Composite Binary Decomposition Networks


  • You Qiaoben Tsinghua University
  • Zheng Wang Fudan University
  • Jianguo Li Intel Labs
  • Yinpeng Dong Tsinghua University
  • Yu-Gang Jiang Fudan University
  • Jun Zhu Tsinghua University



Binary neural networks have great resource and computing efficiency, while suffer from long training procedure and non-negligible accuracy drops, when comparing to the fullprecision counterparts. In this paper, we propose the composite binary decomposition networks (CBDNet), which first compose real-valued tensor of each layer with a limited number of binary tensors, and then decompose some conditioned binary tensors into two low-rank binary tensors, so that the number of parameters and operations are greatly reduced comparing to the original ones. Experiments demonstrate the effectiveness of the proposed method, as CBDNet can approximate image classification network ResNet-18 using 5.25 bits, VGG-16 using 5.47 bits, DenseNet-121 using 5.72 bits, object detection networks SSD300 using 4.38 bits, and semantic segmentation networks SegNet using 5.18 bits, all with minor accuracy drops.1




How to Cite

Qiaoben, Y., Wang, Z., Li, J., Dong, Y., Jiang, Y.-G., & Zhu, J. (2019). Composite Binary Decomposition Networks. Proceedings of the AAAI Conference on Artificial Intelligence, 33(01), 4747-4754.



AAAI Technical Track: Machine Learning