Skip to main content
. Author manuscript; available in PMC: 2020 Oct 23.
Published in final edited form as: Proc Int Jt Conf Neural Netw. 2020 Sep 28;2020:10.1109/ijcnn48605.2020.9206678. doi: 10.1109/ijcnn48605.2020.9206678

Fig. 2.

Fig. 2.

Left: Bar plot that shows how the five models trained on the MNIST dataset agree on the overgeneralized samples (e.g., dog samples of the CIFAR-10 dataset). Right: Bar plot that shows how the five models classify the adversarial examples generated by one of the models (m3); the bars denote the classification of the samples to true labels, adversarial labels and other labels, respectively.