Supervised IAFC neural network based on the fuzzification of learning vector quantization

Abstract

In this paper, a fuzzy LVQ(Learning Vector Quantization) is proposed which is based on the fuzzification of LVQ. The proposed FLVQ(Fuzzy Learning Vector Quantization) uses the different learning rate depending on the correctness of classification. When the classification is correct, the amount of update is determined by consideration of location of the input vector relative to the decision boundary. When the classification is not correct, the amount of update is determined by the degree of belongingness of the input vector to the winning class. The supervised IAFC(Integrated Adaptive Fuzzy Clustering) neural network 3, which uses FLVQ, is introduced in this paper. The supervised IAFC neural network 3 is both stable and plastic because it uses the control structure which is similar to that of Adaptive Resonance Theory(ART)-1 neural network. We used iris data set to compare the performance of the supervised IAFC neural network 3 with those of LVQ algorithm and backpropagation neural network. The supervised IAFC neural network 3 yielded fewer misclassifications than LVQ algorithm and backpropagation neural network. © Springer-Verlag Berlin Heidelberg 2006.