// Daniel Shiffman

// http://natureofcode.com

​

// Simple Perceptron Example

// See: http://en.wikipedia.org/wiki/Perceptron

​

// Code based on text "Artificial Intelligence", George Luger

​

// A list of points we will use to "train" the perceptron

let training = [];

// A Perceptron object

let perceptron;

​

// We will train the perceptron with one "Point" object at a time

let count = 0;

​

// The function to describe a line

function f(x) {

  let y = 0.6 * x + 0.2;

  return y;

}

​

function setup() {

  createCanvas(640, 240);

​

  // The perceptron has 3 inputs -- x, y, and bias

  // Second value is "Learning Constant"

  perceptron = new Perceptron(3, 0.001); // Learning Constant is low just b/c it's fun to watch, this is not necessarily optimal

​

  // Create a random set of training points and calculate the "known" answer

  for (let i = 0; i < 500; i++) {

    let x = random(1);

    let y = random(1);

    let answer = 1;

    if (y < f(x)) {

      answer = -1;

    }

    training[i] = {

      input: [x, y, 1],

      output: answer,

    };

  }

}

​

function draw() {

  background(255);

​

  // Draw the line

  strokeWeight(1);

  stroke(0);

  line(0 * width, f(0) * height, 1 * width, f(1) * height);

​

  // Draw the line based on the current weights

  // Formula is weights[0]*x + weights[1]*y + weights[2] = 0

  stroke(0);

  strokeWeight(2);

  let weights = perceptron.weights;

  y1 = (-weights[2] - weights[0] * 0) / weights[1];

  y2 = (-weights[2] - weights[0] * 1) / weights[1];

  line(0, y1 * height, width, y2 * height);

​

  // Train the Perceptron with one "training" point at a time

  for (let i = 0; i < 5; i++) {

    perceptron.train(training[count].input, training[count].output);

    count = (count + 1) % training.length;

  }

  // Draw all the points based on what the Perceptron would "guess"

  // Does not use the "known" correct answer

  for (let i = 0; i < training.length; i++) {

    stroke(0);

    strokeWeight(1);

    fill(127);

    let guess = perceptron.feedforward(training[i].input);

    if (guess > 0) noFill();

​

    let x = training[i].input[0] * width;

    let y = training[i].input[1] * height;

    circle(x, y, 8);

  }

}

​