Huberized Multiclass Support Vector Machine for Microarray Classification

This paper proposes a new multiclass support vector machine (SVM) for simultaneous gene selection and microarray classification. Combining the huberized hinge loss function and the elastic net penalty, the proposed SVM can perform automatic gene selection and encourages a grouping effect. The coefficient paths of the proposed SVM are shown to be piecewise linear with respect to the single regularization parameter, based on which the solution path algorithm is developed with low computational complexity. Experiments performed on the leukemia data set are provided to verify the obtained results.