My first perceptron that models the AND logical gate

perceptron.js

// AND logic
// @see https://medium.com/@stanleydukor/neural-representation-of-and-or-not-xor-and-xnor-logic-gates-perceptron-algorithm-b0275375fea1
// The perceptron model is x1+x2–1 where weights is 1
// Classify into 1 or -1 where
// 1 = true
// -1 = false
const trainingData = [
    [1, 1, 1],
    [1, 0, -1],
    [0, 1, -1],
    [0, 0, -1]
]

function random(min, max) {
    return Math.random() * (max - min) + min;
}

// activation function = the sign function
function activation(sum) {
    if (sum > 0) return 1
    else return -1
}

// feed forward
// x1, x2 -> activation -> output guessed
function feedForward(inputs, weights) {
    let sum = 0
    for (let i = 0; i < inputs.length; i++) {
        sum += inputs[i] * weights[i];
    }
    return activation(sum);
}

function train(epochs, learningRate) {
    // x1, x2 and the bias
    const weights = [
        random(-1, 1),
        random(-1, 1),
        1
    ];
    let i = 0;
    const bias = -1

    while (i < epochs) {
        i++;
        for ([x1, x2, y] of trainingData) {
            const inputs = [x1, x2, bias];
            const guess = feedForward(inputs, weights);
            const err = y - guess;
            // update the weights
            for (let i = 0; i < weights.length; i++) {
                weights[i] += learningRate * err * inputs[i];
            }
        }
    }

    return weights;
}

function predict(input, weights) {
    const [x1, x2, y] = input;
    const output = feedForward(input, weights);
    console.log(`${x1} AND ${x2} = guessed ${output}, label is ${y}`)
}

// Test the feedForward and activation function based on the expected result with fixed weights
const fixedModel = [1, 1, 1];
console.log('got', feedForward([1, 1, -1], fixedModel), 'expected', 1);
console.log('got', feedForward([1, 0, -1], fixedModel), 'expected', -1);
console.log('got', feedForward([0, 1, -1], fixedModel), 'expected', -1);
console.log('got', feedForward([0, 0, -1], fixedModel), 'expected', -1);

// Train the model
let model = train(1000, 0.01);
console.log(model);

for (let input of trainingData) {
    predict(input, model);
}