Adaptive multi-class support vector machine for microarray classification and gene selection

This paper proposes an adaptive multi-class support vector machine for simultaneous microarray classification and gene selection. By evaluating the gene ranking significance, the adaptive multi-class support vector machine is shown to encourage an adaptive grouping effect in the process of building classifiers, thus leading a sparse multi-classifiers with enhanced interpretability. Based on a reasonable correlation between the two regularization parameters, an efficient solution path algorithm is developed for solving the proposed support vector machine. Experiments performed on the leukaemia data set are provided to verify the obtained results.