Dear all,

Next week we will have Kirtan Padh https://kirtan.netlify.app/ as an invited speaker for the reading group. Kirtan is a PhD candidate at Helmholtz AI in Munich, advised by Prof. Niki Kilbertus (which some of us know from the MPI-IS times), and broadly speaking his work lays on the intersection of machine learning, causality, and fairness.

This was a talk scheduled quite a bit in advance by @Miriam, which will be our host next week. Unfortunately, Kirtan is not able to come in person and, therefore, the reading group *will be online next week.* You only need to access the _Zoom room located at the top of the shared spreadsheet_.

Please let me know if there is any doubts, and I hope to virtually see you all next week!

* *Speaker:* Kirtan Padh * *Title of the talk:* Addressing Fairness in Classification with a Model-Agnostic Multi-Objective Algorithm * *Link to the related paper:* https://proceedings.mlr.press/v161/padh21a/padh21a.pdf https://proceedings.mlr.press/v161/padh21a/padh21a.pdf? * *Abstract:* The goal of fairness in classification is to learn a classifier that does not discriminate against groups of individuals based on sensitive attributes, such as race and gender. One approach to designing fair algorithms is to use relaxations of fairness notions as regularization terms or in a constrained optimization problem. We observe that the hyperbolic tangent function can approximate the indicator function. We leverage this property to define a differentiable relaxation that approximates fairness notions provably better than existing relaxations. In addition, we propose a model-agnostic multi-objective architecture that can simultaneously optimize for multiple fairness notions and multiple sensitive attributes and supports all statistical parity-based notions of fairness. We use our relaxation with the multi-objective architecture to learn fair classifiers. Experiments on public datasets show that our method suffers a significantly lower loss of accuracy than current debiasing algorithms relative to the unconstrained model.

Cheers, Adrián