Sample Selection Based on K-L Divergence for Effectively Training SVM | IEEE Conference Publication | IEEE Xplore