RT journal article
T1 Fair Logistic Regression via Penalized Mean-Field Variational Inference
A1 Carrizosa, Emilio
A1 Jiménez-Llamas, Rafael
A1 Ramírez Cobo, Josefa
A2 Estadística e Investigación Operativa
K1 Fairness
K1 Variational Bayes
K1 Binary Classification
K1 Coordinate Ascent Variational Inference
AB Fairness in Machine Learning (ML) has recently risen in importance due to itsdemand by both society and institutions, either private or public. This is due tothe discrimination or disparities that may arise as a consequence of biases in thedata sets that are used for training the prediction algorithms. As a consequence,the applicability of such algorithms may be limited. In this paper we propose a fairversion of the logistic regression via Mean-Field Variational Bayes (MFVB). Fairnessis defined in terms of a carefully selected metric that, not only is interpretable for theconsidered logistic regression problem, but also leads to tractable MFVB equations.In addition, it allows for the problem to be solved in a private manner, furtherincreasing the applicability of the method. The MFVB equations are solved througha Coordinate Ascent Variational Inference (CAVI) algorithm, where a proper selectionof the hyper-parameters is shown to be crucial to assure accurate classification rates.Numerical experiments illustrate how the novel approach provides a trade-off betweenthe accuracy of the method and the predictions’ level of fairness that eases the decision making process.
YR 2025
FD 2025-08
LK http://hdl.handle.net/10498/38832
UL http://hdl.handle.net/10498/38832
LA eng
DS Repositorio Institucional de la Universidad de Cádiz
RD 07-nov-2026