Deep Learning with Limited Numerical Precision

Suyog Gupta, Ankur Agrawal, Kailash Gopalakrishnan, Pritish Narayanan

ICML 2015 pp. 1737-1746

/icml/2015/gupta2015icml-deep/

Abstract

Training of large-scale deep neural networks is often constrained by the available computational resources. We study the effect of limited precision data representation and computation on neural network training. Within the context of low-precision fixed-point computations, we observe the rounding scheme to play a crucial role in determining the network’s behavior during training. Our results show that deep networks can be trained using only 16-bit wide fixed-point number representation when using stochastic rounding, and incur little to no degradation in the classification accuracy. We also demonstrate an energy-efficient hardware accelerator that implements low-precision fixed-point arithmetic with stochastic rounding

PDF ICML Semantic Scholar

Cite

Text

Gupta et al. "Deep Learning with Limited Numerical Precision." International Conference on Machine Learning, 2015.

Markdown

[Gupta et al. "Deep Learning with Limited Numerical Precision." International Conference on Machine Learning, 2015.](https://mlanthology.org/icml/2015/gupta2015icml-deep/)

BibTeX

@inproceedings{gupta2015icml-deep,
  title     = {{Deep Learning with Limited Numerical Precision}},
  author    = {Gupta, Suyog and Agrawal, Ankur and Gopalakrishnan, Kailash and Narayanan, Pritish},
  booktitle = {International Conference on Machine Learning},
  year      = {2015},
  pages     = {1737-1746},
  volume    = {37},
  url       = {https://mlanthology.org/icml/2015/gupta2015icml-deep/}
}