… a novel algorithm for learning fair representations that can … error rate and conditional alignment
of representations. We show … algorithms on learning fair representations for classification. …

… also considers conditional mutual information and conditional … , and explicitly samples from
conditional distributions (rather … We did not include supervised fair representation models, as …

… in fair machine learning typically focuses on enforcing independence on lowdimensional
features in cases where the conditioning set … This restricts the learning of fair representations to …

… In light of this challenge, in this work, we study learning fair representations that … learning.
Specifically, we first theoretically analyze the connections between learning representations …

… , which incorporates auxiliary information by conditioning on the … task, fair contrastive
learning , we condition on the sensitive attribute to remove its information in the representations. …

… On the upside, we prove that if the group-wise Bayes optimal classifiers are close, then
learning fair representations leads to an alternative notion of fairness, known as the accuracy …

… downstream vendor cannot learn a classifier over the representations z that discriminates …
information as possible from x conditioned on the knowledge of u. By conditioning on u, we …

… As fair representation learning is a common in-processing framework to deal with fairness
issues, we apply the regularizer and handle the conditional fairness constraint in this …

… Flexibly Fair VAE We want to learn fair representations that—beyond being useful for predicting
many test-time task labels y—can be adapted simply and compositionally for a variety of …

… this problem, including fair representation learning, which cleans … fair representation learning
as an intermediate step in fair … We show that fair representation learning in fact addresses a …