首页> 外文会议>Third International Congress on Numerical Methods in Engineering and Applied Sciences (Cimenics'96) 25-29 March 1996 Merida, Venezuela >Sensibility to element degradation in feedforward neural networks and its application to network training
【24h】

Sensibility to element degradation in feedforward neural networks and its application to network training

机译:前馈神经网络中元素退化的敏感性及其在网络训练中的应用

获取原文

摘要

In this work, an algorithm for the approximate calculation of sensibilities to element degradation in feedforward neural networks has been developed. The algorithm has N times the complexity of the Back-Propagation training procedure, being N the number of network outputs; the results obtained from this algorithm are in accordance with the usral definition of sensibility. It has also been used in a minor modification to the training procedure, improving its element utilization, its fault toleranc,e and giving it a better interpolation and/or extrapolation property.
机译:在这项工作中,已经开发了一种近似计算前馈神经网络中元素降解敏感性的算法。该算法的复杂度是反向传播训练过程的N倍,是网络输出的N倍;从该算法获得的结果符合敏感性的一般定义。它也已用于训练过程的较小修改,以提高其元素利用率,容错能力,并使其具有更好的内插和/或外推特性。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

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