A web-based step-by-step interactive visualization of artificial neural networks
Switch branches/tags
Nothing to show
Clone or download
Latest commit f435772 Feb 4, 2016
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
source Added a flag to prevent re-running help while it is already running, … Feb 4, 2016
.gitignore Fixed a typo in the gitignore Feb 1, 2016
Gruntfile.js Build process cleanup Feb 1, 2016
LICENSE Initial commit Dec 22, 2015
README.md Update README.md Feb 2, 2016
bower.json Build process cleanup Feb 1, 2016
package.json Build process cleanup Feb 1, 2016

README.md

A web-based interactive visualization of artificial neural networks

This demo aims to serve as a tool to help visual learners, like myself, to gain visual intuition as to how artificial neural networks work. This demo allows a step-by-step iteration through a fully-connected backpropagation artificial neural network to get some intuition as to what goes on during the process of training. It uses the latest Javascript technologies, particularly the "yield" keyword that allows clean loop pause and resume. The default demo attempts to learn an XOR problem.

The demo is experimental and it uses the latest ES6 features so it won't work on all browsers, also the performance still needs optimization so I recommend running it in Google Chrome.

The code still needs cleanup and further testing so issue reports and pull request are most welcome. You can try it out by visiting http://experiments.mostafa.io/public/ffbpann/

How to install locally

To run this demo on your local machine (technically, it is a client side application so it will always run on your machine, but you get the idea :P), you will need to have nodejs, npm, grunt and bower installed.

After that clone the repository to your local system.

git clone https://github.com/drdrsh/interactive-bpann

navigate to the path where the code resides and run

npm install
grunt
grunt serve

This will install all the dependancies and run grunt server on port 9000, then navigate to the following url

http://localhost:9000/index.html