Data Selection for Nonlinear Proximal Support Vector Machine

An incremental learning method based on a new nonlinear proximal support vector machine (PSVM) classifier was developed, which can be utilized in online learning efficiently. However the memory requirement of this method is proportional to the square of the size of the training data, which makes it impractical in large data set learning problem. In this paper a data selection method, which can select a small fraction of the entire dataset as "support vectors" of PSVM classifiers, is devised. We also proposed a framework for incremental learning using this data selection method. It maintains only a small fraction of a large data set before merging and processing it with new incoming data, which makes online large dataset learning problem solvable for nonlinear PSVM. Mathematical analysis and experimental results demonstrated the effectiveness of our proposed technique both in batch mode and in online learning situation.