Abstract
Global convergence of the sequential minimal optimization (SMO) algorithm for support vector regression (SVR) is studied in this paper. Given ι training samples, SVR is formulated as a convex quadratic programming (QP) problem with ι pairs of variables. We prove that if two pairs of variables violating the optimality condition are chosen for update in each step and subproblems are solved in a certain way, then the SMO algorithm always stops within a finite number of iterations after finding an optimal solution. Also, efficient implementation techniques for the SMO algorithm are presented and compared experimentally with other SMO algorithms.
| Original language | English |
|---|---|
| Pages (from-to) | 971-982 |
| Number of pages | 12 |
| Journal | IEEE Transactions on Neural Networks |
| Volume | 19 |
| Issue number | 6 |
| DOIs | |
| Publication status | Published - 2008 |
| Externally published | Yes |
Fingerprint
Keywords
- Convergence
- Quadratic programming (QP)
- Sequential minimal optimization (SMO)
- Support vector regression (SVR)
ASJC Scopus subject areas
- Artificial Intelligence
- Computer Networks and Communications
- Computer Science Applications
- Software
- Control and Systems Engineering
- Theoretical Computer Science
- Electrical and Electronic Engineering
- Computational Theory and Mathematics
- Hardware and Architecture
Cite this
Global convergence of SMO algorithm for support vector regression. / Takahashi, Norikazu; Guo, Jun; Nishi, Tetsuo.
In: IEEE Transactions on Neural Networks, Vol. 19, No. 6, 2008, p. 971-982.Research output: Contribution to journal › Article
}
TY - JOUR
T1 - Global convergence of SMO algorithm for support vector regression
AU - Takahashi, Norikazu
AU - Guo, Jun
AU - Nishi, Tetsuo
PY - 2008
Y1 - 2008
N2 - Global convergence of the sequential minimal optimization (SMO) algorithm for support vector regression (SVR) is studied in this paper. Given ι training samples, SVR is formulated as a convex quadratic programming (QP) problem with ι pairs of variables. We prove that if two pairs of variables violating the optimality condition are chosen for update in each step and subproblems are solved in a certain way, then the SMO algorithm always stops within a finite number of iterations after finding an optimal solution. Also, efficient implementation techniques for the SMO algorithm are presented and compared experimentally with other SMO algorithms.
AB - Global convergence of the sequential minimal optimization (SMO) algorithm for support vector regression (SVR) is studied in this paper. Given ι training samples, SVR is formulated as a convex quadratic programming (QP) problem with ι pairs of variables. We prove that if two pairs of variables violating the optimality condition are chosen for update in each step and subproblems are solved in a certain way, then the SMO algorithm always stops within a finite number of iterations after finding an optimal solution. Also, efficient implementation techniques for the SMO algorithm are presented and compared experimentally with other SMO algorithms.
KW - Convergence
KW - Quadratic programming (QP)
KW - Sequential minimal optimization (SMO)
KW - Support vector regression (SVR)
UR - http://www.scopus.com/inward/record.url?scp=49149114060&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=49149114060&partnerID=8YFLogxK
U2 - 10.1109/TNN.2007.915116
DO - 10.1109/TNN.2007.915116
M3 - Article
C2 - 18541498
AN - SCOPUS:49149114060
VL - 19
SP - 971
EP - 982
JO - IEEE Transactions on Neural Networks and Learning Systems
JF - IEEE Transactions on Neural Networks and Learning Systems
SN - 2162-237X
IS - 6
ER -