Oxford logo
[BPW+22] Elias Benussi, Andrea Patane, Matthew Wicker, Luca Laurenti and Marta Kwiatkowska. Individual Fairness Guarantees for Neural Networks. In Proc. 31st International Joint Conference on Artificial Intelligence (IJCAI'22). To appear. July 2022. [pdf] [bib]
Downloads:  pdf pdf (2.85 MB)  bib bib
Abstract. We consider the problem of certifying the individual fairness (IF) of feed-forward neural networks (NNs). In particular, we work with the ε-δ-IF formulation, which, given a NN and a similarity metric learnt from data, requires that the output difference between any pair of ε-similar individuals is bounded by a maximum decision tolerance δ ≥ 0. Working with a range of metrics, including the Mahalanobis distance, we propose a method to over- approximate the resulting optimisation problem us- ing piecewise-linear functions to lower and upper bound the NN’s non-linearities globally over the in- put space. We encode this computation as the solution of a Mixed-Integer Linear Programming problem and demonstrate that it can be used to compute IF guarantees on four datasets widely used for fair- ness benchmarking. We show how this formulation can be used to encourage models’ fairness at train- ing time by modifying the NN loss, and empirically confirm our approach yields NNs that are orders of magnitude fairer than state-of-the-art methods.