Libra

Nowadays we are witnessing widespread adoption of decision-making software with far-reaching societal impact, e.g., in fiels such as social welfare, criminal justice, and even health care. It is not difficult to envision that in the future most of the decisions in society will be delegated to software. As decision-making software becomes more and more widespread, a number of recent cases have evidenced the importance of ensuring software fairness^[1][2].

Even software products that we regularly use are biased!

Going forward, decision-making software will be subject to more and more legal regulations as well as administrative audits.

Libra

The goal of the Libra research project is to meet these needs and develop new analyses and tools to reason about and certify fairness of decision-making software.

A prototype static analyzer for certifying fairness of feed-forward neural networks used for classification of tabular data is open-source and available on GitHub.

Completed Projects

Abhinandan Pal (Bachelor Student, IIIT Kalyani, India)
GPU-based Fairness Certification of Neural Networks
Research Internship (remote), 2022
Marco Zanella (PhD Student, Università degli Studi di Padova, Italy)
Fairness of Decision Tree Ensemble Classifiers
PhD Internship, École Normale Supérieure, 2020

Publications

Abhinandan Pal, Francesco Ranzato, Caterina Urban, Marco Zanella. Abstract Interpretation-Based Feature Importance for Support Vector Machines. In VMCAI, 2024.

PDF Project HAL Springer

Abhinandan Pal, Francesco Ranzato, Caterina Urban, Marco Zanella. Abstract Interpretation-Based Feature Importance for SVMs. CoRR abs/2210.12456, 2022.

PDF Project arXiv

Serge Durand, Augustin Lemesle, Zakaria Chihani, Caterina Urban, François Terrier. ReCIPH: Relational Coefficients for Input Partitioning Heuristic. In WFVML, 2022.

Project

Francesco Ranzato, Caterina Urban, Marco Zanella. Fairness-Aware Training of Decision Trees by Abstract Interpretation. In CIKM, 2021.

PDF Code Project HAL ACM

Denis Mazzucato, Caterina Urban. Reduced Products of Abstract Domains for Fairness Certification of Neural Networks. In SAS, 2021.

PDF Code Project Slides Video Artifact HAL Springer

Francesco Ranzato, Caterina Urban, Marco Zanella. Fair Training of Decision Tree Classifiers. CoRR abs/2101.00909, 2021.

PDF Project arXiv

Caterina Urban, Maria Christakis, Valentin Wüstholz, Fuyuan Zhang. Perfectly Parallel Fairness Certification of Neural Networks. In OOPSLA, 2020.

PDF Code Project Slides Video Artifact BibTeX HAL ACM

Caterina Urban, Maria Christakis, Valentin Wüstholz, Fuyuan Zhang. Perfectly Parallel Fairness Certification of Neural Networks. CoRR abs/1912.02499, 2019.

PDF Code Project BibTeX arXiv