Abstract
Generalization error bounds in Support Vector Machines are based on the minimum distance between training points and the separating hyperplane. The error of soft margin algorithm can be bounded by a target margin and some norms of the slack vector. In this paper, we propose a new method controlling allowable error and formulate considering the contamination by noise in data directly. The method can provide desirable separating hyperplanes easily by controlling a restricted slack parameter. Additionally, through an artificial numerical example, we compare the proposed method with a conventional soft margin algorithm.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Bartlett, P., Shawe-Taylor, J. (1999) Generalization Performance of Support Vector Machines and Other Pattern Classifiers, Advances in Kernel Methods-Support Vector Learning (edited by B. Schölkopf, C.J.C. Burges and A.J. Smola ), MIT Press, 43–54
Cristianini, N. and Shawe-Taylor, J. (2000) An Introduction to Support Vector Machine, Cambridge University Press
Shawe-Taylor, J., Cristianini, N. (2000) On the Generalisation of Soft Margin Algorithms, NeuroCOLT2 Technical Report Series, NC-TR-2000–082
Shawe-Taylor, J., Bartlett, P.L., Williamson, R.C., M. Anthony (1998) Structural Risk Minimization over Data-Dependent Hierarchies, IEEE Transactions on Information Theory 44 (5), 1926–1940
Magasarian, O.L. (2000) Generalized Support Vector Machines, Advances in Large Margin Classifiers (edited by A. Smola, P. Bartlett, B. Shoolköpf and D. Schuurmans ), 135–146.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 2003 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Yoon, M., Nakayama, H., Yun, Y. (2003). A Soft Margin Algorithm controlling Tolerance Directly. In: Multi-Objective Programming and Goal Programming. Advances in Soft Computing, vol 21. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-36510-5_41
Download citation
DOI: https://doi.org/10.1007/978-3-540-36510-5_41
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-00653-4
Online ISBN: 978-3-540-36510-5
eBook Packages: Springer Book Archive