Magnification Control in Neural Gas by Winner Relaxing Learning

Abstract.
We transfer the idea of winner relaxing learning from the self-organizing map to the neural gas to enable magnification control independently of the shape of the data distribution. The entropy of the resulting map and the influence of a diagonal term is studied numerically.
Jens Christian Claussen and Thomas Villmann

6.1.2003. 4 pages LaTeX with 2 figures included. Submitted to Proc. ICANN 2003
This paper is not (yet) available online. If you are interested, please do not hesitate to ask the authors (claussen, villmann) for a printed copy.