Code for ICCV2019 Oral paper "PR Product: A Substitute for Inner Product in Neural Networks", containing PR-FC, PR-CNN and PR-LSTM.
The code is implemented based on Pytorch.
The usage of PR-X is the same as the one of nn.X:
import PR
pr_fc = PR.PRLinear(100, 200)
pr_cnn = PR.PRConv2d(32, 64, kernel_size=3, stride=1, padding=1, bias=False)
pr_lstmcell = PR.PRLSTMCell(256, 512)
If you use 'PR Product' in a scientific publication, we would appreciate references to the following paper:
PR Product: A Substitute for Inner Product in Neural Networks. ICCV2019 Oral: PR Product
Biblatex entry:
@inproceedings{wang2019pr,
title={PR Product: A Substitute for Inner Product in Neural Networks},
author={Wang, Zhennan and Zou, Wenbin and Xu, Chen},
booktitle={Proceedings of the IEEE International Conference on Computer Vision},
pages={6013--6022},
year={2019}
}
This code is released under the MIT License (refer to the LICENSE file for details).