causal-ts-237a0a68·1 events·first seen Aliases: Causal-TS
Bloomberg has open-sourced Causal-TS, a Python library for causal discovery in high-dimensional and nonstationary multivariate time series. The library implements four specialized algorithms (CDNOTS, CDNOTS+, CEDAR, GRACE) plus wrappers for established methods, with a unified conditional independence test layer GPU-accelerated via PyTorch. It includes a regime discovery pipeline for structural breaks, synthetic data generators, CLI tooling, and optional DoWhy integration for end-to-end causal effect estimation.