Kernel-based Conditional Independence Test and Application in Causal Discovery
Conditional independence testing is an important problem, especially in Bayesian network learning and causal discovery. Due to the curse of dimensionality, testing for conditional independence of continuous variables is particularly challenging. We propose a Kernel-based Conditional Independence test (KCI-test), by constructing an appropriate test statistic and deriving its asymptotic distribution under the null hypothesis of conditional independence. The proposed method is computationally efficient and easy to implement. Experimental results show that it outperforms other methods, especially when the conditioning set is large or the sample size is not very large, in which case other methods encounter difficulties.
| Author(s): | Zhang, K. and Peters, J. and Janzing, D. and Schölkopf, B. |
| Pages: | 804-813 |
| Year: | 2011 |
| Month: | July |
| Day: | 0 |
| Editors: | FG Cozman and A Pfeffer |
| Publisher: | AUAI Press |
| Bibtex Type: | Conference Paper (inproceedings) |
| Address: | Corvallis, OR, USA |
| Event Name: | 27th Conference on Uncertainty in Artificial Intelligence (UAI 2011) |
| Event Place: | Barcelona, Spain |
| Electronic Archiving: | grant_archive |
| ISBN: | 978-0-9749039-7-2 |
| Links: | |
BibTex
@inproceedings{ZhangPJS2011,
title = {Kernel-based Conditional Independence Test and Application in Causal Discovery},
abstract = {Conditional independence testing is an important problem, especially in Bayesian network learning and causal discovery. Due to the curse of dimensionality, testing for conditional independence of continuous variables is particularly challenging. We propose a Kernel-based Conditional Independence test (KCI-test), by constructing an appropriate test statistic and deriving its asymptotic distribution under the null hypothesis of conditional
independence. The proposed method is computationally efficient and easy to implement. Experimental results show that it outperforms other methods, especially when the conditioning set is large or the sample size is not very large, in which case other methods encounter difficulties.},
pages = {804-813},
editors = {FG Cozman and A Pfeffer},
publisher = {AUAI Press},
address = {Corvallis, OR, USA},
month = jul,
year = {2011},
slug = {zhangpjs2011},
author = {Zhang, K. and Peters, J. and Janzing, D. and Sch{\"o}lkopf, B.},
month_numeric = {7}
}