摘要
为了避免神经网络的收敛速度慢和局部极小点 ,采用统计学习理论中的支持向量机代替梯度下降法对三层神经网络中隐层到输出层的过程进行改进 .分别采用由支持向量机改进的神经网络和传统的神经网络对昆明市“一二一”大街交通交流的实时预测 。
In order to avoid slow convergence and local minimum of three-layered neural network, Supporting Vector Machines of statistical learning theory is used to modify this process from hidden-layer to output-layer instead of gradient descent method. The traffic flow at a crossing of Kunming City was respectively forecast by modified neural network and by classical neural network. And the advantages of modified neural network have been shown and proved by the results of experiment.
出处
《昆明理工大学学报（理工版）》
2004年第6期148-152,共5页
Journal of Kunming University of Science and Technology(Natural Science Edition)
基金
云南省教育厅自然科学基金项目资助 (项目编号 :0 2ZY0 11)
云南大学理 (工 )科校级科研项目资助 (项目编号 :2 0 0 2Q0 19SL)
云南省自然科学基金项目资助 (项目编号 :2 0 0 3E0 0 86M )
关键词
神经网络
支持向量机
函数逼近
结构风险最·J、化
交通流量
neural network
Supporting Vector Machines
function approximation
structural risk minimization
traffic flow
