基于跨越连接的多层前馈神经网络结构分析

doi:10.3778/j.issn.1002-8331.2009.32.014

计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (32): 45-47.DOI: 10.3778/j.issn.1002-8331.2009.32.014

基于跨越连接的多层前馈神经网络结构分析

唐云岚^1，2，高妍方¹，谭旭¹，陈英武¹

1.国防科学技术大学信息系统与管理学院，长沙 410073
2.武警工程学院通信工程系，西安 710086

收稿日期:2008-06-27 修回日期:2008-10-09 出版日期:2009-11-11 发布日期:2009-11-11
通讯作者: 唐云岚

Research on structure of multi-layer feed-forward neural network with cross connections

TANG Yun-lan^1，2，GAO Yan-fang¹，TAN Xu¹，CHEN Ying-wu¹

1.Institute of Information System and Management，National University of Defense Technology，Changsha 410073，China
2.Department of Communication Engineering，Engineering College of the Chinese People’s Armed Police Force，Xi’an 710086，China

Received:2008-06-27 Revised:2008-10-09 Online:2009-11-11 Published:2009-11-11
Contact: TANG Yun-lan

摘要/Abstract

摘要： 任何连接方式的神经网络总可以归结为跨越连接网络。在传统多层前馈神经网络算法的基础上，提出了完全全连接神经网络的概念，给出了基于跨越连接的多层前馈神经网络算法。通过分析多层前馈神经网络的误差函数，从理论上证明了：相对于无跨越连接网络，基于跨越连接的多层前馈神经网络能以更加简洁的结构逼近理想状态。最后，用一个隐层神经元解决了XOR问题。

关键词: 跨越连接, 多层前馈神经网络, 隐层结构, XOR问题

Abstract: Neural networks with any kind of connections can always be sorted as cross-connected ones.According to traditional multi-layer feed-forward neural network，this paper elaborates the concept of completely-fully connected neural network and then puts forward a cross-connected multi-layer feed-forward neural network algorithm.By analyzing the error function of multi-layer feed-forward neural network，it can be theoretical proved that the cross-connected neural network can reach ideal results with more concise framework compared with the non-cross connected neural network.Lastly，the proposed algorithm is tested on the XOR problem，which is well solved by using only one hidden neuron.

Key words: cross connections, multi-layer feed-forward neural network, structure of hidden layer, XOR problem

中图分类号:

TP13

唐云岚^1，2，高妍方¹，谭旭¹，陈英武¹. 基于跨越连接的多层前馈神经网络结构分析[J]. 计算机工程与应用, 2009, 45(32): 45-47.

TANG Yun-lan^1，2，GAO Yan-fang¹，TAN Xu¹，CHEN Ying-wu¹. Research on structure of multi-layer feed-forward neural network with cross connections[J]. Computer Engineering and Applications, 2009, 45(32): 45-47.

[1]	刘自鑫^1，2，吕恕¹，钟守铭¹，叶茂³. 网络化系统控制器设计的矩阵分解方法[J]. 计算机工程与应用, 2010, 46(4): 18-21.
[2]	刘自鑫^1，2，吕恕¹，钟守铭¹. 离散时滞网络化系统H_∞控制[J]. 计算机工程与应用, 2010, 46(1): 18-20.
[3]	陈健，张持健. 三级倒立摆的LQR方法优化参数控制[J]. 计算机工程与应用, 2009, 45(29): 245-248.
[4]	胡爱凤，黎明，李军华. 噪声环境下遗传算法的研究[J]. 计算机工程与应用, 2009, 45(27): 46-49.

基于跨越连接的多层前馈神经网络结构分析

Research on structure of multi-layer feed-forward neural network with cross connections

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 4

编辑推荐

Metrics