CvarAdversarialRL

Official code repository for "A Game-Theoretic Perspective on Risk-Sensitive Reinforcement Learning".

Initial setup

Create a virtual environment using

python3 -m venv ${YOUR_VENVS_DIR}/cvarRL

and activate it

source ${YOUR_VENVS_DIR}/cvarRL/bin/activate

Install the necessary requirements

pip3 install -r requirements.txt

Add the current folder to your PYTHONPATH

export PYTHONPATH="${PYTHONPATH}:${YOUR_PARENT_DIR}/CvarAdversarialRL"

Running the experiments and collecting figures

Scripts are produced to allow easy reproductibility of our results. They can be found in the scripts folder.

To run experiments:

./scripts/run_experiments.sh

To generate figures:

./scripts/generate_figures.sh