Comment on Adversarial Reprogramming of Neural NetworksparentComments−stochastic_monk8yIn light of [0], there is both theoretical and experimental reason to believe that CNNs are mostly picking up fourier decomposition.[0] https://arxiv.org/pdf/1711.11561.pdf
Comments
In light of [0], there is both theoretical and experimental reason to believe that CNNs are mostly picking up fourier decomposition.
[0] https://arxiv.org/pdf/1711.11561.pdf