This repository is for an agent we presented at The 2019 Science of Consciousness conference. Also see our slides from the conference for more information and context.

The DQN agent, model, and training code was borrowed heavily from the repository here.

This was tested only on Ubuntu 16.04 using Python 3.5.

1. Install pipenv: pip3 install pipenv
2. Set up the environment with the required packages: `pipenv install
3. Install and enable the jupyter notebook extensions (optional):

pipenv run jupyter contrib nbextension install --user
pipenv run jupyter nbextension enable codefolding/main
pipenv run jupyter nbextension enable toc2/main
pipenv run jupyter nbextension enable collapsible_headings/main
pipenv run jupyter nbextension enable varInspector/main
pipenv run jupyter nbextension enable spellchecker/main

1. Launch a jupyter notebook server: pipenv run jupyter notebook
2. Open and run the notebook located at notebooks/TSC-2019.ipynb

To train the model run: pipenv run python3 train.py