models by Independent Components Analysis}},
author = {Lacerda, Gustavo and Spirtes, Peter and Ramsey, Joseph D. and Hoyer, Patrik O.},
booktitle = {Conference on Uncertainty in Artificial Intelligence},
year = {2008},
pages = {366-374},
url = {https://mlanthology.org/uai/2008/lacerda2008uai-discovering/}
}