Installation Guide

ToyNN comes with multiple modules that are packaged together.

Requirements

ToyNN requires NodeJS v18.0.0 or higher to work.

Installation

To install/add ToyNN in your project you can use the following command:

npm

# add toynn to your project with npm
npm install toynn

Yarn

# add toynn to your project with yarn
yarn add toynn

Creating Models with ToyNN

ToyNN can be used to create complex neural networks. Let’s try to create a simple AND truth table neural network.

AND Truth Table

X	Y	Output
0	0	0
0	1	0
1	0	0
1	1	1

Implementation

import toynn from 'toynn';

const X = [
    new toynn.NArray([0, 0]).reshape(1, 2),
    new toynn.NArray([0, 1]).reshape(1, 2),
    new toynn.NArray([1, 0]).reshape(1, 2),
    new toynn.NArray([1, 1]).reshape(1, 2),
];

const y = [
    new toynn.NArray([0]),
    new toynn.NArray([0]),
    new toynn.NArray([0]),
    new toynn.NArray([1]),
];

const model = new toynn.NN('and');

const layer1 = new toynn.Layer(2, 3);
layer1.use(toynn.functions.linear);
const layer2 = new toynn.Layer(3, 1);
layer2.use(toynn.functions.sigmoid);

model.add(layer1);
model.add(layer2);

model.train({
    x: X,
    y,
    epochs: 500,
    alpha: 0.001,
    loss: toynn.errors.MSE,
    verbose: true,
});

let newData = new toynn.NArray([1, 0]).reshape(1, 2);
// make prediction
console.log(model.forward(newData).flatten());

Note: You can use ToyNN with typescript as well.