NeuroEvolution Bird

An educational project demonstrating how a Neural Network combined with a Genetic Algorithm can learn to play Flappy Bird from scratch.

Built with pure JavaScript, HTML, and CSS. No external libraries or frameworks.

🧠 How it Works

The Brain (Neural Network)

Each bird has a tiny brain (Neural Network) consisting of:

• Input Layer (4 nodes):
  • Y Position
  • Velocity
  • Distance to next pipe
  • Height of the pipe gap
• Hidden Layer (6 nodes): Processes the inputs using Tanh activation.
• Output Layer (1 node): Decides whether to jump (> 0.5) or not.

The Evolution (Genetic Algorithm)

1. Population: We start with 50 birds with random brains (random weights).
2. Selection: Birds that survive longer get a higher "fitness" score.
3. Reproduction: The best performing birds are selected to be parents.
4. Mutation: The offspring inherit the parents' brains but with slight random "mutations" to their weights. This introduces variation.
5. Loop: This process repeats endlessly, creating smarter birds over time.

🚀 Features

• Visualized Neural Network: See the active brain of the best bird in real-time.
• Speed Control: Speed up the simulation (up to 1000x) to train faster.
• Headless Mode: Disable rendering to train even faster.
• Save/Load: The best brain is automatically saved to local storage (feature in progress).

🛠️ Installation & Usage

1. Clone the repository

   git clone https://github.com/yourusername/neuroevolution-bird.git
   cd neuroevolution-bird

2. Run locally Since this is a static site, you can open index.html directly in your browser, or serve it with a local server for better module support:

   # Using Python
   python3 -m http.server
   
   # Using Node.js (npx)
   npx serve .

3. Open in Browser Go to http://localhost:8000 (or whatever port your server uses).

📄 License

This project is open source and available under the MIT License.