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

DAI Xian-hua

您当前的位置：

首页 >

文章列表页 >

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

更新时间：2025-07-16

- A New Method for Training RNN via Hidden Representation Estimated by EM Algorithm
- Acta Electronica Sinica Vol. 28, Issue 10, Pages: 133-137(2000)
- 作者机构：
  
  1. 汕头大学电子工程系,汕头,515063
  2. 汕头大学电子工程系汕头,515063
- 作者简介：
- 基金信息：
- DOI：
  CLC： TP183
- Published：2000
- 稿件说明：
移动端阅览
DAI Xian-hua. A New Method for Training RNN via Hidden Representation Estimated by EM Algorithm[J]. Acta Electronica Sinica, 2000, 28(10): 133-137.
DOI：

DAI Xian-hua. A New Method for Training RNN via Hidden Representation Estimated by EM Algorithm[J]. Acta Electronica Sinica, 2000, 28(10): 133-137. DOI：

摘要

研究回馈神经网络(RNN)参数估计的新方法.利用隐含观测量

将复杂RNN的训练分解为线性输出层和多个单隐元的参数估计.基于每个隐元激励函数的多点线性近似

RNN可利用统计混合专家网络模型(ME)描述

从而将RNN的参数估计转化为包含隐含观测量的线性系统的最大似然估计问题

最后利用期望最大化(EM)算法获得RNN的隐含观测量及其参数估计.

Abstract

A new method for training recurrent neural network (RNN) has been proposed.By introducing the hidden representation or hidden variables into RNN

training the complicated RNN is decomposed into training a set of single neurons and a linear output layer.Based on linear approximation of RNN hidden units

RNN is remodeled with a "mixture of experts"(ME) model.Morever

training RNN is also changed into a maximum likelihood estimation of the linear systems with hidden variables.Finally

training RNN is fulfilled with the expectation-maximization (EM) algorithm.

关键词

Keywords

references

Views

1384

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Blind Estimation of Parameters in Gaussian Noise

The Resampling Method for Mixtures of Factor Analyzers

Nonlinear System Blind Deconvolution via Cumulants Extrema

Broadband DOA Estimation and Blind Source Separation Based on EM Algorithm

Auto Clustering Method Study of Flow Cytometry Data Based on Skew t-Mixture Models

Related Author

LI Zhi-shun

WANG Hui-gang

JIAO Li-cheng

YUE Bo

DAI Xian-hua

WANG Hua-xing

JIANG Wen-li

LIU Zheng

Related Institution

Marine Engineering College,Northwestern Polytechnic University

Key Laboratory for Radar Signal Processing,Xidian University

Unit 61541 of PLA Beijing China

School of Electronic Science and Engineering National University of Defense Technology Changsha Hunan China

Unit 61541 of PLA

⁰