Learning Vector Quantization with Training Count (LVQTC)

Kohonen's learning vector quantization (LVQ) is modified by attributing training counters to each neuron, which record its training statistics. During training, this allows for dynamic self-allocation of the neurons to classes. In the classification stage training counters provide an estimate of the reliability of classification of the single neurons, which can be exploited to obtain a substantially higher purity of classification. The method turns out to be especially valuable in the presence of considerable overlaps among class distributions in the pattern space. The results of a typical application to high energy elementary particle physics are discussed in detail. Copyright 1997 Elsevier Science Ltd.