本文为美国德州理工大学（作者：CARRIE KNERR）的硕士论文，共61页。

本文研究了人工全连接递归神经网络中嵌入先验知识对非线性时间序列预测的影响，该网络利用反向传播方法进行训练。利用时间序列（如方波、Mackey Glass数据和心电图信号）对两种网络结构进行比较，以确定通过余弦振荡器嵌入更多信息时，网络的预测质量或训练能力是否得到改善。这种设计的可能好处是预测异常的心电图信号，即心脏活动的电测量，这种能力将允许医疗专业人员干预并可能防止异常的心电图信号。改进后的网络能够为Mackey Glass时间序列提供更准确的预测值和训练能力。

This thesis involves the investigation of the effect of prior knowledgeembedded in an artificial fully connected recurrent neural network for the predictionof non-linear time series. The networks utilize the back propagation method fortraining. Two network architectures are compared using time series such as thesquare wave, Mackey Glass data, and an ECG signal to determine if predictionquality or training ability are improved when more information through cosineoscillators are embedded in the network. The benefit of such an exercise may bethe prediction of abnormal ECG signals, which is an electrical measure of heartactivity. Such ability would allow medical professionals to intervene and possiblyprevent abnormal ECG signals. The improved network was able to provideincreased prediction value and training ability for the Mackey Glass time series.

1 引言

1.1 时间序列

1.2 人工神经网络

1.3 递归神经网络

1.4 最小二乘算法

1.5 学习准则

1.6 Mackey Glass方程

1.7 心电图数据

1.8 研究目标

2 研究背景

2.1 训练

2.2 Pearlmutter算法

2.3 最新应用

3 试验结果

3.1 方波

3.2 Mackey Glass

3.3 ECG信号

4 结论

下载英文原文地址：

http://page2.dfpan.com/fs/0l3cfj1222b17289166/ 

更多精彩文章请关注微信号：