A decomposition method for support vector clustering

In this paper we study how to apply decomposition method for support vector clustering (SVC) and compare its performance with that of non-decomposition methods which solves quadratic programming problem using traditional techniques. Decomposition is one of the major methods for solving SVMs and is known for its efficiency. SVC is a novel clustering method that uses SVM approach. SVC problem is formulated as a quadratic optimization with constrains to come up with clusters. When applying SVC to large data sets, traditional ways of solving quadratic programming problem require huge memory and computational time. This problem is eased by the application of the decomposition method. This paper discusses the application of the decomposition method in such cases and gives a detailed discussion on how to apply the method. The method is applied on a few benchmark data sets and results shows a substantial improvement in terms of performance.