CSlearn

A CStree on four binary random variables.

cslearn is a Python package for CStree models [1] [2] – a family of graphical causal models for multivariate discrete data that encode context-specific independence (CSI). CStrees generalize DAG models while remaining tractable.

The package implements CSlearn, a three-phase structure-learning algorithm: DAG pre-screening (PC or GRaSP) to restrict candidate parent sets, order MCMC over topological orderings, and an exact staging search under a sparsity bound.

See Installation to get started and the project README for a quick-start example. The paper experiments and precomputed results (~5.9 GB) are on Zenodo.

References

[1]

Eliana Duarte and Liam Solus. Representation of context-specific causal models with observational and interventional data. Journal of the Royal Statistical Society Series B: Statistical Methodology, 88(2):567–610, April 2026. URL: https://doi.org/10.1093/jrsssb/qkaf059, doi:10.1093/jrsssb/qkaf059.

[2]

Felix L. Rios, Alex Markham, and Liam Solus. Scalable structure learning for sparse context-specific causal systems. 2024. arXiv:2402.07762.