/Robby-The-Robot-Q-learning

AI Reinforcement Learning Using Q Learning

Primary LanguageJupyter NotebookMIT LicenseMIT

Robby-The-Robot-Q-learning

AI Reinforcement Learning Using Q Learning

  • Python version 3.10 or higher
  • Jupyter

  1. Load the libraries numpy, matplotlib, and scikit-learn

  2. Change the Hyper parameters below

N = 5000 # Number of Episodes M = 200 # number of actions to perform in each episode epsilon = 0.1 # Can start with high ε, and decrease it over the run eta = 0.2 gamma = 0.9

decreasing_epsilon = True # Change this to be False to have constant epsilon action_tax = False # Change this to be true to have a -5 action tax

  1. Train the model run This code: training_reward, trained_Q_Matrix = Q_Learning_Robby(N, M, epsilon, eta, gamma, decreasing_epsilon, Q_Matrix, action_tax)

  2. Test the model run This code: test_reward, tested_Q_Matrix = Q_Learning_Robby(N, M, epsilon, eta, gamma, False, trained_Q_Matrix, action_tax)

The Neural Network attempt is in it's own section, it re uses some of the code from above and is ran the same way It's hyper parameters use a smaller number of episodes because of the time to train each seperate neural network