摘要
为提高广义径向基神经网络代理模型的计算精度以及减少这种网络的计算量,提出并构建非径向对称核函数的广义径向基神经网络。采用径向对称的高斯核函数以及非径向对称的核函数对测试函数进行代理模型验证,从训练以及测试网络所需的时间、广义网络的隐层节点数、相对误差以及均方根误差等方面对代理模型进行评价,实验结果表明这种非径向对称的广义径向基神经网络的代理模型具有计算精度高、所需网络节点少、计算比较的次数少等优点。
To improve the calculating accuracy of the agent model based on generalised radial basis function neural networks as well as to reduce its computational complexity,we propose and construct the generalised radial basis function neural networks with non-radial symmetric kernel function.The radial symmetric Gaussian kernel functions and the non-radial symmetric kernel function are used to validate the text function,and the agent model is evaluated in terms of the required time for training and testing network,the node numbers in hidden layer of generalised network,the relative error and the root-mean-square error.Experimental results indicate that the agent model based on non-radial symmetric generalised radial basis function neural networks has many advantages such as high computation accuracy and less required network nodes,and less calculating and comparing time.
出处
《计算机应用与软件》
CSCD
北大核心
2013年第5期171-173,共3页
Computer Applications and Software
基金
河南省自然科学基金项目(122300410310)
关键词
非径向对称
径向基函数
神经网络
代理模型
广义网络结构
Non-radial symmetry Radial basis function(RBF) Neural networks Agent model Generalised network structure