27 February 2025
09:00 Master's Defense Room 85 of IC2
Topic
A Gradient Boosting Machine for Rawlsian Subgroup Fairness
Student
Jansen Silva de Brito Pereira
Advisor / Teacher
Marcos Medeiros Raimundo
Brief summary
In recent years, fairness in machine learning has emerged as a critical concern to ensure that predictive models developed and deployed do not yield disparate predictions for marginalized groups. However, it is essential to avoid biased decisions and promote equitable outcomes when simultaneously dealing with multiple (sub)group attributes (gender, race, etc.). In this work, we consider the application of the Rawlsian subgroup fairness concept to gradient boosting machines designed for supervised learning problems. Our approach expanded upon gradient boosting methodologies to explore a broader range of objective functions, which combine conventional losses, such as those from classification and regression problems, and a min-max fairness term. The optimization process explored primal-dual problems in each boosting round. This generic framework can be adapted to various fairness concepts. The proposed primal-dual min-max gradient boosting algorithm has been empirically shown to be a powerful and flexible approach to address binary and subgroup fairness.
Examination Board
Headlines:
| Marcos Medeiros Raimundo | IC / UNICAMP |
| Saullo Haniell Galvão de Oliveira | FAS/PUC-Campinas |
| Leonardo Tomazeli Duarte | FCA/UNICAMP |
Substitutes:
| Jacques Wainer | IC / UNICAMP |
| Fernando José Von Zuben | FEEC / UNICAMP |