/B-LRP

B-LRP is the repository for the paper How Much Can I Trust You? — Quantifying Uncertainties in Explaining Neural Networks

Primary LanguageJupyter Notebook

B-LRP

How much can I trust you? - Quantifying Uncertainties in Explaining Neural Networks

https://arxiv.org/abs/2006.09000

We provide Pytorch implementations for our paper How Much Can I Trust You? — Quantifying Uncertainties in Explaining Neural Networks

Requirements

To be able to run the code properly, please download the imagenet subset data by following link: https://www.dropbox.com/s/hbo7udo721igpnd/imagenet_data.zip?dl=0 and copy the 'imagenet_data' folder to a directory with notebooks

To install requirements:

pip install -r requirements.txt

All experiments were conducted with Python 3.6.9.

Code structure:

The code is structured in the following way:

  • B-LRP MNIST This notebook consists of the code for visualization of B-LRP method for MNIST

  • B-LRP Imagenet This notebook consists of the code for visualization of B-LRP method for Imagenet

Cite this paper:

To cite this paper use following Bibtex annotation:

@article{bykov2020much,
  title={How Much Can I Trust You?--Quantifying Uncertainties in Explaining Neural Networks},
  author={Bykov, Kirill and H{\"o}hne, Marina M-C and M{\"u}ller, Klaus-Robert and Nakajima, Shinichi and Kloft, Marius},
  journal={arXiv preprint arXiv:2006.09000},
  year={2020}
}