One learning output neuron, no special activation functions, simple code that won't break your browser, easy to understand. Not exactly
