/dcg_supp

Deep Causal Graphs - libraries and supplementary material for the paper "Estimand-Agnostic Causal Query Estimation with Deep Causal Graphs"

Primary LanguageJupyter NotebookMIT LicenseMIT

dcg - Deep Causal Graphs

This repository contains all three libraries necessary for training Deep Causal Graphs (DCG), models for causal query estimation without estimands.

The "paper" release includes a zip file with all experiments included in the paper "Estimand-Agnostic Causal Query Estimation with Deep Causal Graphs".

Installation

Install all three libraries here included with pip (preferably using a virtual environment) in the following order:

pip3 install -e torch-misc
pip3 install -e flow
pip3 install -e dcg

Tutorial

We have created a tutorial for the dcg library, which you can find in the tutorial/ folder.