A novel all-at-once learning method for multi-class Support Vector Machine

In this paper, first we deal with multi-class problems with Support Vector Machine, and then propose a novel and efficient all-at-once method with Support Vector Machine for solving multi-class problems. We evaluate our method for some benchmark data sets and experiment result shows that classification performance of our approach is comparable with one-against-all decomposition solved by the SMO algorithm, and the method not only saves computation time but also keeps accuracy of classification.