Causal Discovery for Linear DAGs with Dependent Latent Variables via Higher-order Cumulants
This addresses causal discovery in complex systems with dependent latent variables, which is an incremental improvement over existing methods that assume independent latents.
The paper tackles the problem of estimating causal directed acyclic graphs in linear non-Gaussian models with latent confounders, proposing a novel algorithm that allows causal structures among latent and observed variables and demonstrates validity through simulations and real-world experiments.
This paper addresses the problem of estimating causal directed acyclic graphs in linear non-Gaussian acyclic models with latent confounders (LvLiNGAM). Existing methods assume mutually independent latent confounders or cannot properly handle models with causal relationships among observed variables. We propose a novel algorithm that identifies causal DAGs in LvLiNGAM, allowing causal structures among latent variables, among observed variables, and between the two. The proposed method leverages higher-order cumulants of observed data to identify the causal structure. Extensive simulations and experiments with real-world data demonstrate the validity and practical utility of the proposed algorithm.