In this event, we'll find out how to establish the error of a neural network's output, and minimize it, by first principles. By so doing, the neural network would be able to approximate the function that maps its inputs to its outputs including the edge cases, and would successfully predict outputs for inputs it had never seen before.