基于集群智能优化的模糊神经网络设计与应用  

Fuzzy Neural Networks Design Methods Based on Swarm Intelligent Optimization Algorithms and Its Application

作　　者：王永海[1] 郭珂[1] 方岳[1] 叶玉玲 Wang Yonghai;Guo Ke;Fang Yue;Ye Yuling(Beijing Institute of Space Long March Vehicle,Beijing 100076,China;No.710 Institute of China Shipbuilding Industry Corporation,Yichang 443003,China)

机构地区：[1]北京航天长征飞行器研究所,北京100076 [2]中船重工集团第七一〇研究所,湖北宜昌443003

出　　处：《航空兵器》2021年第1期87-92,共6页Aero Weaponry

摘　　要：在分析模糊神经网络(FNN)的结构和参数的基础上,通过引入一个布尔变量作为网络的结构参数,进而把FNN的设计问题转化为一个多变量函数优化问题,提出了一种混合集群智能优化算法,给出了其二进制编码(BIOA)和实数编码(RIOA)的实现形式。分别用BIOA和RIOA对FNN的结构参数和前件参数进行协同进化,得到最优前件参数后,采用最小二乘法计算结论参数。采用此方法设计FNN,对太阳黑子数目进行建模。结果表明,所设计的FNN不仅结构简化,而且精度和泛化能力也得到提高。The structure and parameters of fuzzy neural networks(FNN)are analyzed and a Boolean variable is proposed to network as the structure parameter.Then the question of FNN design is transformed to a function optimization question with multi-parameters.A new hybrid swarm intelligent optimization algorithm is proposed,and its binary-coded form(BIOA)and real-coded form(RIOA)are presented.BIOA and RIOA are applied to cooperative optimize the structure parameters and premise parameters of the FNN.The conclusion parameters of the FNN are optimized by least square error algorithm after the premise parameters are obtained.In the experiment,the number of sunspots is modeled by FNN and the results show that the designed FNN not only has simpler structure,but has higher precision and generalization.

关 键 词：模糊神经网络 混合集群智能优化算法 协同进化 最小二乘法 网络结构 

分 类 号：TJ760[兵器科学与技术—武器系统与运用工程] TP18[自动化与计算机技术—控制理论与控制工程]