Train an AI to play chess, and have your AI compete with other people's AIs! Named after https://github.com/ChesleyTan
This project was made for a Software Development class at Stuyvesant High School.
Upon making an account, users will be given a baby AI with almost no knowledge of chess other than the rules. Players will train the AI to improve its skills. Players will be ranked based on an ELO ranking system, gaining ELO points when their AI defeats another player's AI in a competitive match.
- Python Flask
- TensorFlow
- Python Chess
- SciPy
- Mongo
- Sunfish
Hi, it's Ethan (elc1798). When I first started this project, I knew next to nothing about neural networks, machine learning, etc. A large portion of my knowledge came from reading research papers, the TensorFlow tutorial (https://www.tensorflow.org/versions/master/tutorials/index.html), and some open source neural net projects on GitHub. A large portion of the neural net in this project, named Chessley, is based on Erik Bernhardsson's Deep Pink: https://github.com/erikbern/deep-pink. I would also like to thank Keras's source code for helping me port Deep Pink's Theano framework to TensorFlow, where I made some modifications and tweaks (that probably contain some unintentional ones).
KERAS: https://github.com/fchollet/keras
Deep Pink Blog Post: http://erikbern.com/2014/11/29/deep-learning-for-chess/
| Name | Role |
|---|---|
| Ethan Cheng | Deep Learning AI |
| Ishraq Bhuiyan | Flask Web Application Backend |
| Jion Fairchild | HTML and CSS Frontend |
Clone the project and build all dependencies:
$ git clone https://github.com/elc1798/chessley-tan
$ cd chessley-tan
$ make setup
To deploy the website, just do
$ gunicorn -w 4 -b 0.0.0.0:5000 wsgi_handler:app
To access the site, access your IP address or website on port 5000. The port can be changed to suit your needs.
Our version is hosted at http://vocab.csproject.org:5000