Fuzzy Multiple Synapses Neural Network and Fuzzy Clustering

Abstract

Fuzzy multiple synapses neural network is proposed and the limitation of traditional Hopfield network, which handles the quadratic optimization problems, is overcome. First, fuzzy multiple synapses neuron and network are defined; and a hybrid neural network is given which combines multiple synapses neural network with Hopfield network. It may solve constrained optimization problems whose objective functions may not only include high-order form, but also may include logarithmic, sinusoidal and etc.. Second, it is applied to fuzzy clustering and a concrete hybrid neural network is derived. Experiments show that this method is very valid. Finally, conclusion and future work are given.