MRCpy 0.1.3

Minimax Risk Classification

MRCpy: A Library for Minimax Risk Classifiers

MRCpy library implements minimax risk classifiers (MRCs) that are based on robust risk minimization and can utilize 0-1 loss, in contrast to existing libraries for supervised classification using techniques based on empirical risk minimization and surrogate losses.

Such techniques give rise to a manifold of classification methods that can provide tight bounds on the expected loss. MRCpy provides a unified interface for different variants of MRCs and follows the standards of popular Python libraries. This library also provides implementation for popular techniques that can be seen as MRCs such as L1-regularized logistic regression, zero-one adversarial, and maximum entropy machines.

In addition, MRCpy implements recent feature mappings such as Fourier, ReLU, and threshold features. The library is designed with an object-oriented approach that facilitates collaborators and users. The source code is available under the MIT license at https://github.com/MachineLearningBCAM/MRCpy.

IMPORTANT NOTE

When downloading the MRCpy library from PyPI (via pip), the MRCpy.datasets module (including the data loaders) are not downloaded due to space issues. If you want to obtain the datasets module, please download it from MRCpy github page.

Algorithms

• MRC with 0-1 loss (MRC)
• MRC with log loss (MRC)
• MRC with 0-1 loss and fixed instances' marginals (CMRC)
• MRC with log loss and fixed instances' marginals (CMRC)

Installation

From a terminal (OS X & linux), you can install MRCpy and its requirements directly by running the setup.py script as follows

git clone https://github.com/MachineLearningBCAM/MRCpy.git
cd MRCpy
python3 setup.py install

NOTE: CVXpy optimization uses MOSEK optimizer(by default) which requires a license. You can get a free academic license from here.

Dependencies

• Python >= 3.6
• numpy >= 1.18.1, scipy>= 1.4.1, scikit-learn >= 0.21.0, cvxpy, mosek, pandas

Usage

See the MRCpy documentation page for full documentation about installation, API, usage, and examples.

Citations

This repository is the official implementation of Minimax Risk Classifiers proposed in the following papers. If you use MRCpy in a scientific publication, we would appreciate citations to:

• [1] [Mazuelas, S., Zanoni, A., & Pérez, A. (2020). Minimax Classification with 0-1 Loss and Performance Guarantees. Advances in Neural Information Processing Systems, 33, 302-312.] (https://arxiv.org/abs/2010.07964)

    @article{mazuelas2020minimax,
    title={Minimax Classification with 0-1 Loss and Performance Guarantees},
    author={Mazuelas, Santiago and Zanoni, Andrea and P{\'e}rez, Aritz},
    journal={Advances in Neural Information Processing Systems},
    volume={33},
    pages={302--312},
    year={2020}
    }
  

• [2] Mazuelas, S., Shen, Y., & Pérez, A. (2020). Generalized Maximum Entropy for Supervised Classification. arXiv preprint arXiv:2007.05447.

    @article{mazuelas2020generalized,
    title={Generalized Maximum Entropy for Supervised Classification},
    author={Mazuelas, Santiago and Shen, Yuan and P{\'e}rez, Aritz},
    journal={arXiv preprint arXiv:2007.05447},
    year={2020}
    }
  

• [3] Bondugula, K., Mazuelas, S., & Pérez, A. (2021). MRCpy: A Library for Minimax Risk Classifiers. arXiv preprint arXiv:2108.01952.

    @article{bondugula2021mrcpy,
    title={MRCpy: A Library for Minimax Risk Classifiers},
    author={Bondugula, Kartheek and Mazuelas, Santiago and P{\'e}rez, Aritz},
    journal={arXiv preprint arXiv:2108.01952},
    year={2021}
    }
  

Updates and Discussion

You can subscribe to the MRCpy's mailing list for updates and discussion