brain_server

A lightweight HTTP Koa server that uses Q, brain.js, and Redis as a datastore. Can be used as a good starting point for anyone interested in machine learning with Javascript.

1. Install Node.js (at least 11.13)
2. Install Redis
3. git pull https://github.com/acoyfellow/brain_server.git
4. npm install dependencies, node --harmony server.js to start
5. Add training data
6. HTTP GET the root url for to classify { r: 1, g: 0.4, b: 0 }

This is a simple starter point for writing an API that uses brain.js (which uses a neural network + backpropogation algo for training) to do some simple classification.

Here is some training data you can use (from brain.js docs):

{input: { r: 0.03, g: 0.7, b: 0.5 }, output: { black: 1 }}

{input: { r: 0.16, g: 0.09, b: 0.2 }, output: { white: 1 }}

{input: { r: 0.5, g: 0.5, b: 1.0 }, output: { white: 1 }}

To insert data into Redis:

• HTTP POST to the endpoint http://{your_server}/insert in x-www-form-urlencoded format with 4 inputs
• r, g, b, and output.
• r, g, and b are numbers
• output is boolean 0 or 1, where 0 is White and 1 is Black.