A modified fuzzy clustering based on multisynapse neural network
Kai Li, Lijuan Cui, Yu-Fen Zhang · 2005
Traditional Hopfield neural networks has ability of optimization computation, image segmentation, etc. However, there exist some problems in this network, i.e. it can only solve linear or quadratic optimal problems. So, Wei and Fahn proposed a new neural architecture, the multisynapse neural network, to solve optimization problems including high-order, logarithmic, sinusoidal forms, etc. As one of its major applications, a fuzzy bidirectional associative clustering network (FBACN) is presented for fuzzy clustering according to the objective functional method. In this paper, first, FBACN is analyzed in detail in theory and some drawbacks is pointed. Then we present a modified FBACN, named as MFBACN, by using expended Lagrange multipliers method. Moreover, we also propose a method of determining Lagrange multipliers. Finally we conduct the experiments with three datasets. The experimental results show that the convergence of MFBACN holds and it is an effective method.