Reinforcement Learning

Agent

rl1/

TD(random policy): agent_td.py

Sarsa: agent_sarsa.py

Q-Learning: agent_q.py

n-step Sarsa: agent_n_sarsa.py incomplete

Expected Sarsa: agent_exp_sarsa.py incomplete

Dyna-Q: agent_dyna_q.py

Dyna-Q+: agent_dyna_q_plus.py

rl2/

Sarsa(lambda) implemented with Eligibility Trace: agent_ET.py

Environment

Simple Grid: env.py

Cliff: env_cliff.py

Maze: env_maze.py

Experiment

By typing out python3 experiment.py in command line tool.

Acknowledgement

RL Glue framework

rl_glue.py link: https://sites.google.com/a/rl-community.org/rl-glue/Home?authuser=0

Tanner, B., & White, A. (2009). RL-Glue: Language-independent software for reinforcement-learning experiments. The Journal of Machine Learning Research, 10: 2133--2136.

Tile coding

tiles3.py link: http://incompleteideas.net/tiles/tiles3.html

abmushi/rl