Q-learning Jupyter Notebook

This Jupyter Notebook contains an implementation of the Q-learning algorithm for the FrozenLake-v1 environment provided by the OpenAI Gym. The implementation allows the user to train the agent with the algorithm and evaluate its performance by generating a GIF of its behavior.

Dependencies

To run this Jupyter Notebook, the following dependencies are required:

• gym
• matplotlib
• numpy
• random
• imageio

Usage

To use this notebook, simply run all the cells in order. The Q-learning algorithm will be trained on the FrozenLake-v1 environment, and then evaluated to generate a GIF of the agent's behavior. The GIF will be saved in the current working directory under the name FrozenLake-v1(4x4).gif.

Parameters

The following parameters can be adjusted in the code:

• lr: The learning rate used in the Q-learning algorithm.
• gamma: The discount factor used in the Q-learning algorithm.
• max_epsilon: The maximum value of the exploration rate used in the Q-learning algorithm.
• min_epsilon: The minimum value of the exploration rate used in the Q-learning algorithm.
• decay: The decay rate of the exploration rate used in the Q-learning algorithm.
• n_train_episodes: The number of episodes used to train the agent.
• n_eval_episodes: The number of episodes used to evaluate the agent.
• max_steps: The maximum number of steps allowed per episode.

References

This implemenetation was made possible by this article: Introduction to Q-learning with OpenAI Gym