摘要
小波神经网络参数初始值影响着网络收敛速度的快慢,甚至关系到网络能否收敛。为了减少网络训练次数,提高收敛速度,提出了一种更简便易行的选择方法,通过将此方法的仿真结果与采用随机选取初始值的方法所得仿真结果进行对比,证明此方法既可行又有效。
The initial values of wavelet neural network will influence the network convergence speed,even more whether the network can accomplish convergence.In order to reduce the training times of network and improve the convergence speed,the selection of the initial values of neural wavelet is proposed.The simulation results obtained by this method are compared to those obtained by selecting random initial values,which shows the proposed method is feasible and effective.
出处
《电脑开发与应用》
2005年第2期37-38,41,共3页
Computer Development & Applications
基金
太原理工大学科技基金资助项目(编号:予内190101867)。
关键词
小波神经网络
初始值
收敛速度
仿真研究
wavelet neural network,initial value,convergence,convergence speed