A General Class of Transfer Learning Regression without Implementation Cost

We propose a novel framework that unifies and extends existing methods of transfer learning (TL) for regression. To bridge pretrained source model to the on target task, we introduce density-ratio reweighting function, which is estimated through Bayesian with specific prior distribution. By changing two intrinsic hyperparameters choice model, proposed method can integrate three popular TL: TL based cross-domain similarity regularization, probabilistic using estimation, fine-tuning neural networks. Moreover, benefit from its simple implementation without any additional cost; regression be fully trained off-the-shelf libraries supervised in original output variable simply transformed new variable. demonstrate simplicity, generality, applicability various real data applications.