关键词: 模糊聚类 神经网络 统计检验

Abstract:

In this paper, a fuzzy clustering neural network based on statistical tests （FCNN-ST） is presented to resolve the weaknesses of the current learning algorithm of fuzzy clustering neural networks （FCNN）, such as initial condition sensitiveness, inflexible attraction domain and unreasonable converge ence points. Statistical tests of the unimodal distribution model based on the T-square sample are introduced to gradually adjust the structure of netwo rks and determine the optimal number of dusters. The results of simulation demonstrate the effectiveness of FCNN-ST.

Key words: fuzzy clustering, neural network, statistical test