Fig. 7: Effect of task and model complexity on entropy-robustness relationship.
From: Exploring robust architectures for deep artificial neural networks
a Entropy-robustness correlation coefficient (vertical axis) is plotted against the number of classes (horizontal axis) in datasets. As the task becomes complex, entropy becomes significantly more correlated to the robustness of Deep Artificial Neural Networks (DANNs). b The entropy-robustness correlation coefficient is plotted against the number of model parameters for different DANNs (ResNet-29, ResNet-41, and CNNs) using CIFAR-100 dataset. For the same task, as the number of model parameters increases, the entropy-robustness correlation increases significantly (p < 0.05).
