High Frequency Component Helps Explain the Generalization of Convolutional Neural Networks.

We investigate the relationship between the frequency spectrum of image data and the generalization behavior of convolutional neural networks (CNN). We first notice CNN\u0027s ability in capturing the high-frequency components of images. These high-frequency components are almost imperceptible to a human. Thus the observation leads to multiple hypotheses that are related to the generalization behaviors of CNN, including a potential explanation for adversarial examples, a discussion of CNN\u0027s trade-off between robustness and accuracy, and some evidence in understanding training heuristics.