Sampling Method for Fast Training of Support Vector Data Description

Support Vector Data Description (SVDD) is a machine learning technique used for single class classification and outlier detection. The SVDD model for normal data description builds a minimum radius hypersphere around the training data. A flexible description can be obtained by use of Kernel functions. The data description is defined by the support vectors obtained by solving quadratic optimization problem which minimizes the volume enclosed by the hypersphere. The time required to solve the quadratic programming problem is directly related to the number of observations in the training data set. This leads to very high computing time for large training datasets. In this paper we propose a new iterative sampling-based method for SVDD training. The method incrementally learns the training data set description at each iteration by computing SVDD on an independent random sample selected with replacement from the training data set. The experimental results indicate that the proposed method is extremely fast and provides near-identical data description as compared to training using the entire data set in one iteration. Proposed method can be easily implemented as a wrapper code around the core module for SVDD training computations either in a single machine or a multi-machine distributed environment.