The Power and Limitation of Pretraining-Finetuning for Linear Regression under Covariate Shift

Open in new window