[comp.ai.neural-nets] Kohonen Self Organizing Feature Maps

simonof@aplcen.apl.jhu.edu (Simonoff Robert 301 540 1864) (09/13/90)

I have read in Kohonen's Book, "Self Organization and Associative
Memory" that the Self Organizing Feature Map can be used as a
Novelty Detector.  I have implemented said network and have spent
many hours attempting to see how one could use the output
(grid of distances between the input vector and the vector 
represented by the weight vector for a given node) to determine
novelty.  Could someone please give me a bit of help in this?

Thanks!