Clustering for Prototype Selection using Singular Value Decomposition
Data clustering is an important technique for exploratory data analysis. The speed, reliability and consistency with which a clustering algorithm can organize large amounts of data constitute reasons to use it in applications like data mining, document retrieval, signal compression, coding and pattern classification. In this paper, we use clustering for efficient large-scale pattern classification; more specifically, we achieve it by selecting appropriate prototypes and features using Singular Value Decomposition (SVD). It is found that the SVD based clustering not only selects better prototypes, but also reduces the memory and computational requirements by 98% over the conventional Nearest Neighbour Classifier (NNC) (T.M.Cover and P.E.Hart (1967)), on OCR data.
KeywordsClassification Accuracy Training Sample Singular Value Decomposition Pattern Classification Handwritten Digit
Unable to display preview. Download preview PDF.
- DEWILDE, P. and DEPRETTERE, ED.F. (1988): Singular Value Decomposition: An introduction. In: Ed. F. Deprettere, editor, SVD and Signal Processing: Algorithms, Applications, and Architectures. Elsevier Science Publishers, North Holland, pp. 3–41.Google Scholar
- DRINCAS, P., ALAN FRIEZE, RAVI KANNAN, SANTOSH VEMPALA, VINAY, V. (1999): Clustering in large graphs and matrices. Proc. of the symposium on Discrete Algorithms, SIAM Google Scholar
- JAIN, A.K. and CHANDRASEKARAN, R. (1982): Dimensionality and sample size considerations in pattern recognition practice, in: Handbook of Dimensionality. P.R. Krishnaiah and L.N. Kanal, Eds. New York: North-HollandGoogle Scholar
- RICHARD O. DUDA, PETER E. HART and DAVID G. STORK. (2000): Pattern Classification (2nd ed.) Google Scholar