Introduction to Soft Computing: Neuro-Fuzzy and Genetic Algorithms

568 Introduction to Soft Computing

computing the actual output for the said input.  is actual output when compared with the target output

as in

E

i

i

i

∑

∑

1

2

2

()

T

OA

O

i

i

T

O

gives the error for that training pair.  e same is computed for every training

pair and the average is considered for  tness calculation. Since the aim is to minimize the error whereas

GA is a maximization process, we cannot directly use the error E as the  tness measure. An obvious way

out is to take the reciprocal of E as the  tness.

F

E

=

1

(13.1)

 e rest of the process is usual GA. It may be noted that the GA-based learning of multi-layer nets does

not involve any backpro ...