Fair machine learning models: estimation, tuning and prediction.
Marco Scutari
Istituto Dalle Molle di Studi sull'Intelligenza Artificiale (IDSIA)
Maintainer: Marco Scutari scutari@bnlearn.com
fairml implements key algorithms for learning machine learning models while enforcing fairness with respect to a set of observed sensitive (or protected) attributes.
Currently fairml implements the following algorithms (references below):
nclm(): the non-convex formulation of fair linear regression
model from Komiyama et al. (2018).
frrm(): the fair (linear) ridge regression model from Scutari,
Panero and Proissl (2022).
fgrrm(): thefair generalized (linear) ridge regression model
from Scutari, Panero and Proissl (2022), supporting the Gaussian,
binomial, Poisson, multinomial and Cox (proportional hazards) families.
zlrm(): the fair logistic regression with covariance
constraints from Zafar et al. (2019).
zlrm(): a fair linear regression with covariance
constraints following Zafar et al. (2019).
Furthermore, different fairness definitions can be used in frrm()
and fgrrm():
"sp-komiyama": the statistical parity fairness constraint from
Komiyama et al. (2018);
"eo-komiyama": the analogous equality of opportunity constraint
built on the proportion of variance (or deviance) explained by sensitive
attributes;
"if-berk": the individual fairness constraint from Berk et al.
(2017) adapted in Scutari, Panero and Proissl (2022);
user-provided functions for custom definitions.
In addition, fairml implements diagnostic plots, cross-validation,
prediction and methods for most of the generics made available for linear
models from lm() and glm(). Profile plots to trace key model
and goodness-of-fit indicators at varying levels of fairness are available
from
fairness.profile.plot().
Berk R, Heidari H, Jabbari S, Joseph M, Kearns M, Morgenstern J, Neel S,
Roth A (2017). "A Convex Framework for Fair Regression". FATML.
https://www.fatml.org/media/documents/convex_framework_for_fair_regression.pdf
Komiyama J, Takeda A, Honda J, Shimao H (2018). "Nonconvex Optimization for
Regression with Fairness Constraints". Proceedings of the 35th International
Conference on Machine Learning (ICML), PMLR 80:2737--2746.
http://proceedings.mlr.press/v80/komiyama18a/komiyama18a.pdf
Scutari M, Panero F, Proissl M (2022). "Achieving Fairness with a Simple Ridge
Penalty". Statistics and Computing, 32, 77.
https://link.springer.com/content/pdf/10.1007/s11222-022-10143-w.pdf
Zafar BJ, Valera I, Gomez-Rodriguez M, Gummadi KP (2019). "Fairness
Constraints: a Flexible Approach for Fair Classification". Journal of
Machine Learning Research, 30:1--42.
https://www.jmlr.org/papers/volume20/18-262/18-262.pdf