Introduction to Soft Computing: Neuro-Fuzzy and Genetic Algorithms

194 Introduction to Soft Computing

the classes {(0, 1), (1, 0)} and {(0, 0)}.  e technique is shown is Fig. 6.23. Fig. 6.23(b) shows another pair

of lines that solve the problem in a di erent way.

Using this idea, it is possible design a multi-layered perceptron to solve the XOR problem. Such a

multilayered perceptron is shown in Fig. 6.24. Here the  rst perceptron Y  res only when the input is

(1, 1). But this sends a large inhibitive signal of magnitude −2.0 to the second perceptron Z so that the

excitatory signals from x

1

and x

2

to Z are overpowered. As a result the net input to Z attains a negative

value and the perceptron fails t ...