ConvNetJS

Abstraction: JavaScript library for training neural networks in the browser

Key points:

• Written by Andrej Karpathy during his PhD at Stanford; allows formulating and solving neural networks entirely in JavaScript
• Supports running in Node.js (npm package: convnetjs) as well as in-browser
• Extended by community contributions beyond the original author's work
• Available on GitHub under MIT license; accepts pull requests for new features, layers, and demos
• Enables demos and tutorials runnable without any server-side setup

Connections: Andrej Karpathy · Convnetjs · Stanford · Deep Learning · Neural Networks

Source: http://cs.stanford.edu/people/karpathy/convnetjs/