A global optimization of SVM batch active learning

We consider the problem of SVM batch active learning, which involves distinguishing samples chosen and maximum approximate the real normal vector w in feature space. Although several studies are devoted to batch mode active learning, they suffer either from the uncertain parameter set or from the solutions of local optimization. We introduce a new algorithm for performing batch active learning by cluster diversity and most possibly error approximate method. Experimental results showing that employing our active learning method can significantly reduce the computational cost as well as excellent learning performance in comparison with other active learning methods.