Neural Networks: An Overview - Neural Networks: Methodology and Applications - page 32

Fig. 1.15. Training of a feedforward neural network with one input variable and
three hidden neurons. The line is the output of the model, the crosses are the
elements of the training set. ( a ) initial state; ( b ) after one iteration; ( c )after6
iterations; ( d ) after 13 iterations (results obtained with the NeuroOne software
package by NETRAL S.A.)
1.3 Feedforward Neural Networks and Discrimination
(Classification)
In the early stages of the development of neural networks (in the years 1960's),
the main incentive was the development of pattern recognition applications, as
evidenced by the term perceptron that was used for the ancestor of present-day
neural networks. Indeed, the first nontrivial industrial applications of neural
networks, at the beginning of the 1980's, were related to pattern or signal
recognition. Therefore, the present section is devoted to a general presentation
of classification (or, equivalently, discrimination); it will be shown that many
classification problems can be viewed as nonlinear regression problems, which