CheckersRL

A reinforcement learning implementation for the game of Checkers.

Setup Instructions

1. Environment Setup

First ensure you have Python 3.x installed on your system. Then you can run the commands in your terminal:

# Create a virtual environment
python -m venv venv

# Activate the virtual environment
# On Windows:
venv\Scripts\activate
# On macOS/Linux:
source venv/bin/activate

2. Install Requirements

The project requires several dependencies for deep learning, scientific computing, and data processing. Install them using:

pip install -r requirements.txt

3. Running the Training Process

The training process involves two main components:

3.1 Training the Model (training.py)

To start the training process:

python training.py

The script supports four training modes, controlled by modifying the Oppenent variable in training.py:

• "itself" - Self-play training
• "curriculum" - Adaptive difficulty opponent
• "minmax" - Minimax algorithm opponent
• "random" - Random move opponent

To switch between modes:

1. Open training.py
2. Find this section at the bottom of the file:

if "__main__" == __name__ :
    Oppenent = "curriculum"  # Change this value to your desired mode
    print(Oppenent)
    GetModel(Oppenent=Oppenent)

3. Change "curriculum" to any of: "itself", "minmax", or "random"
4. Save and run python training.py

Each mode will:

• Save the model to models/{mode}aa.keras
• Generate statistics in winrates_{mode}.csv
• Create visualization in final_results/training/

3.2 Model Evaluation (round_robin.ipynb)

To evaluate trained models:

1. Navigate to and open round_robin.ipynb
2. Run all cells in the notebook

The notebook will:

• Load trained models from different checkpoints
• Conduct round-robin tournaments with 100 games between each agent
• Generate performance comparisons
• Visualize results with matplotlib

Project Structure

• checkers.py: Core game implementation with game logic and state management
• training.py: Main reinforcement learning training script
• round_robin.ipynb: Jupyter notebook for model evaluation
• requirements.txt: Complete list of project dependencies