Introduction

This project is the python implementation of the model "Adversarial Balancing based representation learning for Causal Effect Inference (ABCEI)"https://arxiv.org/abs/1904.13335. The entire framework is rewritten based on the Counterfactual regression. Evaluation and Hyper-parameter search parts are reused to ensure the fairness of comparison.

Requirements

To run the code, the following libraries are needed:

Python 3.5;
Tensorflow 1.4;
Numpy 1.15;

Run

Three examples are provided for running this code on IHDP, Jobs and Twins datasets:

ihdp.sh;
jobs.sh;
twins.sh;

Those files can be used combining with workload manager like slurm:

sbatch -p [cluster name] [--gres=gpu:1] ihdp.sh