Linear regression learns a linear hypothesis map [math]\hat{h}[/math] having minimal average squared error on a training set. The learnt hypothesis [math]\hat{h}[/math] is then validated on a validation set which is different from the training set.

Can you construct a training set and validation set such that the validation error of [math]\hat{h}[/math] is strictly smaller than the training error of [math]\hat{h}[/math]?