node package manager


A simple perceptron written in Javascript

This is an implementation of the perceptron learning algorithm for node.js.

npm install perceptron

Let's teach a perceptron to implement a boolean AND function:

var perceptron = require('perceptron')
var and = perceptron()
and.train([1, 1], 1)
and.train([0, 1], 0)
and.train([1, 0], 0)
and.train([0, 0], 0)
// practice makes perfect (we hope...)
while(!and.retrain()) {}
and.perceive([1, 1]) // => 1
and.perceive([0, 1]) // => 0
and.perceive([1, 0]) // => 0
and.perceive([0, 0]) // => 0

The perceptron starts with random weights if you don't provide any defaults. Weights are adjusted according to the delta rule each time you call train and the current weights give the wrong answer. Since this adjustment can cause the perceptron to 'unlearn' previously learned inputs, retrain iterates over all previous inputs, calling train again. Both train and retrain return a boolean success value, indicating if the input(s) were learned.