Kohonen Self-organizing Map

st33d over in the Processing forums posted this neat little example of a self-organizing map, utilizing some clever neural network code. It could use a bit of a visual polish, but the behaviour that it exhibits is quite interesting. I wonder if it could be used to generate interesting compositions if the nodes were persistant rather than being refreshed every frame?

I'm starting to get into some neural net work myself, and found the documents linked in the thread to be particularly interesting. Thanks, Aaron!