HUANG Chaohui,WEN Hui*,CHE Yan.Research on improved RBF network algorithm based onpotential function clustering[J].Journal of Yanbian University,2020,46(02):145-149.
基于势函数聚类的改进RBF网络算法研究
- Title:
- Research on improved RBF network algorithm based on potential function clustering
- 文章编号:
- 1004-4353(2020)02-0145-05
- 分类号:
- TP183
- 文献标志码:
- A
- 摘要:
- 为优化径向基函数(RBF)网络结构并改善网络泛化性能,提出了一种基于势函数聚类的改进RBF网络算法.首先使用势函数统计每个模式类别中的样本势值,以此实现样本空间中不同样本的势值度量; 其次以增量学习的方式逐次完成对样本空间的有效覆盖,以此实现网络隐节点个数及参数的自动有效估计.最后将本文算法与KMRBF、FCRBF、MRAN以及 GAP -RBF 学习算法进行了实验对比,结果表明本文算法的网络分类精度更高,克服了KMRBF和FCRBF算法需人工调整网络隐节点来提高分类精度的问题,且比 GAP -RBF和MRAN算法的网络结构更加简单.
- Abstract:
- To optimize the structure of radial basis function(RBF)network and improve its generalization performance, an improved RBF network based on potential function clustering is proposed. Firstly, the potential function is used to count the potential values of samples in each pattern category, so as to measure the potential values of different samples in the sample space. Secondly, the incremental learning method is used to cover the sample space step by step, so as to complete the automatic and effective estimation of the number of hidden nodes and parameters in the network. Finally, the presented algorithm is compared with KMRBF, FCRBF, MRAN and GAP -RBF learning algorithm by experiments. The results and experiments show that the classification accuracy of thealgorithm is higher than that of KMRBF, FCRBF, GAP -RBF and MRAN, which overcomes the problem that KMRBF and FCRBF algorithm need many experiments to adjust the hidden nodes manually to obtain higher classification accuracy, and is more simple than the GAP -RBF and MRAN network structure.
参考文献/References:
[1] MOODY J, DARKEN C J.Fast learning in networks of locally -tuned processing[J].Neural Computation, 1989,1(2):281-294.
[2] 张爱科,符保龙,李辉.基于改进的模糊聚类RBF网络集成的文本分类方法[J].四川大学学报(自然科学版),2012,49(6):1235-1239.
[3] RAITOHARJU J, KIRANYAZ S, GABBOUJ M.Training radial basis function neural networks for classification via class -specific clustering[J].IEEE Transactions on Neural Networks and Learning Systems, 2016,27(12):2458-2471.
[4] LU Y W, SUNDARARAJAN N, SARATCHANDRAN P.A sequential learning scheme for function approximation using minimal radial basis function[J].Neural Computation, 1997,9(2):461-478.
[5] HUANG G B, SARATCHANDRAN P, SUNDA-RARAJAN N.An efficient sequential learning algorithm for growing and pruning RBF(GAP -RBF)networks[J].IEEE Transactions on Systems, Man and Cybernetics.Part B: Cybernetics, 2004,34(6):2284-2292.
[6] HUANG G B, SARATCHANDRAN P, SUNDA-RARAJAN N.A generalized growing and pruning RBF(GAP -RBF)neural network for function approximation[J].IEEE Transactions on Neural Networks, 2005,16(1):57-67.
[7] 乔俊飞,韩红桂.RBF神经网络的结构动态优化设计[J].自动化学报,2010,36(6):865-872.
[8] 韩红桂,乔俊飞,薄迎春.基于信息强度的RBF神经网络结构设计研究[J].自动化学报,2012,38(7):1083-1090.
[9] BLAKE C, MERZ C. UCI repository of machine learning databases[EB/OL].[2019-06-03].http://archive.ics.uci.edu/ml/35.
[10] HAYKIN S. Neural Networks and Learning Machines[M].Third Edition.Beijing: China Machine Press, 2009:61-63.
备注/Memo
收稿日期: 2019-12-25 *通信作者: 闻辉(1981—),男,博士,副教授,研究方向为机器学习、神经网络.
基金项目: 福建省自然科学基金资助项目(2019J01815); 福建省教育厅中青年教师教育科研项目(JT180486); 莆田市科技局项目(2018RP4004,2018ZP10)