ML Anthology

BCNN: A Binary CNN with All Matrix Ops Quantized to 1 Bit Precision

Arthur J. Redfern, Lijun Zhu, Molly K. Newquist
CVPRW 2021 pp. 4604-4612
doi:10.1109/CVPRW53098.2021.00518 /cvprw/2021/redfern2021cvprw-bcnn/

Abstract

This paper describes a CNN where all CNN style 2D convolution operations that lower to matrix matrix multiplication are fully binary. The network is derived from a common building block structure that is consistent with a constructive proof outline showing that binary neural networks are universal function approximators. 71.24% top 1 accuracy on the 2012 ImageNet validation set was achieved with a 2 step training procedure and implementation strategies optimized for binary operands are provided.

PDF CVPRW Semantic Scholar

Cite

Text

Redfern et al. "BCNN: A Binary CNN with All Matrix Ops Quantized to 1 Bit Precision." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2021. doi:10.1109/CVPRW53098.2021.00518

Markdown

[Redfern et al. "BCNN: A Binary CNN with All Matrix Ops Quantized to 1 Bit Precision." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2021.](https://mlanthology.org/cvprw/2021/redfern2021cvprw-bcnn/) doi:10.1109/CVPRW53098.2021.00518

BibTeX

@inproceedings{redfern2021cvprw-bcnn,
  title     = {{BCNN: A Binary CNN with All Matrix Ops Quantized to 1 Bit Precision}},
  author    = {Redfern, Arthur J. and Zhu, Lijun and Newquist, Molly K.},
  booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
  year      = {2021},
  pages     = {4604-4612},
  doi       = {10.1109/CVPRW53098.2021.00518},
  url       = {https://mlanthology.org/cvprw/2021/redfern2021cvprw-bcnn/}
}