Conformal Counterfactual Inference under Hidden Confounding (KDD’24)

Zonghao Chen*, Ruocheng Guo*, Jean-François Ton, Yang Liu

Code for the KDD '24 paper, Conformal Counterfactual Inference under Hidden Confounding.

Slides

2-min Promotional Video

First Work on Handling Hidden Confounding for Conformal Causal Inference without strong assumptions such as Bounds on the Density Ratio

For now, please cite the arxiv version if you find this paper/repository is helpful.

@misc{chen2024conformal,
      title={Conformal Counterfactual Inference under Hidden Confounding}, 
      author={Zonghao Chen and Ruocheng Guo and Jean-François Ton and Yang Liu},
      year={2024},
      eprint={2405.12387},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}

Overview of the work

Dependencies

pip3 install -r requirements.txt

Datasets

Datasets used in this paper include

• synthetic data simulated in run_syn.py
• Yahoo!R3 dataset
• Coat

Running the experiment

On a linux system, you can run the provided bash scripts, for the synthetic data experiment:

bash run_syn.sh

For the recommendation system rating prediction experiments:

bash iDCF/run_coat.sh
bash iDCF/run_yahoo.sh