Abstract:Learning causal effects from data is a fundamental and well-studied problem across science, especially when the cause-effect relationship is static in nature. However, causal effect is less explored when there are dynamical dependencies, i.e., when dependencies exist between entities across time. Identifying dynamic causal effects from time-series observations is computationally expensive when compared to the static scenario. We demonstrate that the computational complexity of recovering the causation structure for the vector auto-regressive (VAR) model is $O(Tn^3N^2)$, where $n$ is the number of nodes, $T$ is the number of samples, and $N$ is the largest time-lag in the dependency between entities. We report a method, with a reduced complexity of $O(Tn^3 \log N)$, to recover the causation structure to obtain frequency-domain (FD) representations of time-series. Since FFT accumulates all the time dependencies on every frequency, causal inference can be performed efficiently by considering the state variables as random variables at any given frequency. We additionally show that, for systems with interactions that are LTI, do-calculus machinery can be realized in the FD resulting in versions of the classical single-door (with cycles), front and backdoor criteria. We demonstrate, for a large class of problems, graph reconstruction using multivariate Wiener projections results in a significant computational advantage with $O(n)$ complexity over reconstruction algorithms such as the PC algorithm which has $O(n^q)$ complexity, where $q$ is the maximum neighborhood size. This advantage accrues due to some remarkable properties of the phase response of the frequency-dependent Wiener coefficients which is not present in any time-domain approach.

Causal Discovery of Stochastic Dynamical Systems: A Markov Chain Approach

Causal Discovery and Forecasting in Nonstationary Environments with State-Space Models

Identifying Nonstationary Causal Structures with High-Order Markov Switching Models

Causal Discovery from Temporally Aggregated Time Series

Data-Driven Performance Monitoring of Dynamical Systems Using Granger Causal Graphical Models

Causal Discovery in Financial Markets: A Framework for Nonstationary Time-Series Data

Amortized Causal Discovery: Learning to Infer Causal Graphs from Time-Series Data

Bayesian inference of causal relations between dynamical systems

Learning dynamic causal mechanisms from non-stationary data

Unified Causality Analysis Based on the Degrees of Freedom

Dynamic Structural Causal Models

Exploring Granger Causality in Dynamical Systems Modeling and Performance Monitoring.

Neural Additive Vector Autoregression Models for Causal Discovery in Time Series

Causal Discovery in Semi-Stationary Time Series

Causal Discovery from Heterogeneous/Nonstationary Data with Independent Changes

Causal Discovery from Conditionally Stationary Time Series

Confidence in Causal Discovery with Linear Causal Models

Complete Inference of Causal Relations between Dynamical Systems

Causal Structure Recovery of Linear Dynamical Systems: An FFT based Approach

Causal Discovery-Driven Change Point Detection in Time Series

Causal Discovery via Causal Star Graphs