Vahid Haghighatdoost - Computational Vision & Intelligence Laboratory Computer Engineering Department Amirkabir University of Technology, Tehran, Iran
Reza Safabakhsh

This paper presents a new Kohonen Clustering Network algorithm. Unlike the well-known method of Kohonen, the new algorithm considers fuzzification of the winning neuron parameters. It consists of a modified version of the Fuzzy Kohonen Clustering Network functional, where an inhibitory term is added to avoid blind updating rule. This term is actually a membership value, giving each neuron its own chance to be the winner. Using this factor results in neurons which are topologically located in the neighbor of the winning neuron in the map, but their synaptic weights' distance is far from the input, don’t update fearlessly. In the proposed method, the updating rule is changed to avoid producing dead neurons in the free space between training samples. Results of experiments show the good performance of the proposed method

Neural network; Fuzzy; Kohonen Clustering Network; Selforganizing maps; Fuzzy clustering; Fuzzy c-means; FuzzyKohonen Clustering Network