Hands-On Artificial Intelligence for Beginners

Weights and bias factors

Two key parts of ANNs are weights and biases. These elements help us squash and stretch our nonlinearities to help us better approximate a function. 

Weights are applied at every transformation in a neural network, and help us stretch a function. They essentially change the steepness of the nonlinearity. Bias factors are important parts of ANNs as well; you've probably noticed them in the diagrams shown so far in this chapter. Bias factors are values that allow us to shift our activation function left or right to help us best approximate a natural function. 

How does this work in practice? Let's say you have a simple two-neuron setup such as the one in the following diagram: 

Let's see how adding weights into the ...