用模糊逻辑方法优化多层前馈网络结构

OPTIMUM DESIGN OF MLP NETWORK STRUCTURE USING FUZZY LOGIC APPROACH

为了优化用于故障分类的多层前馈网络结构配置，提出了一种基于模糊逻辑的优化方法。通过对网络的学习误差、学习时间、灵敏度和网络单元数的模糊化处理，确定了优化网络结构的代价函数，并由此建立起优化网络结构的模糊推理规则，然后应用这些推理规则给出了优化网络结构的框图。优化结果表明，利用模糊逻辑方法优化用于故障分类的多层前馈网络结构是有效的。

In order to optimize the design of MLP network structure for fault classification, an optimum approach based on fuzzy logic is presented.The learning error,the learning time,the sensitivity and the unit number of MLP network are fuzzified.A evaluation function of the network structure is determined, and the fuzzy reasoning rule of the optimum design for the network structure is built.Then the outline flowchart of the optimized network structure based on the reasoning rule is given.The optimum results show that the optimum design of the MLP network structure for the fault classification using the fuzzy logic approach is efficient and the iteration time is short.