• DOI:10.1109/ICCV.2019.00141
Corpus ID: 184487878

Data-Free Quantization Through Weight Equalization and Bias Correction

@article{Nagel2019DataFreeQT,
  title={Data-Free Quantization Through Weight Equalization and Bias Correction},
  author={M. Nagel and Mart van Baalen and Tijmen Blankevoort and M. Welling},
  journal={2019 IEEE/CVF International Conference on Computer Vision (ICCV)},
  year={2019},
  pages={1325-1334}
}

M. Nagel, Mart van Baalen, +1 author M. Welling

Published 2019

Computer Science, Mathematics

2019 IEEE/CVF International Conference on Computer Vision (ICCV)

We introduce a data-free quantization method for deep neural networks that does not require fine-tuning or hyperparameter selection. [...] Key Method Our approach relies on equalizing the weight ranges in the network by making use of a scale-equivariance property of activation functions. In addition the method corrects biases in the error that are introduced during quantization. This improves quantization accuracy performance, and can be applied to many common computer vision architectures with a straight…Expand Abstract

View on IEEE

arxiv.org

Save to Library

Create Alert

Cite

Launch Research Feed