Kohonen maps are a special kind of neural networks, which try to mimic natural brains with respect to the observation which is commonly represented by the "homunculus". This representation shows that those sensory regions which are very sensitive and have a lot of nerve connections to the brain (e.g. the tongue) influence a larger area of the neo cortex than areas which have less connections (e.g. the back). Further, one can see that this representation is maintaining the topology of the body.

Kohonen networks work in a similar way: the topology of the signal space is preserved, and parts of a signal with many samples are assigned to more "neurons" than parts with less details. One special feature of the Kohonen map is, that the neighborhood relationship can be set up by the user. Thus, you can, for example, set up a ring topology which can be used to optimize the classical travelling salesman problem.

Download
English version	[320 kB]
German version	[320 kB]
After downloading please unpack all files of the zipped packages and start the executable.

The program TravSalm applies a circular Kohonen map to the travelling salesman problem. You can set up a number of randomly distributed cities (depicted by the little houses), the size of the Kohonen map (the number of "neurons" in the circular network), and a few parameters of the network and the learning algorithm. For downloading the program click the little floppy disk at right.