Search

Fabio Anselmi

Translational symmetry in convolutions with localized kernels causes an implicit bias toward high frequency adversarial examples
Robust deep learning object recognition models rely on low frequency information in natural images
Local convolutions cause an implicit bias towards high frequency adversarial examples
Local Convolutions Cause an Implicit Bias towards High Frequency Adversarial Examples

Published with Wowchemy — the free, open source website builder that empowers creators.