A Parsimonious Radial Basis Function-Based Neural Network for Data Classification
The radial basis function neural network trained with a dynamic decay adjustment (known as RBFNDDA) algorithm exhibits a greedy insertion behavior as a result of recruiting many hidden nodes for encoding information during its training process. In this chapter, a new variant RBFNDDA is proposed to rectify such deficiency. Specifically, the hidden nodes of RBFNDDA are re-organized through the supervised Fuzzy ARTMAP (FAM) classifier, and the parameters of these nodes are adapted using the Harmonic Means (HM) algorithm. The performance of the proposed model is evaluated empirically using three benchmark data sets. The results indicate that the proposed model is able to produce a compact network structure and, at the same time, to provide high classification performances.
KeywordsRadial basis function neural network Adaptive resonance theory Harmonic mean algorithm Classification
- 1.Bache, K., Lichman, M.: UCI Machine Learning Repository [http://archive.ics.uci.edu/ml]. University of California, School of Information and Computer Science, Irvine (2013)
- 12.Kohavi, R.: A study of cross validation and bootstrap for accuracy estimation and model selection. In: Proceedings of the 14th International Joint Conference Artificial Intelligence (IJCAI), pp. 1137–1145. Morgan Kaufmann (1995)Google Scholar
- 24.Zhang, B.: Generalized k-harmonic mean—boosting in unsupervised learning. Technical Report HPL-2000-137, Hewlett-Packard Laboratories (2000)Google Scholar