mlp-generator

JS Project for Generating Simple Artificial Neural Networks

• Features
• Getting Started
  • Prerequisites
  • Installing
• TODO
• Built With
• Project Structure
• Versioning
• Authors
• License
• Acknowledgments

Features

• Visual mode MLP Artificial Neural Networks generation
• Realtime Training ANNs for SPAM classification
• Visualization of training results
• Downloading JSONs of trained ANNs

Soon

• Multiple datasets
• Dataset uploading
• Downloading JS functions of trained ANNs
• Module for converting images to textual datasets

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Prerequisites

• MongoDB 3.0+
• Node.js 8.0+
• NPM 6.0+

Installing

If you have git, you can install mlp-generator via git clone:

# Get the latest snapshot
git clone -b master https://github.com/Lure-of-Lysergide/mlp-generator

# Change directory
cd mlp-generator

# Install NPM dependencies
npm install

# Then simply start the application
node index.js

TODO

Front-end:

• Review the networks view :
  • Change Datatables plug-in with a better solution
  • Add modals for non-implemented buttons
• Review the visualization code :
  • Use D3.js for charting instead Highcharts.js
  • Don’t hardcode the array with colors
  • Creating of graphs must not be done with different functions for network and network-form views
• Use Webpack or something like that
• Reorganize client-side JavaScript :
  • DRY
  • Implement concatenation of *.js files
  • Implement minification of *.js files
  • Don’t use jQuery
• Maybe rewrite the fron-ent with Angular (or something else)

Back-end:

• Deploy
• Add functionality for multiple dataset processing :
  • Rewrite the db middleware
  • Add functionality for uploading datasets to server
  • Add functionality for editing datasets
  • Extend readable dataset formats to *.json, *.xml, *.csv, *.tsv, *.xls etc.
• Review the Network Controller :
  • Reorganize the async code (maybe with Rx.js)
  • Implement dataset randomization
• Add functionality for continue training after server crash
• Add functionality for listing and deleting Weights (without Networks) in DB
• Add functionality for converting images to textual datasets

Documentation:

• Add CHANGELOG

Built With

Back-end:

• brain.js - artificial neural network library
• Node.js - JavaScript runtime
• Express - server-side framework for Node.js
• MongoDB - document-oriented NoSQL database
• mongoose - MongoDB object modeling library
• pug - template engine

Front-end:

• D3.js - data manipulation and visualization library
• Highcharts.js - charting library
• Datatables - table plug-in for jQuery
• Bootstrap - front-end component library
• jQuery - JavaScript library (added only for supporting Datatables and Bootstrap scripts)

Project Structure

Name	Description
config/	Database URI and connection options
controllers/api.js	Controller for JSON API routes
controllers/network.js	Main controller for Neural Networks
controllers/router.js	Controller for routing pages
db/	The SPAMBASE dateset
middlewares/db.js	Script for reading data from spambase dataset
models/	Mongoose schema and models for Neural Networks
public/	Static assets (css, fonts, img, js, lib, favicon)
public/css/main.css	Compiled and minified css
public/fonts/	Font-Awesome files
public/js/	Client-side JavaScript
public/lib/	Client-side JavaScript dependencies
src/css/	Bootstrap, Font-Awesome and some themes
src/css/main.scss	Main style overrides
views/	Pug templates for each page
views/partials/	Base layout, navbar and footer partial templates
index.js	The entry point of application with cluster module
server.js	The Express server and route definitions

Versioning

This project use SemVer for versioning. For the versions available, see the tags on this repository.

Authors

• Niţa Radu - Initial work - Lysergide

License

This project is licensed under the MIT License - see the LICENSE file for details

Acknowledgments

This project use the Spambase Data Set from UCI Machine Learning Repository