Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/scienceai/neocortex
Run trained deep neural networks in the browser or node.js
https://github.com/scienceai/neocortex
Last synced: 5 days ago
JSON representation
Run trained deep neural networks in the browser or node.js
- Host: GitHub
- URL: https://github.com/scienceai/neocortex
- Owner: scienceai
- License: apache-2.0
- Archived: true
- Created: 2015-09-19T03:09:45.000Z (about 9 years ago)
- Default Branch: master
- Last Pushed: 2016-11-11T02:44:47.000Z (almost 8 years ago)
- Last Synced: 2024-08-01T12:34:47.583Z (3 months ago)
- Language: JavaScript
- Homepage: https://scienceai.github.io/neocortex
- Size: 552 KB
- Stars: 275
- Watchers: 17
- Forks: 26
- Open Issues: 8
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-starred - scienceai/neocortex - Run trained deep neural networks in the browser or node.js (others)
README
### [PROJECT PAGE AND EXAMPLES](https://scienceai.github.io/neocortex)
Run trained deep neural networks in the browser or node.js. Currently supports serialization from trained [Keras](https://github.com/fchollet/keras/) models (note: not up-to-date with the current Keras API).
[](https://travis-ci.org/scienceai/neocortex)
[](https://www.npmjs.com/package/neocortex-js)
### Background
Training deep neural networks on any meaningful dataset requires massive computational resources and lots and lots of time. However, the forward-pass prediction phase is relatively cheap - typically there is no backpropagation, computational graphs, loss functions, or optimization algorithms to worry about.
What do you do when you have a trained deep neural network and now wish to use it to power a part of your client-facing web application? Traditionally, you would deploy your model on a server and call it from your web application through an API. But what if you can deploy it _in the browser_ alongside the rest of your webapp? Computation would be offloaded entirely to your end-user!
Perhaps most users will not be able to run billion-parameter networks in their browsers quite yet, but smaller networks are certainly within the realm of possibility.
The goal of this project is to provide a lightweight javascript library that can take a serialized Keras, Caffe, Torch or [insert other deep learning framework here] model, together with pre-trained weights, pack it in your webapp, and be up and running. Currently supports serialization from trained [Keras](https://github.com/fchollet/keras/) models.
### Examples
- MNIST multi-layer perceptron / [src](https://github.com/scienceai/neocortex/tree/master/examples/mnist_mlp) / [demo](http://scienceai.github.io/neocortex/mnist_mlp)
- CIFAR-10 VGGNet-like convolutional neural network / [src](https://github.com/scienceai/neocortex/tree/master/examples/cifar10_cnn) / [demo](http://scienceai.github.io/neocortex/cifar10_cnn)
- LSTM recurrent neural network for classifying astronomical object names / [src](https://github.com/scienceai/neocortex/tree/master/examples/astro_lstm) / [demo](http://scienceai.github.io/neocortex/astro_lstm)
You can also run the examples on your local machine at [`http://localhost:8000`](http://localhost:8000):
```sh
$ npm run examples-server
```
### Usage
See the source code of the examples above. In particular, the CIFAR-10 example demonstrates a multi-threaded implementation using Web Workers.
In the browser:
```html
// use neural network here
```
In node.js:
```sh
$ npm install neocortex-js
```
```js
import NeuralNet from 'neocortex-js';
```
The core steps involve:
1. Instantiate neural network class
```js
let nn = new NeuralNet({
// relative URL in browser/webworker, absolute path in node.js
modelFilePath: 'model.json',
arrayType: 'float64', // float64 or float32
});
```
2. Load the model JSON file, then once loaded, feed input data into neural network
```js
nn.init().then(() => {
let predictions = nn.predict(input);
// make use of predictions
});
```
### Build
To build the project yourself, for both the browser (outputs to `build/neocortex.min.js`) and node.js (outputs to `dist/`):
```
$ npm install
$ npm run build
```
To build just for the browser:
```
$ npm run build-browser
```
### Keras
A script to serialize a trained [Keras](http://keras.io/) model together with its `hdf5` formatted weights is located in the [`utils/` folder](https://github.com/scienceai/neocortex/blob/master/utils/serialize_keras.py). It currently only supports sequential models with layers listed below. Implementation of graph models is planned.
Functions and layers currently implemented are listed below. More forthcoming.
+ Activation functions: `linear`, `relu`, `sigmoid`, `hard_sigmoid`, `tanh`, `softmax`
+ Advanced activation layers: `leakyReLULayer`, `parametricReLULayer`, `parametricSoftplusLayer`, `thresholdedLinearLayer`, `thresholdedReLuLayer`
+ Basic layers: `denseLayer`, `dropoutLayer`, `flattenLayer`, `mergeLayer`
+ Recurrent layers: `rGRULayer` (gated-recurrent unit or GRU), `rLSTMLayer` (long short-term memory or LSTM)
+ Convolutional layers: `convolution2DLayer`, `maxPooling2DLayer`, `convolution1DLayer`, `maxPooling1DLayer`
+ Embedding layers: `embeddingLayer` (maps indices to corresponding embedding vectors)
+ Normalization layers: `batchNormalizationLayer`
### Tests
```
$ npm test
```
Browser testing is planned.
### Credits
Thanks to @halmos for the logo.
### Citation
[](http://dx.doi.org/10.5281/zenodo.44692)
### License
[Apache 2.0](https://github.com/scienceai/neocortex/blob/master/LICENSE)