Pattern Classification Using a Set of Compact Hyperspheres
Prototype classifiers are one of the simplest and most intuitive approaches in pattern classification. However, they need careful positioning of prototypes to capture the distribution of each class region. Classical methods, such as learning vector quantization (LVQ), are sensitive to the initial choice of the number and the locations of the prototypes. To alleviate this problem, a new method is proposed that represents each class region by a set of compact hyperspheres. The number of hyperspheres and their locations are determined by setting up the problem as a set of quadratic optimization problems. Experimental results show that the proposed approach significantly beats LVQ and Restricted Coulomb Energy (RCE) in most performance aspects.
KeywordsPattern Classification Learn Vector Quantization Quadratic Optimization Problem Class Region Prototype Classifier
Unable to display preview. Download preview PDF.
- 6.Atiya, A., Hashem, S., Fayed, H.: New hyperspheres for pattern classification. In: Proc. 1st International Computer Engineering Conference (ICENCO-2004), Cairo, Egypt, pp. 258–263 (2004)Google Scholar
- 9.Kohonen, T.: Self-organization and Associative Memory, 3rd edn. Springer, Heidelberg (1989)Google Scholar
- 12.Prechelt, L.: Proben1. A Set of Neural-Network Benchmark Problems. Univ. Karlsruhe, Germany (1994), available FTP: ira.uka.de/pub/neuron/proben1.tar.gz