ICA based causality inference between variables
Hongxia Chen · 2017
Several approaches have been proposed to discover the causality of the fixed or time-invariant causal model recent years. However, in many practical situations, such as economics and neuroscience, causal relations between variables might be time-dependent. The paper aims to estimate the time-dependent causal model with more generally non-Gaussian noise from purely observational data. It is shown that, under appropriate assumptions, the model can be identified and can be estimated by the proposed independent component analysis based two stage method. Experimental results on artificial data show the effectiveness of the proposed approach.