Question 3

Module 4. Linear Models Intuitions on linear models
#1

Is it possible to get a perfect fit (zero prediction error on the training set) with a linear classifier on a non-linearly separable dataset?

If we do a basis expansion on the features space we could get a linear classifier which achieve zero prediction error.

1 Like
#2

Indeed, but this is not discussed in the video or slide regarding the “intuitions on linear models”.

It is later introduced in the notebook: “Linear regression for a non-linear features-target relationship” with a section on feature expansion.