Choosing the Weights of Connecting Input Neurons and Hidden Neurons to Reduce the Numbers of Hidden Neurons

摘要：

In most discussions about the learning capability of multi-layer feedforward networks, the weights connecting the input neurons and the hidden neurons can be chosen almost arbitrarily. It means that their functions have not been developed sufficiently yet. This paper discusses the possibility and method of reducing the number of the hidden layer neurons by suitable choosing the weights connecting the input neurons and the hidden neurons. Some conditions upon input samples are given.

作者: Da-Gang Tian

作者单位: College of Management University of Shanghai Science and Technology Shanghai, 516 Jun Gong Road, P.R. China 200093

会议类型: 国际会议

会议名称: 第三届国际脉冲动力系统及应用学术会议

会议地点: 青岛

会议语种:英文

在线出版日期: 2006-07-21（万方平台首次上网日期，不代表论文的发表时间）