Fuzzy learning vector quantization

In this paper, a new supervised competitive learning network model called fuzzy learning vector quantization (FLVQ) which incorporates fuzzy concepts into the learning vector quantization (LVQ) networks is proposed. Unlike the original algorithm, the FLVQ's learning algorithm is derived from optimizing an appropriate fuzzy objective function which takes into accounts of two goals, namely, minimizing the network output error which is the class membership differences of target and actual values and minimizing the distances between training patterns and competing neurons. As compared with the LVQ network, the proposed one consists of several distinctive features: i) stand-alone operation, i.e. employing preprocessing algorithm to obtain a good initial network state is not required; ii) superior classification performance, particularly in overlapping data sets; and iii) avoiding neuron underutilization. These advantages are demonstrated through an artificially generated data set and a vowel recognition data set.