摘要
文章根据隐节点对整个网络输出贡献的相对大小 ,提出删除策略 ,并结合资源分配网络的增长规则 ,使得径向基函数神经网络的隐节点在学习过程中可以自适应地增加或删除 ,从而形成一个网络资源较少、结构紧凑的自适应径向基函数神经网络。将该网络应用于函数拟合和非线性时间序列预测 。
In this paper, an adaptive radial basis function neural network is presented. A prune strategy based on the relative contribution of each hidden unit to the overall network output is proposed, and the growth criterion of the resource allocation network is adopted so that the hidden units can be added or deleted during learning process. When this adaptive network is applied to function mapping and non-linear time series prediction, a good result is obtained.
出处
《合肥工业大学学报（自然科学版）》
CAS
CSCD
2001年第2期244-247,共4页
Journal of Hefei University of Technology：Natural Science
