Skip to main content

SOM (Self-Organizing Maps)

In the field of artificial neural networks, which are also named after Teuvo Kohonen “Kohonen Networks”,  the SOM (Self-Organizing Maps) networks are known.

When training SOM nets, a net starts with random numbers, and the finished net is gradually produced, by frequent application of the training operation.

Here an image file is used instead of an assignment with random numbers as the initial assignment. Three layers are formed from the colour components red, green and blue. Then, this net is trained in according to the method of a Kohonen net, and thus changed.

Here you can see some results.