As with individual perceptrons, multi-layer networks can be used for learning tasks. However, the learning algorithm that we look at (the backpropagation routine) is derived mathematically, using ...