A new error backpropagation learning algorithm for a layered neural network with nondifferentiable units

Electronics and Communications in Japan (Part III: Fundamental Electronic Science) Pub Date : 2007-05-01 DOI:10.1002/ECJC.20318

Hidenori Naganuma, T. Oohori, Kazuhisa Watanabe

{"title":"A new error backpropagation learning algorithm for a layered neural network with nondifferentiable units","authors":"Hidenori Naganuma, T. Oohori, Kazuhisa Watanabe","doi":"10.1002/ECJC.20318","DOIUrl":null,"url":null,"abstract":"This paper proposes a new error backpropagation method (DBP) for a three-layered neural network containing a nondifferentiable binary output unit. In contrast to the conventional simple perceptron, in which the teacher signal is given only to the output layer, in the DBP method the teacher signal is also given to the middle layer so that the output error is decreased. Consequently, it is possible in the DBP method to correct the coupling weights in both the lower layer and the upper layer. This makes it easy to construct a network composed only of binary output units, which results in high-speed operation and is suitable for hardware implementation. When the DBP method is applied to linearly inseparable tasks such as XORing, the learning performance is greatly improved compared to learning by the simple perceptron, and almost the same learning performance as the conventional BP is obtained. © 2007 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(5): 40– 49, 2007; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.20318","PeriodicalId":100407,"journal":{"name":"Electronics and Communications in Japan (Part III: Fundamental Electronic Science)","volume":"52 1","pages":"40-49"},"PeriodicalIF":0.0000,"publicationDate":"2007-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"3","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Electronics and Communications in Japan (Part III: Fundamental Electronic Science)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1002/ECJC.20318","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 3

Abstract

This paper proposes a new error backpropagation method (DBP) for a three-layered neural network containing a nondifferentiable binary output unit. In contrast to the conventional simple perceptron, in which the teacher signal is given only to the output layer, in the DBP method the teacher signal is also given to the middle layer so that the output error is decreased. Consequently, it is possible in the DBP method to correct the coupling weights in both the lower layer and the upper layer. This makes it easy to construct a network composed only of binary output units, which results in high-speed operation and is suitable for hardware implementation. When the DBP method is applied to linearly inseparable tasks such as XORing, the learning performance is greatly improved compared to learning by the simple perceptron, and almost the same learning performance as the conventional BP is obtained. © 2007 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(5): 40– 49, 2007; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.20318

查看原文本刊更多论文

一种新的不可微单元层状神经网络误差反向传播学习算法

针对含有不可微二进制输出单元的三层神经网络，提出了一种新的误差反向传播方法(DBP)。传统的简单感知器仅将教师信号提供给输出层，与之相反，DBP方法将教师信号也提供给中间层，从而降低了输出误差。因此，在DBP方法中，可以对下层和上层的耦合权值进行校正。这使得构建仅由二进制输出单元组成的网络变得容易，从而实现了高速运行，并且适合硬件实现。当DBP方法应用于XORing等线性不可分割的任务时，与简单感知器的学习相比，学习性能有了很大的提高，并且获得了与传统BP几乎相同的学习性能。©2007 Wiley期刊公司电子工程学报，2009,35 (5):444 - 444;在线发表于Wiley InterScience (www.interscience.wiley.com)。DOI 10.1002 / ecjc.20318

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Electronics and Communications in Japan (Part III: Fundamental Electronic Science)

自引率

0.00%

发文量