• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>Annual Computational Neuroscience Meeting(CNS'02); 20020721-20020725; Chicago,IL; US >A novel approach to training neurons with biological plausibility
【24h】

A novel approach to training neurons with biological plausibility

机译：一种具有生物学可行性的神经元训练新方法

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

The dominant approach to training artificial neurons is to search iteratively for single numeric values that satisfy the training conditions based on error. A novel method of fully training neurons that will generalize has been developed that does not find single numeric values for weights, but instead finds regions in the weight-space that satisfy the training conditions. This method provides an alternative to the biologically implausible backpropagation for training feedforward neural networks, as it allows the McCulloch-Pitt neuron model to be used as a basis for a fast training algorithm that dynamically allocates neurons into the network.
机译：训练人工神经元的主要方法是迭代搜索基于误差的满足训练条件的单个数值。已经开发出一种将对神经元进行全面训练的新颖方法，该方法不会找到权重的单个数值，而是会找到权重空间中满足训练条件的区域。这种方法为训练前馈神经网络提供了生物学上难以置信的反向传播方法，因为它允许将McCulloch-Pitt神经元模型用作快速训练算法的基础，该算法将神经元动态分配到网络中。

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