Blackjack Q-Learning Project

Project Description

This project implements a Q-learning algorithm to train an agent to play the game of Blackjack optimally. The agent learns through interaction with the environment, improving its strategy over time by updating a Q-table based on the rewards received from different actions.

Features

• Environment: Utilizes the gymnasium library to simulate the Blackjack environment.
• Q-Learning Algorithm: Implements a Q-learning algorithm with adjustable hyperparameters such as learning rate, discount factor, and exploration rate.
• Performance Tracking: Tracks the mean return over episodes to monitor the learning progress.
• Visualization: Includes a plot of the mean return over time to visualize the agent's learning process.

Installation

To run this project, you need to have Python installed along with the following packages:

• gymnasium
• numpy
• matplotlib

You can install these packages using pip:

pip install gymnasium numpy matplotlib

Usage

1. Clone the repository.
2. Open the black_jack_rl.ipynb notebook.
3. Run the cells to train the agent and visualize the results.

Results

The learning process is visualized in the black_jack_rl_graph.png image, which shows the mean return over time as the agent learns to play Blackjack.

License

This project is licensed under the MIT License.

Acknowledgments

• The gymnasium library for providing the Blackjack environment.
• The Q-learning algorithm for reinforcement learning.