Robustness of neural networks to random noise perturbations of their inputs

Mark Levene, Martyn Harris

We investigate the problem of the robustness of a trained neural network to the perturbation of its input values. More specifically, we examine the interplay between the accuracy o…