An Efficient Approach to Boost Support Vector Data Description

In this paper, we propose an efficient approach to speed up the training of support vector data description (SVDD). Our approach works in two steps. In the first step, we propose k-farthest neighbors method to identify the samples around the hyper-sphere. These samples are mostly likely to be support vectors which contribute to the construction of the classifier. In the second step, we construct SVDD classifier on the selected samples to boost the training of SVDD and guarantee its accuracy as well. In the end, we conduct extensive experiments to investigate the performance of our approach and the results show that our method can speed up the training of SVDD and hold similar accuracy in comparison of the original SVDD method.