Abstract—The efficient classification ability of support vector machines (SVMs) has been shown in many practical applica-tions, but currently they are considerably slower in testing phase than other approaches with similar classification performance due to a large number of support vectors included in the solution. Among different approaches, simplification of support vector machine (SimpSVM) speeds-up the testing phase by replacing original SVM with a simplified one that consists of a smaller number of support vectors. However, the ultimate goal of the simplification is to keep the simplified solution as similar to the original solution as possible. To improve this similarity, in this paper, we propose two improved SimpSVMs that are based on stochastic gradient descent. Experiments on some datasets show improved results by our algorithms.
Index Terms—Support vector machines, simplification of support vector machine, stochastic gradient descent, solution optimization.
Pham Quoc Thang, Hoang Thi Lam are with Tay Bac University, Vienam (e-mail: firstname.lastname@example.org).
Nguyen Thanh Thuy is with VNU University of Engineering and Technology, Vietnam.
Cite: Pham Quoc Thang, Hoang Thi Lam, and Nguyen Thanh Thuy, "Improving Simplification of Support Vector Machine for Classification," International Journal of Machine Learning and Computing vol. 8, no. 4, pp. 372-376, 2018.