Interpreting Deep Neural Networks with Relative Sectional Propagation by Analyzing Comparative Gradients and Hostile Activations [AAAI 2021]
Detail description of this method is provided in our paper https://arxiv.org/abs/2012.03434.
This code provides a pytorch implementation of RSP.
Requirements
pytorch >= 1.2.0
python >= 3.6
matplotlib >= 1.3.1
Paper Citation
When using this code, please cite our paper.
@inproceedings{nam2021interpreting,
title={Interpreting Deep Neural Networks with Relative Sectional Propagation by Analyzing Comparative Gradients and Hostile Activations},
author={Nam, Woo-Jeoung and Choi, Jaesik and Lee, Seong-Whan},
booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
volume={35},
number={13},
pages={11604--11612},
year={2021}
}