期刊文献+

激励函数可调的NARX神经网络 被引量:1

NARX Networks with Trainable Activation Functions
下载PDF
导出
摘要 提出了一种隐层神经元激励函数可调的具有外部输入的非线性回归(NARX)神经网络,它在进行权值调整的同时,还对各隐层神经元激励函数的参数进行自适应调节;并推导出激励函数参数的学习算法,从而使NARX神经网络更符合生物神经网络.通过系统辨识的仿真实例,说明了隐层神经元激励函数对网络性能的影响,还验证了文中提出的NARX神经网络具有更快的收敛速度,并且能有效地避免算法陷入局部最小. The paper proposes a new kind of NARX network, which has trainable activation functions in addition to only trainable weights in the conventional NARX networks. A learning algorithm based on first-order gradient descent was used to adjust the parameters of activation functions. The simulation results show that the shape of activation functions has large effect on the performs of NARX networks, and the new NARX network has a faster convergence rate.
作者 李明 杨成梧
出处 《江南大学学报(自然科学版)》 CAS 2006年第4期445-448,共4页 Joural of Jiangnan University (Natural Science Edition) 
关键词 具有外部输入的非线性回归神经网络 激励函数 学习算法 NARX network activation function learning algorithm
  • 引文网络
  • 相关文献

参考文献12

  • 1Siegelmann H T,Home B G,Giles C L.Computational capabilities of recurrent NARX neural networks[J].IEEE Transactions on Systems,Man and Cybernetics-Part B,1997,27(2):208-215. 被引量:1
  • 2Chen S,Billings S,Grang P.Non-linear system identification using neural network[J].Int J Control,1990,51(6):1191-1214. 被引量:1
  • 3Su H T,Mcavoy T.Identification of chemical processes using recurrent networks[J].American Controls Conf,1991,3:2314-2319. 被引量:1
  • 4Widrow B,Plett G L.Nonlinear adaptive inverse control[C]/ / Proceeding of the 36th IEEE Conference on Decision and Control.San Diego:[s.n.],1997:1032-1037. 被引量:1
  • 5Plett G L,Bottrich H.DDEKF learning for fast nonlinear adaptive inverse control[C].Proceeding of the 2002 International Joint Conference.Honolulu:[s.n.],2002:2092-2097. 被引量:1
  • 6Liu Yaqiu,Ma Guangfu,Jiang Xueyuan.A design method for adaptive inverse control using NARX neural networks[C]/ / Proceeding of the 5th World Congress on Intelligent Control and Automation(WCICA).Hangzhou:Zhejiang University,2004:459-463. 被引量:1
  • 7Wlodzislaw D,Norvbert J.Survey of neural transfer functions[J].Neural Computing Surveys,1999,2:163-212. 被引量:1
  • 8吴佑寿,赵明生,丁晓青.一种激励函数可调的新人工神经网络及应用[J].中国科学(E辑),1997,27(1):55-60. 被引量:26
  • 9Sulee G,Danilo M P.Recurrent neural networks with trainable amplitude of activation functions[J].Neural Networks,2003(16):1095-1100. 被引量:1
  • 10Wu Youshou,Zhao Mingsheng,Ding Xiaoqing.Beyond weights adaptation:a new neuron model with trainable activation function and its supervised learning[J].IEEE International Conference on Neural Networks,1997,2:1152-1157. 被引量:1

二级参考文献3

  • 1吴佑寿,中国科学.E,1996年,26卷,2期,140页 被引量:1
  • 2吴佑寿,Proceeding of ICONIP’95,1995年,1期 被引量:1
  • 3Chen Q C,Neural Networks,1994年,7卷,1477页 被引量:1

共引文献25

同被引文献9

引证文献1

相关主题

;
使用帮助 返回顶部