摘要
在数字复合正交神经网络的基础上提出一种模拟复合正交神经网络,并用于曲线重建。由于模拟神经网络采用连续学习算法,故网络学习收敛速度快。仿真结果表明,在单变量和多变量复杂函数曲线重建中,用模拟复合正交神经网络方法重建的曲线具有很高的逼近精度。本文提出的曲线重建方法是一种快速有效的方法。由于该模拟神经网络可望用模拟电路实现硬件化,因此在图象图形实时处理中具有很好的工程应用前景。
An analog compound orthogonal neural network was presented based on the digital compound orthogonal neural network and was applied in the curve reconstruction. By adopting continuously learning algorithm for the analog compound orthogonal neural network, the new network learning algorithm learns fast. The simulation results show that the reconstructed curve using the analog com- pound orthogonal neural network method has high precision in the curve reconstruction of complicated function with single-and multivariable. The reconstructed curve presented here is a rapid and effective method. It has excellent engineering application foreground in real-time operation of image and graphics because hardware can be realized with analog circuits.
出处
《机床与液压》
北大核心
2008年第2期25-26,93,共3页
Machine Tool & Hydraulics
基金
浙江省自然科学基金资助项目(M603070)
关键词
模拟复合正交神经网络
曲线重建
连续学习算法
Analog compound orthogonal neural network
Curve reconstruction
Continuously learning algorithm