Experts-as-Priors: Sequential Decision Making with Expert Demonstrations under Unobserved Heterogeneity

Installation • Experiments • Citation

Installation

Set up a virtual environment and load it

python -m venv $ENV_NAME
source $ENV_NAME/bin/activate

Install jax==0.4.14:

pip install jax==0.4.14 jaxlib==0.4.14+cuda11.cudnn86 -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

Then, install the flashbax package via

pip install git+https://github.com/instadeepai/flashbax.git

Finally, other requirements are installed with pip install -r requirements.txt

Experiments

For bandit experiments, follow the notebook notebooks/bandit.ipynb. For deep sea experiments, run notebooks/deep_sea.ipynb.