Causal-learn (documentation, paper) is an open-source platform for causal learning with both classical and state-of-the-art causal discovery algorithms. It aims to recover causal structure from observational data, without requiring interventional experiments, while offering provable correctness guarantees.
The package is actively being developed. Feedback (issues, suggestions, etc.) would be greatly appreciated.
Our causal-learn implements methods for causal discovery:
- Constraint-based causal discovery methods.
- Score-based causal discovery methods.
- Causal discovery methods based on constrained functional causal models.
- Hidden causal representation learning.
- Permutation-based causal discovery methods.
- Granger causality.
- Multiple utilities for building your own method, such as independence tests, score functions, graph operations, and evaluations.
Causal-learn needs the following packages to be installed beforehand:
- python 3 (>=3.7)
- numpy
- networkx
- pandas
- scipy
- scikit-learn
- statsmodels
- pydot
(For visualization)
- matplotlib
- graphviz
To use causal-learn, we could install it using pip:
pip install causal-learn
Please kindly refer to causal-learn Doc for detailed tutorials and usages.
For search methods in causal discovery, there are various running examples in the ‘tests’ directory, such as TestPC.py and TestGES.py.
For the implemented modules, such as (conditional) independent test methods, we provide unit tests for the convenience of developing your own methods.
For the convenience of our community, CMU-CLeaR group maintains a list of benchmark datasets including real-world scenarios and various learning tasks. Please refer to the following links:
- https://github.com/cmu-phil/example-causal-datasets (maintained by Joseph Ramsey)
- https://www.cmu.edu/dietrich/causality/projects/causal_learn_benchmarks
Please feel free to let us know if you have any recommendations regarding causal datasets with high-quality. We are grateful for any effort that benefits the development of the causality community.
Please feel free to open an issue if you find anything unexpected. And please create pull requests, perhaps after passing unittests in 'tests/', if you would like to contribute to causal-learn. We are always targeting to make our community better!
Although causal-learn provides Python implementations for many causal discovery algorithms, there are more in the classical Java-based Tetrad program. For users who would like to incorporate arbitrary Java code in Tetrad as part of a Python workflow, we strongly recommend considering py-tetrad. Here is a list of reusable examples of how to painlessly benefit from the most comprehensive Tetrad Java codebase.
Please cite as:
@article{zheng2024causal,
title={Causal-learn: Causal discovery in python},
author={Zheng, Yujia and Huang, Biwei and Chen, Wei and Ramsey, Joseph and Gong, Mingming and Cai, Ruichu and Shimizu, Shohei and Spirtes, Peter and Zhang, Kun},
journal={Journal of Machine Learning Research},
volume={25},
number={60},
pages={1--8},
year={2024}
}