A Recursive Gradient Method of Least Square Support Vector Machine and its Application

Abstract:

Least Square Support Vector Machine (LS-SVM) is an important machine of Support Vector Machine (SVM). But this method can not be used for online identification, and maybe lead to calculation inflation. A gradient recursive method of LS-SVM is presented by combining the LS-SVM method with the gradient method. This method can overcome the influence of bad data to the parameter estimation, has a stronger robustness, and improves the calculation speed of LS-SVM. The presented method is applied to the modeling of chaotic series. The simulation example validates the validity of the presented method.

• Periodical: Applied Mechanics and Materials (Volume 339)
• Pages: 349-354
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.339.349
• Online since: July 2013
• Authors: Xiang Yong Luo, Wei Min Lv, Zhao Qing Song, Shi Wei Jiang
• Keywords: Gradient, Least Square (LS), Recursion, Support Vector Machine (SVM)

References

[1] V.N. Vapnik, in: The Nature of Statiscal Learning Theory. Springer-Verlag, New York, (1995) in press.

[2] Nello Cristinaini, John Shawe-Yaylo, in: An Introduction to Support Vector Machines. China Machine Press, Beijing, (2005) in press.

[3] V.N. Vapnik, in: Statistical Learning Theory. Wiley, (1998) in press.

[4] J.A.K. Suykens, J. Vandewalle: Least squares support vector machine classifiers. Neural Processing Letters Vol.9 (1999), p.293~300

[5] T.V. Gestel, J.A.K. Suykens, et al. Benchmarking Least Squares Support Vector Machine Classifiers. Machine Learning Vol. 54 (2004), p.5~32

DOI: 10.1023/b:mach.0000008082.80494.e0

[6] D. Anguita1, A. Boni: Digital Least Squares Support Vector Machines. Neural Processing Letters Vol. 18 (2003), p.65~72

[7] D. Tsujinishi, S.Abe: Fuzzy least squares support vector machines for multi-class problems. Neural Networks Vol. 16 (2003), p.785~792

DOI: 10.1016/s0893-6080(03)00110-2