To preprocess the eICU dataset for 24-hour mortality prediction, follow the cohort construction and preprocessing steps defined in the YAIB framework.

You can compute KL-divergence–based scores between datasets in both plaintext and encrypted settings. To compute divergence scores without encryption, run:

python kl_utils.py --score --plaintext --dataset eicu --data_path /path/to/data

To compute divergence scores in a privacy-preserving (encrypted) setting, run:

python kl_utils.py --score --encrypted --dataset eicu --data_path /path/to/data