Causal CPC

Causal Contrastive Learning for Counterfactual Regression Over Time.

The code extends the project of Causal Transformer avialable at https://github.com/Valentyn1997/CausalTransformer/tree/main?tab=MIT-1-ov-file and released under MIT license.

Installations

conda create -n myenv python=3.10
conda activate myenv
pip install -r requirements.txt

The project uses the Hydra package to manage configuration files. All model configurations are in config/backbone, and those of the data are in config/dataset.

For MIMIC III data, please place (all_hourly_data.h5) to data/processed/

The config/config.yaml provides general configuration parameters, especially Hydra defaults that specify which model backbone to use, the dataset, and hyperparameter values.

An example of use for Causal CPC is

PYTHONPATH=. python3 runnables/train_ccpc_rep_est.py exp.seed=10