摘要
RBF神经网络作为一种采用局部调节来执行函数映射的人工神经网络,在逼近能力、分类能力和学习速度等方面都有良好的表现,但由于RBF网络的隐节点的个数和隐节点的中心难以确定,从而影响了整个网络的精度,极大地制约了该网络的广泛应用.为此本文提出基于GEP优化的RBF神经网络算法,对其中心向量及连接权值进行优化.实验表明,本文所提算法比RBF算法的预测误差平均减少了48.96%.
RBF neural network,as an artificial neural network which adopts partial regulation to implement function mapping,specializes in approximation capability,classification capability and learning speed. But because of the difficulty of confirming the number of hidden nodes and the center of hidden nodes,it affects the accuracy of the entire network,which greatly restricted the use of the network RBF. This paper proposes Radial Basis Function Algorithms based on Gene Expression Programming （GEP-RBFA）,by taking advantage of the powerful global search ability of GEP to optimize the parameters of the center vectors and the weights in RBF Neural Network. Experiments show that the average error value of GEP-RBFA is reduced 48.96% in comparison with RBF algorithm.
出处
《小型微型计算机系统》
CSCD
北大核心
2010年第5期950-954,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(60763012)资助
广西自然科学基金项目(桂科自0731028)资助
广西高等学校优秀人才资助计划项目(RC2007022)资助
