https://github.com/gmamaladze/deep-snake

A snake game trained using simple deep learning implemented in client side javascript.
https://github.com/gmamaladze/deep-snake

convnetjs d3js deep-learning example

Last synced: about 2 months ago
JSON representation

A snake game trained using simple deep learning implemented in client side javascript.

• Host: GitHub
• URL: https://github.com/gmamaladze/deep-snake
• Owner: gmamaladze
• License: mit
• Created: 2017-06-29T17:14:01.000Z (almost 8 years ago)
• Default Branch: master
• Last Pushed: 2017-10-04T08:44:17.000Z (over 7 years ago)
• Last Synced: 2025-03-26T09:11:45.349Z (2 months ago)
• Topics: convnetjs, d3js, deep-learning, example
• Language: JavaScript
• Size: 540 KB
• Stars: 76
• Watchers: 4
• Forks: 16
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# deep-snake

A snake game trained using simple deep learning implemented in client side javascript.

![](deep-snake2.gif)

### [Demo: deep-snake.herokuapp.com](https://deep-snake.herokuapp.com/)

*NOTE:* The page will only run correctly in ES6 compatible (newer) browsers.

## What is this?

My goal was to create a simple demo of deep learning which is:

- very small, but still funny

- not too artificial

- demonstrating learning effect in minutes

- easy to read and modify

It is a toy. You can see how the snake is trained and gets gradually better.

The whole deep learning "magic" is happening in your browser. By reloading the page you are wiping Snake's

"brain". The complete code including visualization and the logic of the game is less then 500 lines.

# Frameworks Used

- [ConNetJS](https://cs.stanford.edu/people/karpathy/convnetjs/) - for deep learning.

- [D3.js](https://d3js.org/) - for visualization.