Solid Convex-Hull Sequential Support Vector Machine

Support Vector Machine (SVM) is a useful classification tool. The main disadvantage of SVM algorithms is that it's time-consuming to train large data set because of the optimization(QP) problem. Hence, to accelerate the speed of SVM, simplify the dataset is an available method. In fact, what we need to build the SVM hyper plane are support vectors, which are only a small part of the whole data. How to keep the useful vectors and discard useless ones as much as possible is still a problem. If we save time but lose too much accuracy, this method is meaningless. In this article, we proposed a method to reduce the training time and keep the accuracy simultaneously.