Computational Intelligence Techniques - Computational Intelligence in Biomedical Engineering - page 81

(b)

(a)

(c)

(d)

FIGURE 3.6
Examples of activation functions that have been used in artificial neurons.
(a) Step function; (b) Gaussian; (c) piecewise linear; and (d) sigmoid.

c. Output . The output y i represents the strength of the electrical impulse

travelling along the axon.

d. Feedback . In some ANNs, there is some provision to feed the output

back into the input so that the neural network is adaptive.

Learning in artificial neurons, perceptrons, or neural networks is by adjust-
ing the neuron weights so that the network output matches the required or true
output. In a pattern classification problem (see Section 3.1.2), for example,
each training pattern consists of the pair

and the aim is to have the
neuron learn the complex relationship between the training inputs and the
class labels. The error between the output of the neuron, f ( u i ) and the actual
class label y i is usually taken to be the difference

{ x i ,y i }

E i = y i − f ( u i ) (3.3)

If the number of training patterns is n , the sum of errors for a single neuron
E SN is written as

n n

E SN = E i = y i − f ( u i ) (3.4)

i i