Adding biases
	• A linear neuron is a more flexible model if we include a bias. • We can avoid having to figure out a separate learning rule for the bias by using a trick: – A bias is exactly equivalent to a weight on an extra input line that always has an activity of 1.