Setup Notes

• Use Python 3
• Use Colaboratory if you don't want a local setup
• Use pip install to install packages

Module Information

• Gym: OpenAI Gym, a collection of environments to test RL algorithms on (e.g Atari games)
• Matplotlib: Fundamental plotting library for Python
• Numpy: Fundamental scientific computing package for python

Cartpole

• Run one episode on with a random policy
• "Train a policy" by creating random policies until you get one that gives a good return for an episode
• Test how well the above algorithm creates good policies by plotting the number episodes is required to produce a good policy, for 1000 trials

QCartpole

• Q-Learning on Cartpole
• DQN on Cartpole