Exploring Frequency Domain Interpretation of Convolutional Neural Networks
This work addresses model interpretability for researchers and practitioners by providing a novel frequency-domain perspective, though it is incremental as it builds on existing spatial-domain methods.
The paper tackles the problem of interpreting convolutional neural networks (CNNs) in the frequency domain by analyzing filter frequency proportions, revealing that controlling these proportions impacts robustness to corruptions, with experiments showing 10.97% average robustness gains on CIFAR-10-C for ResNet-18.
Many existing interpretation methods of convolutional neural networks (CNNs) mainly analyze in spatial domain, yet model interpretability in frequency domain has been rarely studied. To the best of our knowledge, there is no study on the interpretation of modern CNNs from the perspective of the frequency proportion of filters. In this work, we analyze the frequency properties of filters in the first layer as it is the entrance of information and relatively more convenient for analysis. By controlling the proportion of different frequency filters in the training stage, the network classification accuracy and model robustness is evaluated and our results reveal that it has a great impact on the robustness to common corruptions. Moreover, a learnable modulation of frequency proportion with perturbation in power spectrum is proposed from the perspective of frequency domain. Experiments on CIFAR-10-C show 10.97% average robustness gains for ResNet-18 with negligible natural accuracy degradation.