An accelerated decomposition algorithm for robust support vector Machines
Authors:
- 18 May 2004
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 51 (5) , 234-240
Abstract
This paper proposes an accelerated decomposition algorithm for the robust support vector machine (SVM). Robust SVM aims at solving the overfitting problem when there is outlier in the training data set, which makes the decision surface less contoured and results in sparse support vectors. Training of the robust SVM leads to a quadratic optimization problem with bound and linear constraint. Osuna provides a theorem which proves that the Standard SVM's quadratic programming (QP) problem can be broken down into a series of smaller QP subproblems. This paper derives the Kuhn-Tucker condition and decomposition algorithm for the robust SVM. Furthermore, a pre-selection technique is incorporated into the algorithm to speed up the calculation. The experiment using standard data sets shows that the accelerated decomposition algorithm makes the training process more efficient.Keywords
This publication has 11 references indexed in Scilit:
- Robust support vector regression networks for function approximation with outliersIEEE Transactions on Neural Networks, 2002
- An improved training algorithm for support vector machines2002
- Robust support vector machine with bullet hole image classificationIEEE Transactions on Systems, Man and Cybernetics, Part C (Applications and Reviews), 2002
- Least Squares Support Vector Machine ClassifiersNeural Processing Letters, 1999
- Adaptive margin support vector machines for classification1999
- Making Large-Scale Support Vector Machine Learning Practical1998
- A Tutorial on Support Vector Machines for Pattern RecognitionData Mining and Knowledge Discovery, 1998
- Support-vector networksMachine Learning, 1995
- A training algorithm for optimal margin classifiers1992
- Optimization Theory and Applications1984