Individual fairness guarantees for neural networks
We consider the problem of certifying the individual fairness (IF) of feed-forward neural networks (NNs). In particular, we work with the epsilon-delta-IF formulation, which, given a NN and a similarity metric learnt from data, requires that the output difference between any pair of epsilon-similar...
| Main Authors: | , , , , |
|---|---|
| Format: | Conference item |
| Language: | English |
| Published: |
International Joint Conferences on Artificial Intelligence
2022
|
| _version_ | 1797107438114045952 |
|---|---|
| author | Benussi, E Patane, A Wicker, M Laurenti, L Kwiatkowska, M |
| author_facet | Benussi, E Patane, A Wicker, M Laurenti, L Kwiatkowska, M |
| author_sort | Benussi, E |
| collection | OXFORD |
| description | We consider the problem of certifying the individual fairness (IF) of feed-forward neural networks (NNs). In particular, we work with the epsilon-delta-IF formulation, which, given a NN and a similarity metric learnt from data, requires that the output difference between any pair of epsilon-similar individuals is bounded by a maximum decision tolerance delta >= 0. Working with a range of metrics, including the Mahalanobis distance, we propose a method to overapproximate the resulting optimisation problem using piecewise-linear functions to lower and upper bound the NN's non-linearities globally over the input 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 fairness benchmarking. We show how this formulation can be used to encourage models' fairness at training 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. |
| first_indexed | 2024-03-07T07:16:05Z |
| format | Conference item |
| id | oxford-uuid:f1ceaf52-9ccd-436f-8816-37cf325f2025 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T07:16:05Z |
| publishDate | 2022 |
| publisher | International Joint Conferences on Artificial Intelligence |
| record_format | dspace |
| spelling | oxford-uuid:f1ceaf52-9ccd-436f-8816-37cf325f20252022-07-28T16:58:16ZIndividual fairness guarantees for neural networksConference itemhttp://purl.org/coar/resource_type/c_5794uuid:f1ceaf52-9ccd-436f-8816-37cf325f2025EnglishSymplectic ElementsInternational Joint Conferences on Artificial Intelligence2022Benussi, EPatane, AWicker, MLaurenti, LKwiatkowska, MWe consider the problem of certifying the individual fairness (IF) of feed-forward neural networks (NNs). In particular, we work with the epsilon-delta-IF formulation, which, given a NN and a similarity metric learnt from data, requires that the output difference between any pair of epsilon-similar individuals is bounded by a maximum decision tolerance delta >= 0. Working with a range of metrics, including the Mahalanobis distance, we propose a method to overapproximate the resulting optimisation problem using piecewise-linear functions to lower and upper bound the NN's non-linearities globally over the input 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 fairness benchmarking. We show how this formulation can be used to encourage models' fairness at training 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. |
| spellingShingle | Benussi, E Patane, A Wicker, M Laurenti, L Kwiatkowska, M Individual fairness guarantees for neural networks |
| title | Individual fairness guarantees for neural networks |
| title_full | Individual fairness guarantees for neural networks |
| title_fullStr | Individual fairness guarantees for neural networks |
| title_full_unstemmed | Individual fairness guarantees for neural networks |
| title_short | Individual fairness guarantees for neural networks |
| title_sort | individual fairness guarantees for neural networks |
| work_keys_str_mv | AT benussie individualfairnessguaranteesforneuralnetworks AT patanea individualfairnessguaranteesforneuralnetworks AT wickerm individualfairnessguaranteesforneuralnetworks AT laurentil individualfairnessguaranteesforneuralnetworks AT kwiatkowskam individualfairnessguaranteesforneuralnetworks |