Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Inria, Soda team, CREST, ENSAE, IP Paris, Inria, Fairplay joint team
2
0
2
Choosing between Wasserstein and Total Variation penalties in fair regression isn't just a technical detail – it reflects fundamentally different fairness philosophies, with Wasserstein offering smooth compromises and Total Variation enforcing strict parity for select individuals.
Achieving counterfactual fairness in regression boils down to satisfying demographic parity conditioned on a latent variable, unlocking a surprisingly tractable solution via optimal transport.