C++ Neural Networks and Fuzzy Logic: Preface

1 0 0 0 1

1 0 0 0 1

1 1 1 1 1

1 0 0 0 1

1 0 0 0 1

For use in the Kohonen program, we need to serialize the rows, so that all entries appear on one line.

C++ Neural Networks and Fuzzy Logic:Preface

Chapter 12 Application to Pattern Recognition

244

---

For the characters A and X you would end up with the following entries in the input file, input.dat:

0 0 1 0 0  0 1 0 1 0  1 0 0 0 1  1 0 0 0 1  1 1 1 1 1  1 0 0 0 1  1 0 0 0 1

 << the letter A

1 0 0 0 1  0 1 0 1 0  0 0 1 0 0  0 0 1 0 0  0 0 1 0 0  0 1 0 1 0  1 0 0 0 1

 << the letter X

Monitoring the Weights

We will present the Kohonen map with many such characters and find the response in output. You will be

able to watch the Kohonen map as it goes through its cycles and learns the input patterns. At the same time,

you should be able to watch the weight vectors for the winner neurons to see the pattern that is developing in

the weights. Remember that for a Kohonen map the weight vectors tend to become aligned with the input

vectors. So after a while, you will notice that the weight vector for the input will resemble the input pattern

that you are categorizing.