Welcome to Quantus documentation!

Quantus is an eXplainable AI toolkit for responsible evaluation of neural network explanations.

Figure: a) Simple qualitative comparison of XAI methods is often not sufficient to distinguish which gradient-based method — Saliency, Integrated Gradients, GradientShap or FusionGrad is preferred. With Quantus, we can obtain richer insights on how the methods compare b) by holistic quantification on several evaluation criteria and c) by providing sensitivity analysis of how a single parameter, e.g., pixel replacement strategy of a faithfulness test influences the ranking of XAI methods. 📑 Shortcut to paper!

This documentation is complementary to the README.md in the Quantus repository and provides documentation for how to install Quantus, how to contribute and details on the API. For further guidance on what to think about when applying Quantus, please read the user guidelines. Do you want to get started? Please have a look at our simple toy example with PyTorch using MNIST data. For more examples, check the tutorials folder.

Quantus can be installed from PyPI (this way assumes that you have either PyTorch or Tensorflow installed on your machine):

pip install quantus

For a more in-depth guide on how to install Quantus, read more here. This includes instructions for how to install a desired deep learning framework such as PyTorch or tensorflow together with Quantus.

Contents

Installation

• Quick installation

Getting Started

• Getting Started

API Reference

• quantus package

Developer Documentation

• Contribute to Quantus

Guidelines

• User guidelines
• Disclaimers

Citation

If you find this toolkit or its companion paper Quantus: An Explainable AI Toolkit for Responsible Evaluation of Neural Network Explanations interesting or useful in your research, please use the following Bibtex annotation to cite us:

@article{hedstrom2023quantus,
  author  = {Anna Hedstr{\"{o}}m and Leander Weber and Daniel Krakowczyk and Dilyara Bareeva and Franz Motzkus and Wojciech Samek and Sebastian Lapuschkin and Marina Marina M.{-}C. H{\"{o}}hne},
  title   = {Quantus: An Explainable AI Toolkit for Responsible Evaluation of Neural Network Explanations and Beyond},
  journal = {Journal of Machine Learning Research},
  year    = {2023},
  volume  = {24},
  number  = {34},
  pages   = {1--11},
  url     = {http://jmlr.org/papers/v24/22-0142.html}
}

When applying the individual metrics of Quantus, please make sure to also properly cite the work of the original authors. You can find the relevant citations in the documentation of each respective metric here.