一种基于蚁群聚类的径向基神经网络  

Radial Basis Function Neural Network Based on Ant Colony Clustering

作　　者：武方方[1] 赵银亮[1] 

机构地区：[1]西安交通大学新型计算机研究所,西安710049

出　　处：《西安交通大学学报》2006年第4期386-389,共4页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(60173066)

摘　　要：提出了一种基于蚁群聚类算法的径向基神经网络.利用蚁群算法的并行寻优特征和挥发系数方法的自适应更改信息量的能力,并以球面聚类的方式确定了径向基神经网络中基函数的位置,同时通过比较隐层神经元的相似性、合并相似性较为接近的2个神经元来约简隐含层的神经元,以达到简化径向基神经网络结构的目的.实验比较了几种不同聚类算法的径向基神经网络,结果表明,所提神经网络的整体训练时间至少可缩短40%,学习的准确率可提高1%以上,而且网络结构更加精简.A novel radial basis function neural network based on ant colony clustering （RBF ACC） was proposed. Based on the ant colony algorithm and its feature of parallel search optimum, the center of each basis function of RBF can be defined by using the sphere clustering and the dynamic method to adjust the parameter of evaporation coefficient. Meanwhile, in order to simplify the structure of RBF network, the two neurons that are most similar each other are combined to form a single neuron. Experiment results show that compared with the other RBF neural networks in different clustering algorithms, the accuracy of RBF-ACC increases by more than 1%, the total training time of RBF network can be reduced at least 40%, and the structure of this novel RBF network is simpler due to the reduction of the hidden layers.

关 键 词：径向基神经网络 蚁群聚类算法 基函数 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]