AI Paper Atlas

Research map / Adversarial robustness and security / Certified and empirical robustness

On the Structural Sensitivity of Deep Convolutional Networks to the Directions of Fourier Basis Functions

Yusuke Tsuzuku, Issei Sato

CVPR 2019 · Adversarial robustness and security › Certified and empirical robustness · approach: Decision-boundary geometry of robustness · 58 citations (OpenAlex)

Check your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use a selected Fourier basis perturbation instead of random noiseCNN classifiers · CIFAR10fool ratio: 5.0 -> 63.3; 8.1 -> 82.3; 6.8 -> 72.2; 5.4 -> 50.7Improved
Use a selected Fourier basis perturbation instead of random noiseCNN classifiers · CIFAR100fool ratio: 13.0 -> 83.4; 26.4 -> 93.7; 25.9 -> 95.8; 22.5 -> 72.3Improved
Use a selected Fourier basis perturbation instead of random noiseCNN classifiers · MNISTfool ratio: 0.1 -> 0.4; 55.8 -> 90.2; 0.0 -> 0.1; 0.1 -> 0.2Mixed
Use a selected Fourier basis perturbation instead of random noiseCNN classifiers · Fashion MNISTfool ratio: 5.4 -> 12.5; 11.4 -> 48.1; 8.3 -> 83.7; 12.6 -> 56.9Improved
Use a selected Fourier basis perturbation instead of random noiseCNN classifiers · SVHNfool ratio: 3.1 -> 64.9; 5.2 -> 90.8; 0.0 -> 0.0; 4.6 -> 50.5Mixed
Replace SFA's Fourier basis with its signed version (ablation)SFA on ILSVRC2015 · ILSVRC2015fool ratio: 34.6 -> 44.1; 38.7 -> 40.1; 49.7 -> 53.3; 36.8 -> 39.5Mixed
Increase perturbation size from 10/255 to 20/255SFA on ILSVRC2015 · ILSVRC2015fool ratio: 34.6 -> 62.3; 38.7 -> 68.5; 49.7 -> 76.3; 36.8 -> 63.5Improved
Use a Fourier basis perturbation rather than a standard pixel basis directionCNNs · Various datasetsfool ratioImproved
Use convolutional networks rather than an MLPFourier-direction sensitivity · Various datasetsfool ratioImproved
Use a single Fourier basis perturbation on MNISTCNN classifiers · MNISTfool ratioHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy