基于连续沃尔什变换的过程神经元网络训练  

Training of procedure neural network based on continuous Walsh conversion

作　　者：李盼池[1] 许少华[1] 

机构地区：[1]大庆石油学院计算机科学与工程学院,黑龙江大庆163318

出　　处：《计算机工程与设计》2005年第3期702-703,713,共3页Computer Engineering and Design

摘　　要：为解决过程神经元网络训练涉及的时域聚合运算问题,提出了过程神经元网络的一种学习算法。对网络的过程式输入函数及时变的权函数,实施连续沃尔什变换,用变换后的数据训练网络。经过这种变换,能使网络输入的时域聚合运算大为简化,有效地避免了复杂的积分过程,使过程神经元网络的训练等同于普通网络的训练。Aiming at procedure neural network( PNN) 's aggregation in time field, a training method of PNN is shown. First, both input function of network and weight function of hide layer are converted to walsh base function series, and then network is trained by coefficient of walsh base function. Training of PNN would be simplified in a large extent by the method and complex integral is avoided. It is come true to turn training of PNN into training of general network.

关 键 词：过程神经元 网络训练 学习算法 输入 运算 变换 数据 连续 积分 权函数 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP183[自动化与计算机技术—控制科学与工程]