Implementation for our paper A First-Occupancy Representation for Reinforcement Learning.
dependencies: numpy
, scipy
, matplotlib
, seaborn
files:
agents.py
: basic agent classes for GPI, value iteration, the FR, and the SRenvs.py
: FourRoom and Escape environment classesfrp.py
: FR planning (FRP) agent classutils.py
: basic helper functionsrunners.py
: functions for running experimentsfour_rooms.ipynb
: result notebook
If you find this code useful, please cite using:
@misc{moskovitz2021firstoccupancy,
title={A First-Occupancy Representation for Reinforcement Learning},
author={Ted Moskovitz and Spencer R. Wilson and Maneesh Sahani},
year={2021},
eprint={2109.13863},
archivePrefix={arXiv},
primaryClass={cs.LG}
}