A threshold function changes its value sharply when its input reaches a particular value (the threshold). The simplest is a step function zero (or –1) when it input is less than zero and 1 thereafter. A hard threshold is simple to implement, but can be hard the learn; so neural networks typcally use a non-linear sigmod function instead.
Used on page 113
Also known as thresholding function
Step threshold function
Sigmoid threshold function