Adding in a Bias Term
This chapter introduces a feature of a typical neural network into a code. It includes an extra node in each layer called the bias node. Thus, whereas before, when we had two input variables, we used a 2–3–1 architecture, now we add a special node into all layers except the outer layer. The chapter takes the inputs and builds a model to predict the outputs. As a quick digression, actually there are plenty of other models that will fit this data perfectly. The pattern of the change in this weight cannot be understood in isolation of the changes in all the other weights, and in fact the challenge of understanding a complicated system of weights is generally not attempted.