Spectral Leakage and Rethinking the Kernel Size in CNNs
This addresses performance degradation in CNNs for image classification tasks, but it is incremental as it applies a known signal processing technique to an existing method.
The paper tackles the problem of spectral leakage in CNNs due to small kernel sizes, which degrades performance, and shows that using larger kernels with a Hamming window improves classification accuracy on datasets like Fashion-MNIST, CIFAR-10, CIFAR-100, and ImageNet, and increases robustness to adversarial attacks.
Convolutional layers in CNNs implement linear filters which decompose the input into different frequency bands. However, most modern architectures neglect standard principles of filter design when optimizing their model choices regarding the size and shape of the convolutional kernel. In this work, we consider the well-known problem of spectral leakage caused by windowing artifacts in filtering operations in the context of CNNs. We show that the small size of CNN kernels make them susceptible to spectral leakage, which may induce performance-degrading artifacts. To address this issue, we propose the use of larger kernel sizes along with the Hamming window function to alleviate leakage in CNN architectures. We demonstrate improved classification accuracy on multiple benchmark datasets including Fashion-MNIST, CIFAR-10, CIFAR-100 and ImageNet with the simple use of a standard window function in convolutional layers. Finally, we show that CNNs employing the Hamming window display increased robustness against various adversarial attacks.