Contextualized Policy Recovery

This repository contains the code for replicating Contextualized Policy Recovery.

Installation

cd contextualized_policy_recovery
conda env create -f environment.yml
conda activate cpr
pip install --no-build-isolation --disable-pip-version-check -e .

Experiments

All other experiments to create the relevant figures can be found in notebooks/MIMIC_figures.py and notebooks/ADNI_figures.py

The code to run the bootstrap runs can be found in src/ADNI_bootstrap.py and src/MIMIC_bootstrap.py and is used to create ADNI_results.txt and MIMIC_results.txt.

The code to run the simulations can be found in notebooks/simulation_mdp.py and the code to plot the figure is notebooks/plot.py. Choose alpha=1 for using local information only and 1>alpha>0 for using global information.

Data

1. ADNI: After getting Data access from the Alzheimer's Disease Neuroimaging Initiative (ADNI), place the ADNIMERGE.csv file in the data/adni/raw/ folder.
2. MIMIC: After getting Data access for the MIMIC-III Clinical Database, follow the instruction for installing it in a Postgres database.