Robust estimation of Gaussian linear structural equation models with equal error variances

다국어 초록 (Multilingual Abstract)

This study develops a new approach to learning Gaussian linear structural equation models (SEMs) with equal error variances from possibly corrupted observations by outliers.
More precisely, we consider the two types of corrupted Gaussian linear SEMs depending on the outlier type and develop a structure learning algorithm for the models.
The proposed algorithm consists of two steps in which the effect of outliers is eliminated: Step (1) infers the ordering using conditional variances, and Step (2) estimates the presence of edges using conditional independence relationships.
Various numerical experiments verify that the proposed algorithm is empirically consistent even when corrupted samples exist.
It is further confirmed that the proposed algorithm performs better than the state-of-the-art US, GDS, PC, and GES algorithms in noisy data settings.
Through the corrupted real examination marks data, we also demonstrate that the proposed algorithm is well-suited to capturing the interpretable relationships between subjects.

참고문헌 (Reference)

1 Friedman, N., "Using Bayesian networks to analyze expression data" 7 (7): 601-620, 2000

2 Tsamardinos, I., "Towards principled feature selection: Relevancy, flters and wrappers" Morgan Kaufmann Publishers 1 (1): 300-307, 2003

3 Zhang, J., "The three faces of faithfulness" 193 (193): 1011-1027, 2016

4 Kalisch, M., "Robustifcation of the pc-algorithm for directed acyclic graphs" 17 (17): 773-789, 2008

5 Miyamura, M., "Robust gaussian graphical modeling" 97 (97): 1525-1550, 2006

6 Mooij, J., "Regression by dependence minimization and its application to causal inference in additive noise models" ACM 745-752, 2009

7 Pearl, J., "Probabilistic reasoning in intelligent systems: Networks of plausible inference" Elsevier 2014

8 Chickering, D. M., "Optimal structure identifcation with greedy search" 3 : 507-554, 2003

9 Chen, W., "On causal discovery with an equal-variance assumption" 106 (106): 973-980, 2019

10 Hoyer, P. O., "Nonlinear causal discovery with additive noise models" 1 : 689-696, 2009

11 Mardia, K. V., "Multivariate analysis" Academic Press 1979

12 Park, G., "Learning quadratic variance function (qvf) dag models via overdispersion scoring (ods)" 18 (18): 1-44, 2018

13 Ghoshal, A, "Learning linear structural equation models in polynomial time and sample complexity"

14 Park, G., "Learning large-scale Poisson dag models based on overdispersion scoring" 1 : 631-639, 2015

15 Ghoshal, A., "Learning identifable gaussian Bayesian networks in polynomial time and sample complexity" 1 : 6457-6466, 2017

16 Park, G., "Learning high-dimensional gaussian linear structural equation models with heterogeneous error variances" 154 : 107084-, 2021

17 Park, G., "Learning a high-dimensional linear structural equation model via l1-regularized regression" 22 (22): 1-41, 2021

18 박건웅 ; 김영환, "Identifiability of Gaussian linear structural equation models with homogeneous and heterogeneous error variances" 한국통계학회 49 (49): 276-292, 2020

19 Park, G., "Identifability of generalized hypergeometric distribution (ghd) directed acyclic graphical models" 89 : 158-166, 2019

20 Peters, J., "Identifability of gaussian structural equation models with equal error variances" 101 (101): 219-228, 2014

21 Peters, J., "Identifability of causal graphs using functional models"

22 Park, G., "Identifability of additive noise models using conditional variances" 21 (21): 1-34, 2020

23 Loh, P. -L., "High-dimensional learning of linear causal networks via inverse covariance estimation" 15 (15): 3065-3105, 2014

24 Uhler, C., "Geometry of the faithfulness assumption in causal inference" 41 (41): 436-463, 2013

25 Spirtes, P., "Directed cyclic graphical representations of feedback models" Morgan Kaufmann Publishers Inc 1 : 491-498, 1995

26 Spirtes, P., "Causation, prediction, and search" MIT Press 2000

27 Zhang, K., "Causal discovery in the presence of measurement error: Identifability conditions"

28 Doya, K., "Bayesian brain: Probabilistic approaches to neural coding" MIT Press 2007

29 Saeed, B., "Anchored causal inference in the presence of measurement error" PMLR 619-628, 2020

30 Kuipers, J., "Addendum on the scoring of gaussian directed acyclic graphical models" 42 (42): 1689-1691, 2014

31 Shimizu, S., "A linear non-Gaussian acyclic model for causal discovery" 7 : 2003-2030, 2006