Decomposition of Learning of Complex Neural Network into a Set of Learningof the Simple Single Neuron via the Estimation of Hidden Observed Data

Dai Xianhua

您当前的位置：

首页 >

文章列表页 >

Decomposition of Learning of Complex Neural Network into a Set of Learningof the Simple Single Neuron via the Estimation of Hidden Observed Data

更新时间：2025-12-08

- Decomposition of Learning of Complex Neural Network into a Set of Learningof the Simple Single Neuron via the Estimation of Hidden Observed Data
- Acta Electronica Sinica Issue 7, (1999)
- 作者机构：
  
  汕头大学电子工程系
- 作者简介：
- 基金信息：
- DOI：
  CLC： TP18
- Published：1999
- 稿件说明：
移动端阅览
[1]戴宪华.利用隐含观测量估计将复杂神经网络分解为多个单隐元学习的新方法[J].电子学报,1999(07):60-63.

Dai Xianhua. Decomposition of Learning of Complex Neural Network into a Set of Learningof the Simple Single Neuron via the Estimation of Hidden Observed Data[J]. Acta Electronica Sinica, 1999, (7).
[1]戴宪华.利用隐含观测量估计将复杂神经网络分解为多个单隐元学习的新方法[J].电子学报,1999(07):60-63. DOI：

Dai Xianhua. Decomposition of Learning of Complex Neural Network into a Set of Learningof the Simple Single Neuron via the Estimation of Hidden Observed Data[J]. Acta Electronica Sinica, 1999, (7). DOI：

摘要

本文从统计学的角度研究多层多隐元前馈神经网络（ＮＮ）的参数估计学习问题．利用ＮＮ激励函数的折线线性近似，提出一种求解多隐层多隐元ＮＮ每个隐元指导信号（隐含观测量）的新方法，利用每个隐元的指导信号估计可以将多隐层多隐元ＮＮ的参数估计学习转化为多个相互独立的单隐元ＮＮ参数估计学习训练问题．从而将复杂系统参数估计问题转化为简单系统的参数估计问题而得以解决

Abstract

The paper investigates the learning of the complex neural network （NN） from the point of the statistics.With the help of linear approximation of the activation function

a new method to estimate the supervised signal （hidden or missing data） of each hidden unit of the complex NN is proposed.By virtue of the supervised signal estimation of hidden unit

the learning of the complex NN is decomposed into the learning of a set of single neural neurons

and the complicated problem of NN learning is solved through the simple learning of a set of single neural neurons.

关键词

Keywords

references

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Nonlinear System Blind Deconvolution via Cumulants Extrema

A New Method for Training RNN via Hidden Representation Estimated by EM Algorithm

Related Author

DAI Xian-hua

Related Institution

Dept.of Electronic Eng.Shantou UniversityShantou 515063China

⁰