Computational Intelligence

Neural Network Components and Terminology 173

x

1

y

j

x

2

y

j

= F(X, Vj, Wj)

x

3

x

n

w

jn

v

jn

w

j3

w

j2

v

j2

v

j3

w

j1

v

j1

Figure 5.6 A PE with mean–variance connections.

y

j

= g



n



i = 1



w

ji

− x

i

v

ji



2



(5.6)

g (x) = exp



−x

2

2



(5.7)

Note that it is possible to remove one of the two connections in a mean-variance

network, if the variance is known and stationary, by dividing by the variance prior to

neural network processing. Gaussian nonlinear functions are described in the next

section.

Processing Element Activation Functions

Processing element activation functions, also sometimes referred to as threshold

functions or squashing functions, map a PE’s (possibly) infinite domain to a pre- ...