Covariate-shift correction by adversarial-validation / propensity weighting toward the target distribution. Honest: helps under misspecification+shift, ties null when well-specified; never uses target y. Held-out + null validated. numpy-only.
machine-learning kaggle domain-adaptation covariate-shift adversarial-validation importance-weighting
-
Updated
Jun 27, 2026 - Python