Comb Convolution for Efficient Convolutional Architecture

TL;DR

This paper introduces comb convolution, a novel operator that reduces redundant computations in CNNs by eliminating half of the spatial connections within channels, significantly improving efficiency without sacrificing accuracy.

Contribution

The paper proposes comb convolution, a new convolutional operator that exploits intra-channel spatial sparsity to reduce FLOPs by 50% while maintaining accuracy.

Findings

Achieves 50% FLOPs reduction on state-of-the-art CNNs.

Maintains comparable accuracy with standard convolutions.

Compatible with existing architectures for enhanced efficiency.

Abstract

Convolutional neural networks (CNNs) are inherently suffering from massively redundant computation (FLOPs) due to the dense connection pattern between feature maps and convolution kernels. Recent research has investigated the sparse relationship between channels, however, they ignored the spatial relationship within a channel. In this paper, we present a novel convolutional operator, namely comb convolution, to exploit the intra-channel sparse relationship among neurons. The proposed convolutional operator eliminates nearly 50% of connections by inserting uniform mappings into standard convolutions and removing about half of spatial connections in convolutional layer. Notably, our work is orthogonal and complementary to existing methods that reduce channel-wise redundancy. Thus, it has great potential to further increase efficiency through integrating the comb convolution to existing architectures. Experimental results demonstrate that by simply replacing standard convolutions with comb convolutions on state-of-the-art CNN architectures (e.g., VGGNets, Xception and SE-Net), we can achieve 50% FLOPs reduction while still maintaining the accuracy.