摘要
针对径向基神经网络(RBF NN)的隐节点数难确定的问题,提出了一种基于相似矩阵确定RBF网络隐层节点数的新方法。采用基于输入输出全部信息的模糊相似矩阵的平均矩阵元法自适应确定隐节点数。建立自动确定隐节点数、中心、宽度、权值等网络参数的RBF网络降水预报模型,对广西5月的平均日降水量进行仿真实验,结果表明,该方法确定的隐节点数,网络规模较小,模型切实可行,在预报性能上明显优于同期的T213(中国气象局的全球中期天气数值预报产品预报值)降水预报,可为RBF神经网络隐节点数确定提供新思路,在降水预报上有一定的参考价值。
The number of hidden layer nodes in the radial basis function neural network is diffieuh Io identify. In order to solve the problem, this paper proposed a new method to determine the number of hidden layer nodes of RBF network based on the similarity marix, using the average matrix ele ment method of fuzzy similar matrix based or, all of the input and output information to adaptively de termine the number of hidden nodes. The establishment of automatic determination of the number of hidden nodes, RBF network prediction rnodel to automatically determine the center, width, weight and olher network parameters, the simulation experiment, the average daily precipitation of Guangxi in May showed that the method for determining the number of hidden nodes, when the network scale is small, the mode/ is feasible, is superior in the prediction performance of lhe T218global medium range numerical forecasting product forecast value from China Meteorological Bureau）precipitation forecast, which provides new ideas for the number of hidden nodes of RBF neural network, and has certain reference value in precipitation forecast.
出处
《广西师范学院学报（自然科学版）》
2013年第2期48-54,共7页
Journal of Guangxi Teachers Education University(Natural Science Edition)
基金
广西自然科学基金青年项目(#2012GXNSFBA053161)
广西教育厅项目(#201204LX501)
柳州师范高等专科学校项目(NO.LSZ2012B003)
关键词
相似矩阵
平均矩阵元
隐节点确定
径向基函数神经网络
降水预报
similarily matrix
average matrix element
hidden nodes determined
radial basisfunction neural nmwork
prediction of precipilation
