Assuming we are using non-parametric models like gradient boosted tree, can retraining the model solves each of the dataset shift (1. covariate 2. prior probability 3. concept shift)?
$\begingroup$
$\endgroup$
