Circulant Binary Convolutional Networks: Enhancing the Performance of 1-bit DCNNs with Circulant Back Propagation

TL;DR

This paper introduces Circulant Binary Convolutional Networks (CBCNs) that use circulant filters and back propagation to improve the accuracy of 1-bit deep convolutional neural networks, making them more suitable for resource-limited devices.

Contribution

The paper proposes a novel circulant filter design and a circulant binary convolution method, enhancing binarized CNNs' performance and representational capacity.

Findings

CBCNs achieve 61.4% top-1 accuracy on ImageNet with ResNet18.

CBCNs outperform state-of-the-art XNOR networks by up to 10% in accuracy.

The approach effectively minimizes the performance gap between 1-bit and full-precision DCNNs.

Abstract

The rapidly decreasing computation and memory cost has recently driven the success of many applications in the field of deep learning. Practical applications of deep learning in resource-limited hardware, such as embedded devices and smart phones, however, remain challenging. For binary convolutional networks, the reason lies in the degraded representation caused by binarizing full-precision filters. To address this problem, we propose new circulant filters (CiFs) and a circulant binary convolution (CBConv) to enhance the capacity of binarized convolutional features via our circulant back propagation (CBP). The CiFs can be easily incorporated into existing deep convolutional neural networks (DCNNs), which leads to new Circulant Binary Convolutional Networks (CBCNs). Extensive experiments confirm that the performance gap between the 1-bit and full-precision DCNNs is minimized by increasing the filter diversity, which further increases the representational ability in our networks. Our experiments on ImageNet show that CBCNs achieve 61.4% top-1 accuracy with ResNet18. Compared to the state-of-the-art such as XNOR, CBCNs can achieve up to 10% higher top-1 accuracy with more powerful representational ability.