Séminaire de Statistique et Optimisation

Adaptive Conformal Inference through the Lens of Blackwell Approachability

par Guillaume Principato (Laboratoire de Mathématiques d'Orsay)

Europe/Paris
Salle K. Johnson (1R3, 1er étage)

Salle K. Johnson

1R3, 1er étage

Description
We study conformal inference in non-exchangeable environments through the lens of Blackwell’s theory of approachability. We first recast adaptive conformal inference (ACI, Gibbs and Candès, 2021) as a repeated two-player vector-valued finite game and characterize attainable coverage– efficiency tradeoffs. We then construct coverage and efficiency objectives under potential restrictions on the adversary’s play, and design a calibration-based approachability strategy to achieve these goals.