Modular neural system, based on a fuzzy clustering network, for classification

Deals with the application of a modular neural network system to the classification of poor, low dimensional remote sensed data. The main objective is the reduction of the computational complexity of the neural learning stage, which is influenced by the characteristics of the training data. The classification task is decomposed in two phases. In the first phase, a fuzzy Kohonen network module is used for organizing training patterns into clusters. In the second phase, a feedforward network based on the backpropagation rule, is employed for labelling the clusters obtained in the first phase. The attention is focused on the effectiveness of the fuzzy network module, in applications where clusters touch or overlap. The performance of the modular system have been evaluated in comparison with those of a multilayer perceptron network (MLP). Experimental results have confirmed that the modular network system, supported by the fuzzy clustering module, improves the classification accuracy compared to the results obtained by the supervised MLP alone