Robustness of Selected Learning Models under Label-Flipping Attack