Causal-TS: A Python Library for Causal Discovery in High-Dimensional and Nonstationary Time Series
Causal-TS packages regime-aware causal discovery for multivariate time series into a pip-installable Python library.
The paper says the library includes four specialized algorithms: CDNOTS, CDNOTS+, CEDAR, and GRACE. It also wraps GES, Granger, LASSO-VAR, and LGES behind a shared conditional-independence test layer with PyTorch GPU acceleration. Its regime pipeline detects structural breaks, then runs discovery separately per regime with regime-specific parameters. The package also includes a CLI, synthetic data generators, and optional DoWhy integration for moving from raw time series to causal effect estimates. ArXiv · AI/CL/LG's note
The paper says the library includes four specialized algorithms: CDNOTS, CDNOTS+, CEDAR, and GRACE. It also wraps GES, Granger, LASSO-VAR, and LGES behind a shared conditional-independence test layer with PyTorch GPU acceleration. Its regime pipeline detects structural breaks, then runs discovery separately per regime with regime-specific parameters. The package also includes a CLI, synthetic data generators, and optional DoWhy integration for moving from raw time series to causal effect estimates. ArXiv · AI/CL/LG's note
score 5