Constructing Fast Network through Deconstruction of Convolution

TL;DR

This paper introduces Active Shift Layer (ASL), a learnable shift operation that deconstructs convolution to create faster, resource-efficient neural networks suitable for limited environments like mobile devices.

Contribution

It proposes a novel learnable shift operation, ASL, that replaces traditional convolutions, enabling the construction of lightweight, high-performance neural networks.

Findings

ASL outperforms existing lightweight networks in accuracy and speed.

The method allows end-to-end training of shift parameters.

The proposed network surpasses state-of-the-art models in resource-constrained settings.

Abstract

Convolutional neural networks have achieved great success in various vision tasks; however, they incur heavy resource costs. By using deeper and wider networks, network accuracy can be improved rapidly. However, in an environment with limited resources (e.g., mobile applications), heavy networks may not be usable. This study shows that naive convolution can be deconstructed into a shift operation and pointwise convolution. To cope with various convolutions, we propose a new shift operation called active shift layer (ASL) that formulates the amount of shift as a learnable function with shift parameters. This new layer can be optimized end-to-end through backpropagation and it can provide optimal shift values. Finally, we apply this layer to a light and fast network that surpasses existing state-of-the-art networks.