Kohonen Map Question- Preference for Corners of the Map
Winning neurons tend to be the ones in the far end of every corner
Hello everyone,
I was wondering if any of you has encountered the same problem as me, and how you solved it.
I have programmed a Kohonen map in Java, and even though I am just testing it with 14 patterns on a 10x10 grid, there seems to be a preference for the corners of the map. This is to an extent that many of the patterns which are in the same cluster as each other end up on the same coordinate on the map!
Is this normal?
What could cause this?
Any help is appreciated!




But maybe adjust the neighborhood functions so that there are no corners. Maybe if it wrapped around?
Jeff