Deep Learning In Javascript 🧠

Introduction

Brain.js is a powerful JavaScript library that allows you to create and deploy neural networks in the browser or in a Node.js environment. It efficiently utilizes the GPU when available, and falls back to pure JavaScript when necessary.

The Advantages of Deep Learning in JavaScript

The primary benefit of using Brain.js is the ability to leverage machine learning models in JavaScript. As more companies harness the power of machine learning, JavaScript remains a popular programming language in the industry. Thus, combining the capabilities of both provides numerous opportunities for innovation.

Getting Started with Brain.js

Before diving in, ensure you have the following prerequisites:

• Node.js (v14.0.0 or greater)
• Python 2.7
• basic understanding of JavaScript

Windows

• Microsoft Visual Studio
• d3dcompiler_47.dll should be in c:\windows\system32, but if isn't then you can find another copy in the deps/ folder

MacOS

• XCode

To begin, install the Brain.js package using npm or yarn:

npm install brain.js

yarn add brain.js

Now, let's create a simple example to illustrate the use of Brain.js.

const brain = require("brain.js");

const net = new brain.NeuralNetwork({ hiddenLayers: [3] });
const trainingData = [
  { input: [0, 0], output: [0] },
  { input: [0, 1], output: [1] },
  { input: [1, 0], output: [1] },
  { input: [1, 1], output: [1] },
];
net.train(trainingData);
console.log(net.run([0, 0]));
console.log(net.run([0, 1]));
console.log(net.run([1, 0]));
console.log(net.run([1, 1]));

Understanding the Code

The code above first declares a new NeuralNetwork instance and then trains the model with trainingData. The training data consists of a simple set of combinations containing 0 or 1; if a 1 exists in either position, the output is 1, otherwise, it is 0. When we run this example, we can observe that the model makes fairly accurate predictions.

Float32Array(1) [ 0.11244010180234909 ]
Float32Array(1) [ 0.9405856728553772 ]
Float32Array(1) [ 0.9397708773612976 ]
Float32Array(1) [ 0.9918596744537354 ]

This is just the tip of the iceberg; the Brain.js library offers a plethora of features for experimentation and creating innovative applications.

The repo has a lot more examples on training NeuralNetworks.

Brain.js also makes it easy to make predictions on data using forecast. I decided to feed it some stock market information in a more in-depth example you can see here.

Last note - if you'd like to sponsor the project or even make a one time donation you may do so here!