Is LVQ really good for classification?-an interesting alternative
Werner Poechmueller, Manfred Glesner, H. Juergs · 2002
Learning vector quantization (LVQ), developed by T. Kohonen (1989), is a neural network based method to find a good set of reference vectors to be stored as a nearest neighbor classifier's reference set. An efficient method of finding reference vectors along class boundaries instead of finding vectors representing class distribution, as LQV does, is described. A quantitative comparison with LVQ is given.>