Neura is an intuitive, fast and customizable neural network for JavaScript


License
MIT
Install
npm install neura@1.1.1

Documentation

neura

Neura is the fast and customizable neural network for JavaScript

Requirements:

Node.js 4.2+

Installation:

npm i neura

Usage:

Initialize

const Neura = require('neura');
let n = new Neura({layers: 2, iterations: 10000});

Add train data and run

n.train({input: [[1, 0, 0],[0, 1, 1],[1, 1, 0]], output: [0, 0, 1]});
n.run([1,0,0]);

or

n.train({input: [[1, 0, 0],[0, 1, 1],[1, 1, 0]], output: [0, 0, 1]})
  .run([1, 0, 0]);

Data can be saved in file

n.train({input: [[1, 0, 0],[0, 1, 1],[1, 1, 0]], output: [0, 0, 1]})
  .saveToFile('./output.json');

Saved neural network synapse can be used again

n.readFromFile('./output.json')
  .run([1, 0, 0]);

For example:

Lets create a real estate scoring (chance to be sold) Yes or no denoted by 1/0

id Price in M$ Rooms Area Sold
1 1.12 3 25 0
2 25.2 4 116 1
... ... ... ... ...
100000 4.1 1 65 1

Input is 2, 3 and 4 columns, output is 5 column. Now we have to train our network

let data = sellingChanceData;
n.train(data)

finnaly

n.train(data).run([18, 2, 95]); // 0.85 => This house is likely to be sold

Contributing:

Create issue or PR

git clone https://github.com/AlexanderGY/neura.git
cd neura && npm install

Coming soon

Image recognition