eCDANs: Efficient Temporal Causal Discovery from Autocorrelated and Non-stationary Data (Student Abstract)

Conventional temporal causal discovery (CD) methods suffer from high dimensionality, fail to identify lagged relationships, and often ignore dynamics in relations. In this study, we present a novel constraint-based CD approach for autocorrelated non-stationary time series data (eCDANs) capable of detecting contemporaneous relationships along with changes. eCDANs addresses dimensionality by optimizing the conditioning sets while conducting conditional independence (CI) tests identifies changes relations introducing surrogate variable represent dependency. Experiments on synthetic real-world show that can influence outperform baselines.