首页> 外国专利> method for training a neural network procedure for classification of a sequence of input parameter using a neural networkneural network, a neural network to train and arrangement

method for training a neural network procedure for classification of a sequence of input parameter using a neural networkneural network, a neural network to train and arrangement

机译:神经网络的训练神经网络过程的方法,用于对输入参数序列进行分类神经网络,训练和布置的神经网络

摘要

The neural network is trained for a first period in such a way that a discrimination value is maximized, wherein the discrimination value is dependent on impulses formed by the pulsed neurons during the first period. The first period is shortened in an iterative manner and a second discrimination value is formed for the second period until the second discrimination value is smaller than the maximum discrimination value. The trained neural network is the neural network of the last iteration in which the second discrimination value equals the maximum discrimination value.
机译:在第一时间段内训练神经网络,以使得鉴别值最大化,其中,鉴别值取决于在第一时间段内由脉冲神经元形成的脉冲。第一周期以迭代方式缩短,并且在第二周期中形成第二鉴别值,直到第二鉴别值小于最大鉴别值。训练后的神经网络是最后一次迭代的神经网络,其中第二辨别值等于最大辨别值。

著录项

  • 公开/公告号DE59908130D1

    专利类型

  • 公开/公告日2004-01-29

    原文格式PDF

  • 申请/专利权人 SIEMENS AG;

    申请/专利号DE19995008130T

  • 发明设计人 DECO GUSTAVO;SCHUERMANN BERND;

    申请日1999-07-01

  • 分类号G06N3/04;

  • 国家 DE

  • 入库时间 2022-08-21 22:41:57

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号