Journal
INFORMATION SCIENCES
Volume 181, Issue 18, Pages 3967-3980Publisher
ELSEVIER SCIENCE INC
DOI: 10.1016/j.ins.2011.05.004
Keywords
Twin support vector machine; Empirical risk minimization; Sparse control; Back-fitting strategy; Primal space
Categories
Funding
- Shanghai Municipal Education Commission [11YZ81]
- Natural Science Foundation of SHNU [SK200937, SK201030]
- Shanghai Leading Academic Discipline Project [S30405]
Ask authors/readers for more resources
Twin support vector machines (TSVM) obtain faster training speeds than classical support vector machines (SVM). However, TSVM augmented vectors lose sparsity. In this paper, a rapid sparse twin support vector machine (STSVM) classifier in primal space is proposed to improve the sparsity and robustness of TSVM. Based on a simple back-fitting strategy, the STSVM iteratively builds each nonparallel hyperplanes by adding one support vector (SV) from the corresponding class at one time. This process is terminated using an adaptive and stable stopping criterion. STSVM learning is implemented by linear equation computing systems through introducing a quadratic function to approximate the empirical risk. The computational results on several synthetic and benchmark datasets indicate that the STSVM obtains a sparse separating hyperplane at a low cost without sacrificing its generalization performance. Crown Copyright (C) 2011 Published by Elsevier Inc. All rights reserved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available