The idea in backpropagation with gradient descent is to consider the entire network as a multivariate function that provides input to a loss function. The loss function calculates a number ...