Projection Convolutional Neural Networks for 1-bit CNNs via Discrete Back Propagation


  • Jiaxin Gu Beihang University
  • Ce Li China University of Mining and Technology
  • Baochang Zhang Beihang University
  • Jungong Han Beihang University
  • Xianbin Cao Beihang University
  • Jianzhuang Liu Huawei Technologies Company, Ltd.
  • David Doermann State University of New York at Buffalo



The advancement of deep convolutional neural networks (DCNNs) has driven significant improvement in the accuracy of recognition systems for many computer vision tasks. However, their practical applications are often restricted in resource-constrained environments. In this paper, we introduce projection convolutional neural networks (PCNNs) with a discrete back propagation via projection (DBPP) to improve the performance of binarized neural networks (BNNs). The contributions of our paper include: 1) for the first time, the projection function is exploited to efficiently solve the discrete back propagation problem, which leads to a new highly compressed CNNs (termed PCNNs); 2) by exploiting multiple projections, we learn a set of diverse quantized kernels that compress the full-precision kernels in a more efficient way than those proposed previously; 3) PCNNs achieve the best classification performance compared to other state-ofthe-art BNNs on the ImageNet and CIFAR datasets.




How to Cite

Gu, J., Li, C., Zhang, B., Han, J., Cao, X., Liu, J., & Doermann, D. (2019). Projection Convolutional Neural Networks for 1-bit CNNs via Discrete Back Propagation. Proceedings of the AAAI Conference on Artificial Intelligence, 33(01), 8344-8351.



AAAI Technical Track: Vision