An experimental comparison of some efficient approaches for training support vector machines

Support Vector Machines (SVMs) are one of the most widely used techniques for developing classification and regression models. A significant portion of the recent research on SVMs is devoted to the development of efficient computational approaches for SVM training. This paper performs an experimental analysis of some approaches recently developed for training SVM classification models, including decomposition algorithms, explicit solution techniques, and linear programming. The analysis involves the generalizing performance of the SVM models and the computational efficiency of the algorithms. The results lead to useful conclusions on the performance of the training techniques and to the applicability of linear and non-linear SVM models.