Results folder.

Figs and text results of experiments saved here.

During training, real-time results can be displayed via tensorboard:

tensorboard --logdir=/results/log/tensorboard/<tb_log_exp_name>

Logs of hyper-parameter tuning runs are found in /log and can similarly be displayed (RayTune docs).

Reproducibility Instructions

To run all experiments from the paper:

• Main results

  posh uv run main.py --domain_shift "hospital" uv run main.py --domain_shift "region" uv run main.py --domain_shift "ward" uv run main.py --domain_shift "ethnicity" uv run main.py --domain_shift "ethnicity_coarse" uv run main.py --domain_shift "age"

• Alternative outcome definitions:

  posh uv run main.py --outcome "Shock_4h" uv run main.py --outcome "Shock_12h" uv run main.py --outcome "ARF_4h" uv run main.py --outcome "ARF_12h"

• Additional sequential models:

  posh uv run main.py --models "RNN" "GRU" "LSTM"

• Regularization experiments:

  posh uv run main.py --dropout

• Class incremental experiments:

  posh uv run main.py --class_shift