GARLIC: Graph Attention-based Relational Learning of Multivariate Time Series in Intensive Care

This repository contains the implementation of GARLIC.

Requirements

Install dependencies using pip:

pip install -r requirements.txt

The code is tested with Python 3.11.

Datasets

PhysioNet Challenge 2012 & 2019

Download from:

• https://www.physionet.org/content/challenge-2012/1.0.0/
• https://physionet.org/content/challenge-2019/1.0.0/

MIMIC-III

Request access from:

• https://physionet.org/content/mimiciii/1.4/

Preparation

To reproduce results, place the corresponding raw datasets in the following directories:

./data/rawdata/P12/
./data/rawdata/P19/
./data/rawdata/MIMICIII/

Preprocessed data will be automatically generated in ./data/processed_data/ upon the first run.

Run the Model

To train and evaluate the model on each dataset, run the corresponding shell scripts:

bash run.sh
bash interpretability_evaluation.sh

Each script initializes the model, preprocesses data if needed, trains the model, and reports AUROC and AUPRC.