English

WRPN & Apprentice: Methods for Training and Inference using Low-Precision Numerics

Computer Vision and Pattern Recognition 2018-03-02 v1 Machine Learning Neural and Evolutionary Computing

Abstract

Today's high performance deep learning architectures involve large models with numerous parameters. Low precision numerics has emerged as a popular technique to reduce both the compute and memory requirements of these large models. However, lowering precision often leads to accuracy degradation. We describe three schemes whereby one can both train and do efficient inference using low precision numerics without hurting accuracy. Finally, we describe an efficient hardware accelerator that can take advantage of the proposed low precision numerics.

Keywords

Cite

@article{arxiv.1803.00227,
  title  = {WRPN & Apprentice: Methods for Training and Inference using Low-Precision Numerics},
  author = {Asit Mishra and Debbie Marr},
  journal= {arXiv preprint arXiv:1803.00227},
  year   = {2018}
}

Comments

Tech report

R2 v1 2026-06-23T00:37:45.597Z