idevede / relate_cate

ReLATE_CATE

This is the code of paper "Learning Minimal Sufficient Representations from Mixed Covariates for Treatment Effect Estimation". In the spirit of open research and transparency, we are committed to releasing all of our code, including the model architectures and hyperparameters, training scripts, evaluation metrics, and baseline implementations upon acceptance of our paper.

Code Transparency

1. The code in the SpeedData_Exp folder is used to reproduce Experiment 5.2 and Appendix F.2 in our paper. This includes:
   • The Mod 1 dataset used in the experiments
   • The proposed model architectures and hyperparameters
   • The full training and evaluation scripts
   • Outputs collection from multiple runs to demonstrate reproducibility

To run this code:

bash run_relate.sh

2. The Jupyter notebook file in the Covid_Exp_Appendix folder contains the code to reproduce the Covid-related experiments discussed in Appendix D. This includes:
   • Data processing steps
   • example of Model implementations
   • Full results and analyses