From the 1 of 6 linked papers with an AI index.
6 papers
OT-FairBoost: Optimal Transport-Guided Gradient Boosting for Fairness Regularization on Tabular Data
Veronika Shilova, Abdoulaye Sakho, Younes Boumoussou +3
The paper proposes OT-FairBoost, an in-processing method that adds a Wasserstein-2 distance penalty to gradient-boosted tree training to improve group fairness while maintaining ac…
Discovering Geometric Biases in 3D Face Reconstruction: A Curvature-Aware Spectral Framework for Fairness Evaluation
Veronika Shilova, Emmanuel Malherbe, Giovanni Palma +3
3D Morphable Models (3DMMs) remain the standard parametric shape priors for many state-of-the-art 3D face reconstruction algorithms. However, as these models are derived from a fin…
Exposing the Illusion of Fairness: Auditing Vulnerabilities to Distributional Manipulation Attacks
Valentin Lafargue, Adriana Laurindo Monteiro, Emmanuelle Claeys +2
The rapid deployment of AI systems in high-stakes domains, including those classified as high-risk under the The EU AI Act (Regulation (EU) 2024/1689), has intensified the need for…
Explanation of Dynamic Physical Field Predictions using WassersteinGrad: Application to Autoregressive Weather Forecasting
Younes Essafouri, Laure Raynaud, Luciano Drozda +1
As the demand to integrate Artificial Intelligence into high-stakes environments continues to grow, explaining the reasoning behind neural-network predictions has shifted from a th…
Fairness-Aware Grouping for Continuous Sensitive Variables: Application for Debiasing Face Analysis with respect to Skin Tone
Veronika Shilova, Emmanuel Malherbe, Giovanni Palma +2
Within a legal framework, fairness in datasets and models is typically assessed by dividing observations into predefined groups and then computing fairness measures (e.g., Disparat…
TaCo: Targeted Concept Erasure Prevents Non-Linear Classifiers From Detecting Protected Attributes
Fanny Jourdan, Louis Béthune, Agustin Picard +2
Ensuring fairness in NLP models is crucial, as they often encode sensitive attributes like gender and ethnicity, leading to biased outcomes. Current concept erasure methods attempt…