A General Framework for Fair Regression
This work addresses fairness in machine learning for regression tasks, offering a general framework applicable to existing models, though it appears incremental as it extends fairness constraints to various methods without a major paradigm shift.
The paper tackles the problem of incorporating group fairness constraints into kernel regression methods, including Gaussian processes, support vector machines, neural networks, and decision trees, and shows that the approach preserves computational and memory complexity while bounding perturbations based on tree leaves.
Fairness, through its many forms and definitions, has become an important issue facing the machine learning community. In this work, we consider how to incorporate group fairness constraints in kernel regression methods, applicable to Gaussian processes, support vector machines, neural network regression and decision tree regression. Further, we focus on examining the effect of incorporating these constraints in decision tree regression, with direct applications to random forests and boosted trees amongst other widespread popular inference techniques. We show that the order of complexity of memory and computation is preserved for such models and tightly bound the expected perturbations to the model in terms of the number of leaves of the trees. Importantly, the approach works on trained models and hence can be easily applied to models in current use and group labels are only required on training data.