Spectral leakage and rethinking the kernel size in cnns

N Tomen, JC van Gemert - Proceedings of the IEEE/CVF …, 2021 - openaccess.thecvf.com
Convolutional layers in CNNs implement linear filters which decompose the input into
different frequency bands. However, most modern architectures neglect standard principles …

The effects of image distribution and task on adversarial robustness

O Kunhardt, A Deza, T Poggio - arxiv preprint arxiv:2102.10534, 2021 - arxiv.org
In this paper, we propose an adaptation to the area under the curve (AUC) metric to
measure the adversarial robustness of a model over a particular $\epsilon $-interval …

[PDF][PDF] Quantifying Adversarial Robustness of a Model as a Function of the Image Distribution

O Kunhardt, A Deza, T Poggio - arturodeza.wdfiles.com
In this paper, we propose an adaptation to the area under the curve (AUC) metric to
measure the adversarial robustness of a model over a particular∈-interval [∈ 0,∈ …