Description

The repository provides the python implementation for the algorithms in the paper "Causal Abstraction Learning based on the Semantic Embedding Principle" by D'Acunto et al.

Organization

The material is organized into two main folders:

• src, containing the source code, that is, the algorithms implementation and the utils.py file with useful functions (such as the metrics);
• data, containing the results saved from the example.ipynb notebook, stored as parquet files to avoid versioning problems.

Additionally,

• example.ipynb shows how to apply the algorithms, in both the full-prior and partial-prior settings;
• environment.yml allows to install the conda environment. Please refer to conda User guide;
• .gitignore is the usual file that specifies intentionally untracked files that Git should ignore.