| 双层BP人工神经网络连接权值的性质 |
| |
| 引用本文： | 孙效功,冯天瑾,陈佐林.双层BP人工神经网络连接权值的性质[J].中国海洋大学学报(自然科学版),1994(Z2). |
| |
| 作者姓名： | 孙效功 冯天瑾 陈佐林 |
| |
| 作者单位： | 河口海岸带研究所 |
| |
| 摘 要： | 从理论上分析和探讨具有Ｓ型连结函数的双层ＢＰ网络连接权值和阈值的性质，并进行实例验证，本文有助于神经网络内部行为的研究。
|
| 关 键 词： | 神经网络，连接权值，S型连结函数，双层BP网络 |
THE CHARACTERS OF BOTTOM-UP WEIGHTS OF DOUBLE-LAYERRED BP ARTIFICIAL NEURAL NETWORKS |
| |
| Sun Xiaogong, Feng Tianjin, Chen zuolin.THE CHARACTERS OF BOTTOM-UP WEIGHTS OF DOUBLE-LAYERRED BP ARTIFICIAL NEURAL NETWORKS[J].Periodical of Ocean University of China,1994(Z2). |
| |
| Authors: | Sun Xiaogong Feng Tianjin Chen zuolin |
| |
| Institution: | 1. Estuary and Seacoast Institute 2. Electrical Engineering Department |
| |
| Abstract: | The characters of bottom-up weights and biases of double-layerred BP artificial neural networks with S-shaped activation function are studied in theory and testified by the actual examples. The results show that the algebraic sum of the bottomup weights from each input node to all output nodes is zero and that generally the sum of all biases is not zero, which is helpful to studying the internal behaviour of artificial neural networks. |
| |
| Keywords: | Neural networks bottom--up weight S-shaped activation function double-layerred BP network |
| 本文献已被 CNKI 等数据库收录！ |
