Nearsighted Paramecium Multiverse
Miss any of our Open RFC calls?Watch the recordings here! »

browsernn

1.0.1 • Public • Published

browserNN.js

browserNN is an easy to use javascript library that allows you to train deep neural networks in your own browser.

For full documentation and a demo, please see: https://carbonati.github.io/posts/browsernn/

Example

model = new browsernn.Model();
 
layers = []; 
 
layers.push({type: 'input', n: 1, d: 2, depth: 1}); 
layers.push({type: 'dense', n_neurons: 8, activation: 'relu'}); 
layers.push({type: 'dense', n_neurons: 4, activation: 'tanh'}); 
layers.push({type: 'softmax', n_classes: 2}); 
 
params = {optimizer: 'sgd', 
        learning_rate: .01, 
        momentum: 0.01, 
        batch_size: 10, 
        l2_decay: 0.01};
        
trainer = new browsernn.Trainer(model, layers, params); 

Use in Node

  1. install: npm install browsernn
  2. Use: var browsernn = require("browsernn");

License

MIT

Install

npm i browsernn

DownloadsWeekly Downloads

1

Version

1.0.1

License

MIT

Unpacked Size

56.6 kB

Total Files

20

Last publish

Collaborators

  • avatar