Welcome to the INC (Impetus and Concepts) Machine Learning Project repository! In this project, we have implemented the NEAT (NeuroEvolution of Augmenting Topologies) algorithm from scratch, using the official research paper as our primary guide. Our goal was to explore and demonstrate the power of NEAT in evolving neural networks for tasks like playing games.
NEAT is a powerful evolutionary algorithm for evolving artificial neural networks. It is especially well-suited for problems where the network's architecture itself needs to evolve, making it a valuable tool in reinforcement learning tasks, like game playing. Our implementation closely follows the principles outlined in the original NEAT paper:
To test the capabilities of our NEAT implementation, we created a Flappy Bird clone using the Pygame library. Flappy Bird is a simple yet challenging game where a bird must navigate through pipes by jumping. Our goal was to evolve neural networks capable of playing the game and achieving high scores autonomously.
neat: Folder containing implementation of the NEAT algorithm.game: The Flappy Bird game folder (All Game specific files).__main__.py: The main script to run NEAT on the Flappy Bird game.README.md: You're reading it right now!
To get started with this project, follow these steps:
-
Clone this repository to your local machine:
git clone https://github.com/kshitijaucharmal/FlappyNEAT.git
-
Install the necessary dependencies. You can use
pipto install them:pip install pygame
-
Run the NEAT algorithm on the Flappy Bird game:
python .This will start the training process and display the progress of the evolving neural networks as they learn to play the game.
!!! Not implemented yet !!!
You can fine-tune the NEAT algorithm by modifying the parameters in the config.txt file.
These parameters include population size, mutation rates, and various other settings that can influence the evolution process.
- Kshitij Aucharmal - Algorithm Implementation
- Alisha Shaikh - Game Development
- Tanish Chaudhari - Project Management
Contributions are always welcome!!
Just fork the repository and create a pull request with your desired changes, we'll be sure to review them as soon as possible!
We would like to express our gratitude to the creators of the NEAT algorithm for their pioneering work in neuroevolution. This project wouldn't have been possible without their research.
This project is licensed under the GPL 3.0 License - see the LICENSE file for details.
Feel free to explore the code, experiment with the NEAT parameters, and enjoy watching neural networks learn to play Flappy Bird! If you have any questions or suggestions, please don't hesitate to open an issue or reach out to us. Happy coding!