Out-of-Distribution Generalization in Kernel Regression

In real word applications, the data generating process for training a machine learning model often differs from what the model encounters in the test stage.