Robustness with Intersection Sets

The two folders contain experiments for supervised learning and bandits (supervised_cb), and reinforcement learning (rl).

Run Instructions

The two folders have run scripts with the extension *.sh.

To run supervised learning experiment, run

cd supervised_cb
bash run_supervised.sh

For bandits, run

cd supervised_cb
bash run_cb.sh
bash run_warfarin.sh

For reinforcement learning, run

cd rl
bash run_batch_cliff.sh
bash run_batch_sepsis.sh

Uncomment the code in each script to run with the full dataset or for multiple runs.

Packages Required

Instal by running pip install torch numpy scipy scikit-learn pandas joblib pymdptoolbox matplotlib seaborn tqdm

Acknowledgements

The code includes publicly available scripts from the following three codebases.

Supervised Learning and Contextual Bandits

Code for the experiments is adapted from Srivastava et al.'s implementation.

Reinforcement Learning

Code for the experiments is adapted from Roy et al.'s implementation.

Sepsis simulator is written by Oberst and Sontag.