Conformal prediction coverage gaps under joint distribution shift can be upper bounded by Wasserstein distance and reduced by importance weighting plus a Wasserstein-regularized training objective.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Wasserstein-regularized Conformal Prediction under General Distribution Shift
Conformal prediction coverage gaps under joint distribution shift can be upper bounded by Wasserstein distance and reduced by importance weighting plus a Wasserstein-regularized training objective.