基于SVD_TLS及EKF算法的动态自组织模糊神经网络  

Application of Dynamic Self-organizing Fuzzy Neural Network in Dynamic System Based on SVD-TLS and EKF Algorithm

作　　者：李云[1] 

机构地区：[1]宿迁高等师范学校,江苏宿迁223800

出　　处：《成都大学学报（自然科学版）》2012年第2期154-157,共4页Journal of Chengdu University（Natural Science Edition）

摘　　要：提出了将奇异值分解总体最小二乘法(SVD_TLS)及扩展卡尔曼滤波(EKF)相结合的动态自组织模糊神经网络.首先给出了STD_DSFNN的结构及各层的含义;其次,用EKF算法学习非线性参数,SVD_TLS算法学习线性参数的同时提取重要模糊规则;最后,通过典型的Machey-Glass时间序列预测实例验证SVD_TLS及EKF相结合的动态自组织模糊神经网络(STE_DSFNN),同时与DFNN、ANFIS及UKF_DFNN相对比,结果表明STE_DSFNN网络结构更紧凑,具有更好的泛化能力.A dynamic self-organizing fuzzy neural network was proposed with the combination of SVD_ TLS and EKF. Firstly, the structure and meanings of each layer were given. Secondly, important fuzzy rules were extracted when using EKF algorithm and SVD _ TLS algorithm to learn nonlinear and linear parameters respectively. Finally, the STE _ DSFNN was verified through the typical Machey-Glass time series prediction examples, which was compared with the DFNN, ANFIS and UKF _ DFNN at the same time. The results show that the STE_ DSFNN network structure is more compact and has better generalization ability.

关 键 词：奇异值分解_总体最小二乘法(SVD_TLS) 扩展卡尔曼滤波(EKF) Machey-Glass时间序列预测 

分 类 号：TP273.4[自动化与计算机技术—检测技术与自动化装置]