On Estimation of Functional Causal Models: Post-Nonlinear Causal Model as an Example

Abstract

Compared to constraint-based causal discovery, causal discovery based on functional causal models is able to identify the whole causal model under appropriate assumptions. Functional causal models represent the effect as a function of the direct causes together with an independent noise term. Examples include the linear non-Gaussian a cyclic model (LiNGAM), nonlinear additive noise model, and post-nonlinear (PNL) model. Currently there are two ways to estimate the parameters in the models, one is by dependence minimization, and the other is maximum likelihood. In this paper, we show that for any a cyclic functional causal model, minimizing the mutual information between the hypothetical cause and the noise term is equivalent to maximizing the data likelihood with a flexible model for the distribution of the noise term. We then focus on estimation of the PNL causal model, and propose to estimate it with the warped Gaussian process with the noise modeled by the mixture of Gaussians. As a Bayesian nonparametric approach, it outperforms the previous one based on mutual information minimization with nonlinear functions represented by multilayer perceptrons, we also show that unlike the ordinary regression, estimation results of the PNL causal model are sensitive to the assumption on the noise distribution. Experimental results on both synthetic and real data support our theoretical claims.