This is the code for the design, evaluation and implementation of contextual bandits to learn optimal physical exercises regimes that decrease pain in endometriosis patients.
For more information, see our arxiv submission.
If you do not want to setup the python environment, evaluation.html
still allows an interactive view of the intervention allocations in the browser.
To run the jupyter notebook, you can install the necessary dependencies with:
python3 -m pip install -e .[dev]
Now, to start a jupyter lab, run
jupyter lab
We provide the used dataset from the paper under data/2023-09-20-series.json
.
The evaluation notebook can be found under notebooks/evaluation.ipynb
.
The data was generated using a custom simulation library.
You can view the setup of the bayesian and data generating models in pymc under notebooks/data-generation.ipynb
.