Overall structure
----------------------------------

- /configs contains the configuration files for MENSA and the baseline models.
- /data contains the data files used for training the models.
- /notebooks contains Jupyter notebooks to plot the results in the paper.
- /scripts contains bash scripts to run the experiments.
- /src contains the source code.
- /src/data contains Python files to load and preprocess the raw data, once it has been put in /data.

Steps to reproduce the results
----------------------------------

The code was tested in virtual environment with Python 3.9 and PyTorch 1.13.1.
By default, the experiments are performed 10 times with randomization seeds (0-9) for splitting the dataset.

1. First, install the required packages specified in the requirements.txt file.
2. Install the src directory by runnning: pip install -e .
3. Please refer to src/config.py to set appropriate paths.
4. Download the data needed to train the models. Rotterdam is provided in the /data directory. The MIMIC-IV,
SEER and PRO-ACT datasets need to be downloaded from their respective website and preprocessed according to the
description in Appendix E.1.
5. Generate the results by running the scripts in the /scripts folder.