计算机工程 ›› 2010, Vol. 36 ›› Issue (06): 192-195.doi: 10.3969/j.issn.1000-3428.2010.06.065

• 人工智能及识别技术 • 上一篇    下一篇

基于参数调整的动态模糊神经网络算法

张德丰1,周 灵1,孙亚民2,马子龙3   

  1. (1. 佛山科学技术学院计算机系,佛山 528000;2. 南京理工大学计算机科学与技术学院,南京 210094;3. 哈尔滨工业大学电子工程系,哈尔滨150001)
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2010-03-20 发布日期:2010-03-20

Dynamic Fuzzy Neural Network Algorithm Based on Parameter Adjusting

ZHANG De-feng1, ZHOU Ling1, SUN Ya-min2, MA Zi-long3   

  1. (1. Department of Computer, Foshan University, Foshan 528000; 2. School of Computer and Technololgy, Nanjing University of Science and Technology, Nanjing 210094;3. Department of Electrical Engineering, Harbin Institute of Technology, Harbin 150001)
  • Received:1900-01-01 Revised:1900-01-01 Online:2010-03-20 Published:2010-03-20

摘要: 模糊逻辑与神经网络结合形成的模糊神经网络同时具有易于表达人类知识、存储与学习分布信息的优点,基于此,提出一种基于参数调整的动态模糊神经网络算法。采用扩展卡尔曼滤波器法将全局算法划分为线性和非线性部分,线性参数由最小二乘法和滤波器法决定,非线性参数由训练样本和启发式法直接决定,线性和非线性参数可进行实时更新。仿真结果表明,该算法能保证更简洁的结构和更短的学习时间。

关键词: 动态模糊神经网络, 模糊逻辑, 扩展卡尔曼滤波器

Abstract: The union of fuzzy logic and the neural network forms the fuzzy neural network, it simultaneously has the advantages of expressing the human knowledge, storing and learning distribution information storage. This paper proposes a Dynamic Fuzzy-Neural Network(D-FNN) algorithm based on parameter adjusting. It uses Extended Kalman Filter(EKF) method to divide the overall algorithm into the linearity and the misalignment part. The linear parameter is decided by Least Squares(LS) method and the filter method, the misalignment parameter is directly decided by the training sample and heuristic method, the linearity and the misalignment parameter can carry on the real-time renewal. Simulation results indicate that this algorithm can guarantee more succinct structure and shorter learning time.

Key words: Dynamic Fuzzy-Neural Network(D-FNN), fuzzy logic, Extended Kalman Filter(EKF)

中图分类号: