Most Simple Neural Network-Like Thing, no backprop

net.js

// Train data
var xor_data = [
	[[0, 0], [0]],
	[[0, 1], [1]],
	[[1, 0], [1]],
	[[1, 1], [0]],
];

// helper functions
function randomRange(count) {
	var range = [];
	for(var i = count; i--;){
		range.push(Math.random());
	}
	return range;
}
function randomMatrix(width, height){
	var matrix = [];
	for(var y = height; y--;) {
		matrix.push(randomRange(width));
	}
	return matrix;
}

function Net(in_count, hidden_count, out_count) {
	
	this.in_layer = new Array(in_count);
	this.hidden_layer = new Array(hidden_count);
	this.out_layer = new Array(out_count);

	// weights
	this.in_to_hidden = randomMatrix(in_count, hidden_count);
	this.hidden_to_out = randomMatrix(hidden_count, out_count);
}

Net.prototype.run = function(input) {
	if(this.in_layer.length != input.length){
		console.log("Input length doesn't match Net input length");
		return;
	}

	for(var i = this.in_layer.length; i--;) {
		this.in_layer[i] = input[i];
	}

	// Input -> Hidden
	for(var i = this.hidden_layer.length; i--;) {
		var sum = 0.0;
		for(var j = this.in_layer.length; j--;) {
			sum += this.in_layer[j] * this.in_to_hidden[i][j];
		}
		this.hidden_layer[i] = sum;
	}

	// Hidden -> Output
	for(var i = this.out_layer.length; i--;) {
		var sum = 0.0;
		for(var j = this.hidden_layer.length; j--;) {
			sum += this.hidden_layer[j] * this.hidden_to_out[i][j];
		}
		this.out_layer[i] = sum;
	}

	return this.out_layer;
};

var net = new Net(2, 3, 1);
console.log(net.run(xor_data[3][0]));