Malfunctioning NN. I gave it my best shot at translating from python.

NN.js

let weights = Array.from({
	length: 3
}, () => Math.random() * 2 - 1)

// let weights = [-0.16595599, 0.44064899, -0.99977125]
let inputs = [
	[0, 1, 1],
	[1, 1, 1],
	[1, 0, 1],
	[0, 1, 1]
];
let outputs = [0, 1, 1, 0];

let exp = x => Math.E ** x;
let sig = x => x.map(i => 1 / (1 + (exp(-i))))

let sigP = x => x.map(i => i * (1 - i))

let add = (x, y) => x.map((i, a) => i + y[a]);
let sub = (x, y) => x.map((i, a) => i - y[a]);
let mlt = (x, y) => x.map((i, a) => i * y[a]);

let dot = (x, y) => x.map(a => a.map((b, i) => b * y[i])).map(a => a.reduce((i, b) => i + b));

// x.map(a => a.map((b, i) => b * y[i]))

console.log("weights", weights)

let outputLayer;

for (let i = 0; i < 20000; i++) {
	outputLayer = sig(dot(inputs, weights));
	let err = sub(outputs, outputLayer);
	let change = mlt(err, sigP(outputLayer))
	weights = add(weights, change);
}

let guess = inputs => sig(dot(inputs, weights))

console.log("weights", weights)
console.log("outputs", outputLayer)

console.log("guess", guess([1, 0, 0]))