DoReFa-Net: Training low bitwidth convolutional neural networks with low bitwidth gradients

Zekun Ni
Zekun Ni
Xinyu Zhou
Xinyu Zhou
He Wen
He Wen
Yuheng Zou
Yuheng Zou

arXiv preprint arXiv:1606.06160, 2016.

Cited by: 703|Bibtex|Views63
Other Links: arxiv.org

Abstract:

We propose DoReFa-Net, a method to train convolutional neural networks that have low bitwidth weights and activations using low bitwidth parameter gradients. In particular, during backward pass, parameter gradients are stochastically quantized to low bitwidth numbers before being propagated to convolutional layers. As convolutions during ...More

Code:

Data:

Full Text
Your rating :
0

 

Tags
Comments