Home

Awesome

👋 Sobol Attribution Method (NeurIPS 2021)

This repository contains code for the paper:

Look at the Variance! Efficient Black-box Explanations with Sobol-based Sensitivity Analysis, Thomas Fel*, Rémi Cadène*, Mathieu Chalvidal, Matthieu Cord, David Vigouroux & Thomas Serre. NeurIPS 2021, [arXiv].

The code is implemented and available for Pytorch & Tensorflow. A notebook for each of them is available: notebook Pytorch, notebook Tensorflow.

<img src="./assets/images.png" width="500px"> <img src="./assets/explanations.png" width="500px"> 
@inproceedings{fel2021sobol,
      title={Look at the Variance! Efficient Black-box Explanations with Sobol-based Sensitivity Analysis}, 
      author={Thomas Fel and Remi Cadene and Mathieu Chalvidal and Matthieu Cord and David Vigouroux and Thomas Serre},
      year={2021},
      booktitle={Advances in Neural Information Processing Systems (NeurIPS)}
}

Other Attribution methods

The code for the metrics and the other attribution methods used in the paper come from the Xplique toolbox.

<a href="https://github.com/deel-ai/xplique"> <img src="./assets/banner.png" width="500px"> </a>

Authors