EZCrop: Energy-Zoned Channels for Robust Output Pruning
This work addresses the need for efficient and robust channel pruning in CNNs, offering a novel interpretation and method that is incremental but improves upon existing techniques.
The paper tackled the problem of understanding and exploiting the constant rank phenomenon in CNN feature maps for channel pruning, resulting in a new FFT-based metric that achieves consistently better pruning results than state-of-the-art methods.
Recent results have revealed an interesting observation in a trained convolutional neural network (CNN), namely, the rank of a feature map channel matrix remains surprisingly constant despite the input images. This has led to an effective rank-based channel pruning algorithm, yet the constant rank phenomenon remains mysterious and unexplained. This work aims at demystifying and interpreting such rank behavior from a frequency-domain perspective, which as a bonus suggests an extremely efficient Fast Fourier Transform (FFT)-based metric for measuring channel importance without explicitly computing its rank. We achieve remarkable CNN channel pruning based on this analytically sound and computationally efficient metric and adopt it for repetitive pruning to demonstrate robustness via our scheme named Energy-Zoned Channels for Robust Output Pruning (EZCrop), which shows consistently better results than other state-of-the-art channel pruning methods.