An Optimized Approach on Reduced Kernel Matrix to ClusterSVM
Yali Qi, Wei He, Hou Shu · 2008
For classification problem clustering method divides the dataset into many clusters based on correlation attribute of all elements of the dataset. How to classify data within the same clustering number as close as possible and data in different clusters as depart as possible is the key of clustering method. Clustering support vector machines (ClusterSVM) partition the training data into disjoint clusters first, then train an initial support vectors using representatives of these clusters. These initial support vectors which give us a global picture of the solution can approximately identify the support vectors and non-support vectors. The training process is accelerated by replacing non-support vectors with few data. The initial SVM of cluster is the key of training ClusterSVM. This paper proposed a reduced kernel matrix to generate a nonlinear separating surface which depends on a small randomly selected portion of the dataset, and used this kernel to train the initial SVM of clusters. Computational results indicate computational times as well as the number of training data are much smaller for the improved method than that of a conventional ClusterSVM.