EZCrop: Energy-Zoned Channels for Robust Output Pruning

TL;DR

EZCrop introduces a frequency-domain based, efficient channel importance metric for CNN pruning, demonstrating superior robustness and performance over existing methods through iterative pruning.

Contribution

The paper proposes a novel FFT-based importance metric for CNN channel pruning and introduces EZCrop, a robust iterative pruning scheme that outperforms state-of-the-art methods.

Findings

EZCrop achieves higher accuracy retention after pruning.

The FFT-based metric correlates well with channel importance.

EZCrop demonstrates robustness across multiple CNN architectures.

Abstract

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.