Kohonen, T.,
MATLAB Implementations and Applications of the Self-Organizing Map, Unigrafia
Oy, Helsinki , Finland
===
Self-organizing maps is a quite famous neural network, it is based on the observation that our brain organizes neurons mainly in a 2D configuration. It is a unsupervised neural network, which means, no professor is required to train the net. As Kohonen on the abovementioned book highlights,
It produces
low-dimensional projection images of high-dimensional data distributions,
On this short article, it is presented some excerpts from the abovementioned book.
This is not a mathematical treatise but a guide. It is neither any handbook of the syntax of computer programming, but rather a starter, which contains a number of exemplary program codes. You may call it a recipe book.
Since 1989, many SOM software packages have been published by various parties. Usually some diagnostic and other auxiliary programs have been included with the basic SOM algorithms. Some of these packages are freeware, others are commercial, and many researchers use specific SOM programs developed by themselves for particular applications. One might think that the methodology
would already have been established and standardized, but in practice one has encountered following kinds of problems: mentioned on the book.
The SOM has mainly been used by experts of mathematical statistics and programming. However, with a little of guidance, even non-specialists are expected to be able to use it correctly. So this is not a textbook, which is trying to define the syntax of the complete SOM Toolbox. The purpose of this discourse is to give the first advice in the correct application of the SOM, using exemplary scripts relating to different application areas.
