Beginning Anomaly Detection Using

b (called bias) is added 

to the dot product result. From there, it passes through an 

activation function that 

decides if the entire node sends data or not. In this case, the activation function only 

varies between 0 and 1 depending on whether or not the dot product plus the bias 

reaches a certain value or not (threshold). It is possible to have other activation functions 

such as a sigmoid function, which outputs some value between 0 and 1.

Calling the output y and the input x, the basic function for each node can be 

represented by the equation in Figure 

3-5

.

An artificial neural network is comprised of interlinked layers of these nodes and can 

look like Figure 

3-6

.