Department of Applied Mathematics & Physics, Kyoto University

Technical Report 2005-011 (September 22, 2005)

Second Order Cone Programming Formulations for Robust Multi-class Classification
by Ping Zhong and Masao Fukushima

pdf File

Multi-class classification is an important and on-going research subject in machine learning. Current support vector methods for multi-class classification implicitly assume that the parameters in the optimization problems to be known exactly. However, in practice, the parameters have perturbations since they are estimated from the training data which are usually subject to measurement noise. In this paper, we propose linear and nonlinear robust formulations for multi-class classification based on M-SVM method. The preliminary numerical experiments concerm the robustness of the proposed method.